blob: e901d7eeafd8ff00e0da25b30754cab5c2ff8007 [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 Tarreaub3066502017-11-26 19:50:17 +01005 version 1.9
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreau0b787922017-11-26 19:25:23 +01007 2017/11/26
Willy Tarreau6a06a402007-07-15 20:15:28 +02008
9
10This document covers the configuration language as implemented in the version
Davor Ocelice9ed2812017-12-25 17:49:28 +010011specified above. It does not provide any hints, examples, or advice. For such
Willy Tarreau0ba27502007-12-24 16:55:16 +010012documentation, please refer to the Reference Manual or the Architecture Manual.
Davor Ocelice9ed2812017-12-25 17:49:28 +010013The summary below is meant to help you find sections by name and navigate
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014through 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
Davor Ocelice9ed2812017-12-25 17:49:28 +010022 sometimes useful to prefix all output lines (logs, console outputs) with 3
23 closing angle brackets ('>>>') in order to emphasize the difference between
24 inputs and outputs when they may be ambiguous. If you add sections,
Willy Tarreau62a36c42010-08-17 15:53:10 +020025 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
Davor Ocelice9ed2812017-12-25 17:49:28 +0100341.2.1. The request line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200351.2.2. The request headers
361.3. HTTP response
Davor Ocelice9ed2812017-12-25 17:49:28 +0100371.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200381.3.2. 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
Davor Ocelice9ed2812017-12-25 17:49:28 +0100595. Bind and server options
Willy Tarreau086fbf52012-09-24 20:34:51 +0200605.1. Bind options
615.2. Server and default-server options
Baptiste Assmann1fa66662015-04-14 00:28:47 +0200625.3. Server DNS resolution
635.3.1. Global overview
645.3.2. The resolvers section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020065
666. HTTP header manipulation
67
Willy Tarreau74ca5042013-06-11 23:12:07 +0200687. Using ACLs and fetching samples
697.1. ACL basics
707.1.1. Matching booleans
717.1.2. Matching integers
727.1.3. Matching strings
737.1.4. Matching regular expressions (regexes)
747.1.5. Matching arbitrary data blocks
757.1.6. Matching IPv4 and IPv6 addresses
767.2. Using ACLs to form conditions
777.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200787.3.1. Converters
797.3.2. Fetching samples from internal states
807.3.3. Fetching samples at Layer 4
817.3.4. Fetching samples at Layer 5
827.3.5. Fetching samples from buffer contents (Layer 6)
837.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200847.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020085
868. Logging
878.1. Log levels
888.2. Log formats
898.2.1. Default log format
908.2.2. TCP log format
918.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100928.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100938.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200948.3. Advanced logging options
958.3.1. Disabling logging of external tests
968.3.2. Logging before waiting for the session to terminate
978.3.3. Raising log level upon errors
988.3.4. Disabling logging of successful connections
998.4. Timing events
1008.5. Session state at disconnection
1018.6. Non-printable characters
1028.7. Capturing HTTP cookies
1038.8. Capturing HTTP headers
1048.9. Examples of logs
105
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001069. Supported filters
1079.1. Trace
1089.2. HTTP compression
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +02001099.3. Stream Processing Offload Engine (SPOE)
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200110
William Lallemand86d0df02017-11-24 21:36:45 +010011110. Cache
Cyril Bonté7b888f12017-11-26 22:24:31 +010011210.1. Limitation
11310.2. Setup
11410.2.1. Cache section
11510.2.2. Proxy section
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200116
1171. Quick reminder about HTTP
118----------------------------
119
Davor Ocelice9ed2812017-12-25 17:49:28 +0100120When HAProxy is running in HTTP mode, both the request and the response are
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200121fully analyzed and indexed, thus it becomes possible to build matching criteria
122on almost anything found in the contents.
123
124However, it is important to understand how HTTP requests and responses are
125formed, and how HAProxy decomposes them. It will then become easier to write
126correct rules and to debug existing configurations.
127
128
1291.1. The HTTP transaction model
130-------------------------------
131
132The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100133to one and only one response. Traditionally, a TCP connection is established
Davor Ocelice9ed2812017-12-25 17:49:28 +0100134from the client to the server, a request is sent by the client through the
135connection, the server responds, and the connection is closed. A new request
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200136will involve a new connection :
137
138 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
139
140In this mode, called the "HTTP close" mode, there are as many connection
141establishments as there are HTTP transactions. Since the connection is closed
142by the server after the response, the client does not need to know the content
143length.
144
145Due to the transactional nature of the protocol, it was possible to improve it
146to avoid closing a connection between two subsequent transactions. In this mode
147however, it is mandatory that the server indicates the content length for each
148response so that the client does not wait indefinitely. For this, a special
149header is used: "Content-length". This mode is called the "keep-alive" mode :
150
151 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
152
153Its advantages are a reduced latency between transactions, and less processing
154power required on the server side. It is generally better than the close mode,
155but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200156a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200157
Willy Tarreau95c4e142017-11-26 12:18:55 +0100158Another improvement in the communications is the pipelining mode. It still uses
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200159keep-alive, but the client does not wait for the first response to send the
160second request. This is useful for fetching large number of images composing a
161page :
162
163 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
164
165This can obviously have a tremendous benefit on performance because the network
166latency is eliminated between subsequent requests. Many HTTP agents do not
167correctly support pipelining since there is no way to associate a response with
168the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100169server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170
Willy Tarreau95c4e142017-11-26 12:18:55 +0100171The next improvement is the multiplexed mode, as implemented in HTTP/2. This
172time, each transaction is assigned a single stream identifier, and all streams
173are multiplexed over an existing connection. Many requests can be sent in
174parallel by the client, and responses can arrive in any order since they also
175carry the stream identifier.
176
Willy Tarreau70dffda2014-01-30 03:07:23 +0100177By default HAProxy operates in keep-alive mode with regards to persistent
178connections: for each connection it processes each request and response, and
179leaves the connection idle on both sides between the end of a response and the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100180start of a new request. When it receives HTTP/2 connections from a client, it
181processes all the requests in parallel and leaves the connection idling,
182waiting for new requests, just as if it was a keep-alive HTTP connection.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200183
Willy Tarreau70dffda2014-01-30 03:07:23 +0100184HAProxy supports 5 connection modes :
185 - keep alive : all requests and responses are processed (default)
186 - tunnel : only the first request and response are processed,
187 everything else is forwarded with no analysis.
188 - passive close : tunnel with "Connection: close" added in both directions.
189 - server close : the server-facing connection is closed after the response.
190 - forced close : the connection is actively closed after end of response.
191
Davor Ocelice9ed2812017-12-25 17:49:28 +0100192For HTTP/2, the connection mode resembles more the "server close" mode : given
193the independence of all streams, there is currently no place to hook the idle
Willy Tarreau95c4e142017-11-26 12:18:55 +0100194server connection after a response, so it is closed after the response. HTTP/2
195is only supported for incoming connections, not on connections going to
196servers.
197
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200198
1991.2. HTTP request
200-----------------
201
202First, let's consider this HTTP request :
203
204 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100205 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200206 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
207 2 Host: www.mydomain.com
208 3 User-agent: my small browser
209 4 Accept: image/jpeg, image/gif
210 5 Accept: image/png
211
212
2131.2.1. The Request line
214-----------------------
215
216Line 1 is the "request line". It is always composed of 3 fields :
217
218 - a METHOD : GET
219 - a URI : /serv/login.php?lang=en&profile=2
220 - a version tag : HTTP/1.1
221
222All of them are delimited by what the standard calls LWS (linear white spaces),
223which are commonly spaces, but can also be tabs or line feeds/carriage returns
224followed by spaces/tabs. The method itself cannot contain any colon (':') and
225is limited to alphabetic letters. All those various combinations make it
226desirable that HAProxy performs the splitting itself rather than leaving it to
227the user to write a complex or inaccurate regular expression.
228
229The URI itself can have several forms :
230
231 - A "relative URI" :
232
233 /serv/login.php?lang=en&profile=2
234
235 It is a complete URL without the host part. This is generally what is
236 received by servers, reverse proxies and transparent proxies.
237
238 - An "absolute URI", also called a "URL" :
239
240 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
241
242 It is composed of a "scheme" (the protocol name followed by '://'), a host
243 name or address, optionally a colon (':') followed by a port number, then
244 a relative URI beginning at the first slash ('/') after the address part.
245 This is generally what proxies receive, but a server supporting HTTP/1.1
246 must accept this form too.
247
248 - a star ('*') : this form is only accepted in association with the OPTIONS
249 method and is not relayable. It is used to inquiry a next hop's
250 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100251
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200252 - an address:port combination : 192.168.0.12:80
253 This is used with the CONNECT method, which is used to establish TCP
254 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
255 other protocols too.
256
257In a relative URI, two sub-parts are identified. The part before the question
258mark is called the "path". It is typically the relative path to static objects
259on the server. The part after the question mark is called the "query string".
260It is mostly used with GET requests sent to dynamic scripts and is very
261specific to the language, framework or application in use.
262
Willy Tarreau95c4e142017-11-26 12:18:55 +0100263HTTP/2 doesn't convey a version information with the request, so the version is
Davor Ocelice9ed2812017-12-25 17:49:28 +0100264assumed to be the same as the one of the underlying protocol (i.e. "HTTP/2").
Willy Tarreau95c4e142017-11-26 12:18:55 +0100265However, haproxy natively processes HTTP/1.x requests and headers, so requests
266received over an HTTP/2 connection are transcoded to HTTP/1.1 before being
267processed. This explains why they still appear as "HTTP/1.1" in haproxy's logs
268as well as in server logs.
269
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200270
2711.2.2. The request headers
272--------------------------
273
274The headers start at the second line. They are composed of a name at the
275beginning of the line, immediately followed by a colon (':'). Traditionally,
276an LWS is added after the colon but that's not required. Then come the values.
277Multiple identical headers may be folded into one single line, delimiting the
278values with commas, provided that their order is respected. This is commonly
279encountered in the "Cookie:" field. A header may span over multiple lines if
280the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
281define a total of 3 values for the "Accept:" header.
282
Davor Ocelice9ed2812017-12-25 17:49:28 +0100283Contrary to a common misconception, header names are not case-sensitive, and
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200284their values are not either if they refer to other header names (such as the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100285"Connection:" header). In HTTP/2, header names are always sent in lower case,
286as can be seen when running in debug mode.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200287
288The end of the headers is indicated by the first empty line. People often say
289that it's a double line feed, which is not exact, even if a double line feed
290is one valid form of empty line.
291
292Fortunately, HAProxy takes care of all these complex combinations when indexing
293headers, checking values and counting them, so there is no reason to worry
294about the way they could be written, but it is important not to accuse an
295application of being buggy if it does unusual, valid things.
296
297Important note:
Lukas Tribus23953682017-04-28 13:24:30 +0000298 As suggested by RFC7231, HAProxy normalizes headers by replacing line breaks
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200299 in the middle of headers by LWS in order to join multi-line headers. This
300 is necessary for proper analysis and helps less capable HTTP parsers to work
301 correctly and not to be fooled by such complex constructs.
302
303
3041.3. HTTP response
305------------------
306
307An HTTP response looks very much like an HTTP request. Both are called HTTP
308messages. Let's consider this HTTP response :
309
310 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100311 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200312 1 HTTP/1.1 200 OK
313 2 Content-length: 350
314 3 Content-Type: text/html
315
Willy Tarreau816b9792009-09-15 21:25:21 +0200316As a special case, HTTP supports so called "Informational responses" as status
317codes 1xx. These messages are special in that they don't convey any part of the
318response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100319continue to post its request for instance. In the case of a status 100 response
320the requested information will be carried by the next non-100 response message
321following the informational one. This implies that multiple responses may be
322sent to a single request, and that this only works when keep-alive is enabled
323(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
324correctly forward and skip them, and only process the next non-100 response. As
325such, these messages are neither logged nor transformed, unless explicitly
326state otherwise. Status 101 messages indicate that the protocol is changing
327over the same connection and that haproxy must switch to tunnel mode, just as
328if a CONNECT had occurred. Then the Upgrade header would contain additional
329information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200330
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200331
Davor Ocelice9ed2812017-12-25 17:49:28 +01003321.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200333------------------------
334
335Line 1 is the "response line". It is always composed of 3 fields :
336
337 - a version tag : HTTP/1.1
338 - a status code : 200
339 - a reason : OK
340
341The status code is always 3-digit. The first digit indicates a general status :
Davor Ocelice9ed2812017-12-25 17:49:28 +0100342 - 1xx = informational message to be skipped (e.g. 100, 101)
343 - 2xx = OK, content is following (e.g. 200, 206)
344 - 3xx = OK, no content following (e.g. 302, 304)
345 - 4xx = error caused by the client (e.g. 401, 403, 404)
346 - 5xx = error caused by the server (e.g. 500, 502, 503)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200347
Lukas Tribus23953682017-04-28 13:24:30 +0000348Please refer to RFC7231 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100349"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200350found there, but it's a common practice to respect the well-established
351messages. It can be composed of one or multiple words, such as "OK", "Found",
352or "Authentication Required".
353
Davor Ocelice9ed2812017-12-25 17:49:28 +0100354HAProxy may emit the following status codes by itself :
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200355
356 Code When / reason
357 200 access to stats page, and when replying to monitoring requests
358 301 when performing a redirection, depending on the configured code
359 302 when performing a redirection, depending on the configured code
360 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100361 307 when performing a redirection, depending on the configured code
362 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200363 400 for an invalid or too large request
364 401 when an authentication is required to perform the action (when
365 accessing the stats page)
366 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
367 408 when the request timeout strikes before the request is complete
368 500 when haproxy encounters an unrecoverable internal error, such as a
369 memory allocation failure, which should never happen
370 502 when the server returns an empty, invalid or incomplete response, or
371 when an "rspdeny" filter blocks the response.
372 503 when no server was available to handle the request, or in response to
373 monitoring requests which match the "monitor fail" condition
374 504 when the response timeout strikes before the server responds
375
376The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3774.2).
378
379
3801.3.2. The response headers
381---------------------------
382
383Response headers work exactly like request headers, and as such, HAProxy uses
384the same parsing function for both. Please refer to paragraph 1.2.2 for more
385details.
386
387
3882. Configuring HAProxy
389----------------------
390
3912.1. Configuration file format
392------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200393
394HAProxy's configuration process involves 3 major sources of parameters :
395
396 - the arguments from the command-line, which always take precedence
397 - the "global" section, which sets process-wide parameters
398 - the proxies sections which can take form of "defaults", "listen",
399 "frontend" and "backend".
400
Willy Tarreau0ba27502007-12-24 16:55:16 +0100401The configuration file syntax consists in lines beginning with a keyword
402referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200403delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100404
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200405
William Lallemandf9873ba2015-05-05 17:37:14 +02004062.2. Quoting and escaping
407-------------------------
408
409HAProxy's configuration introduces a quoting and escaping system similar to
410many programming languages. The configuration file supports 3 types: escaping
411with a backslash, weak quoting with double quotes, and strong quoting with
412single quotes.
413
414If spaces have to be entered in strings, then they must be escaped by preceding
415them by a backslash ('\') or by quoting them. Backslashes also have to be
416escaped by doubling or strong quoting them.
417
418Escaping is achieved by preceding a special character by a backslash ('\'):
419
420 \ to mark a space and differentiate it from a delimiter
421 \# to mark a hash and differentiate it from a comment
422 \\ to use a backslash
423 \' to use a single quote and differentiate it from strong quoting
424 \" to use a double quote and differentiate it from weak quoting
425
426Weak quoting is achieved by using double quotes (""). Weak quoting prevents
427the interpretation of:
428
429 space as a parameter separator
430 ' single quote as a strong quoting delimiter
431 # hash as a comment start
432
William Lallemandb2f07452015-05-12 14:27:13 +0200433Weak quoting permits the interpretation of variables, if you want to use a non
434-interpreted dollar within a double quoted string, you should escape it with a
435backslash ("\$"), it does not work outside weak quoting.
436
437Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200438quoting.
439
440Strong quoting is achieved by using single quotes (''). Inside single quotes,
441nothing is interpreted, it's the efficient way to quote regexes.
442
443Quoted and escaped strings are replaced in memory by their interpreted
444equivalent, it allows you to perform concatenation.
445
446 Example:
447 # those are equivalents:
448 log-format %{+Q}o\ %t\ %s\ %{-Q}r
449 log-format "%{+Q}o %t %s %{-Q}r"
450 log-format '%{+Q}o %t %s %{-Q}r'
451 log-format "%{+Q}o %t"' %s %{-Q}r'
452 log-format "%{+Q}o %t"' %s'\ %{-Q}r
453
454 # those are equivalents:
455 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
456 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
457 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
458 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
459
460
William Lallemandb2f07452015-05-12 14:27:13 +02004612.3. Environment variables
462--------------------------
463
464HAProxy's configuration supports environment variables. Those variables are
465interpreted only within double quotes. Variables are expanded during the
466configuration parsing. Variable names must be preceded by a dollar ("$") and
467optionally enclosed with braces ("{}") similarly to what is done in Bourne
468shell. Variable names can contain alphanumerical characters or the character
469underscore ("_") but should not start with a digit.
470
471 Example:
472
473 bind "fd@${FD_APP1}"
474
475 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
476
477 user "$HAPROXY_USER"
478
William Lallemanddaf4cd22018-04-17 16:46:13 +0200479A special variable $HAPROXY_LOCALPEER is defined at the startup of the process
480which contains the name of the local peer. (See "-L" in the management guide.)
481
William Lallemandb2f07452015-05-12 14:27:13 +0200482
4832.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200484----------------
485
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100486Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100487values are generally expressed in milliseconds (unless explicitly stated
488otherwise) but may be expressed in any other unit by suffixing the unit to the
489numeric value. It is important to consider this because it will not be repeated
490for every keyword. Supported units are :
491
492 - us : microseconds. 1 microsecond = 1/1000000 second
493 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
494 - s : seconds. 1s = 1000ms
495 - m : minutes. 1m = 60s = 60000ms
496 - h : hours. 1h = 60m = 3600s = 3600000ms
497 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
498
499
Lukas Tribusaa83a312017-03-21 09:25:09 +00005002.5. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200501-------------
502
503 # Simple configuration for an HTTP proxy listening on port 80 on all
504 # interfaces and forwarding requests to a single backend "servers" with a
505 # single server "server1" listening on 127.0.0.1:8000
506 global
507 daemon
508 maxconn 256
509
510 defaults
511 mode http
512 timeout connect 5000ms
513 timeout client 50000ms
514 timeout server 50000ms
515
516 frontend http-in
517 bind *:80
518 default_backend servers
519
520 backend servers
521 server server1 127.0.0.1:8000 maxconn 32
522
523
524 # The same configuration defined with a single listen block. Shorter but
525 # less expressive, especially in HTTP mode.
526 global
527 daemon
528 maxconn 256
529
530 defaults
531 mode http
532 timeout connect 5000ms
533 timeout client 50000ms
534 timeout server 50000ms
535
536 listen http-in
537 bind *:80
538 server server1 127.0.0.1:8000 maxconn 32
539
540
541Assuming haproxy is in $PATH, test these configurations in a shell with:
542
Willy Tarreauccb289d2010-12-11 20:19:38 +0100543 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200544
545
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005463. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200547--------------------
548
549Parameters in the "global" section are process-wide and often OS-specific. They
550are generally set once for all and do not need being changed once correct. Some
551of them have command-line equivalents.
552
553The following keywords are supported in the "global" section :
554
555 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200556 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200557 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200558 - crt-base
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200559 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200560 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200561 - description
562 - deviceatlas-json-file
563 - deviceatlas-log-level
564 - deviceatlas-separator
565 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900566 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200567 - gid
568 - group
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100569 - hard-stop-after
Willy Tarreau6a06a402007-07-15 20:15:28 +0200570 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200571 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100572 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200573 - lua-load
Willy Tarreau6a06a402007-07-15 20:15:28 +0200574 - nbproc
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200575 - nbthread
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200576 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200577 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100578 - presetenv
579 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200580 - uid
581 - ulimit-n
582 - user
Willy Tarreau1d549722016-02-16 12:41:57 +0100583 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200584 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200585 - ssl-default-bind-ciphers
586 - ssl-default-bind-options
587 - ssl-default-server-ciphers
588 - ssl-default-server-options
589 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100590 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100591 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100592 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100593 - 51degrees-data-file
594 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200595 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200596 - 51degrees-cache-size
scientiamobiled0027ed2016-11-04 10:55:08 +0100597 - wurfl-data-file
598 - wurfl-information-list
599 - wurfl-information-list-separator
600 - wurfl-engine-mode
601 - wurfl-cache-size
602 - wurfl-useragent-priority
Willy Tarreaud72758d2010-01-12 10:42:19 +0100603
Willy Tarreau6a06a402007-07-15 20:15:28 +0200604 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200605 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200606 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200607 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100608 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100609 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100610 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200611 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200612 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200613 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200614 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200615 - noepoll
616 - nokqueue
617 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100618 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300619 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +0000620 - noreuseport
Willy Tarreaufe255b72007-10-14 23:09:26 +0200621 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200622 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200623 - server-state-file
Grant Zhang872f9c22017-01-21 01:10:18 +0000624 - ssl-engine
Grant Zhangfa6c7ee2017-01-14 01:42:15 +0000625 - ssl-mode-async
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200626 - tune.buffers.limit
627 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200628 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200629 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100630 - tune.comp.maxlevel
Willy Tarreaufe20e5b2017-07-27 11:42:14 +0200631 - tune.h2.header-table-size
Willy Tarreaue6baec02017-07-27 11:45:11 +0200632 - tune.h2.initial-window-size
Willy Tarreau5242ef82017-07-27 11:47:28 +0200633 - tune.h2.max-concurrent-streams
Willy Tarreau193b8c62012-11-22 00:17:38 +0100634 - tune.http.cookielen
Stéphane Cottin23e9e932017-05-18 08:58:41 +0200635 - tune.http.logurilen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200636 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100637 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100638 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100639 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100640 - tune.lua.session-timeout
641 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200642 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100643 - tune.maxaccept
644 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200645 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200646 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200647 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100648 - tune.rcvbuf.client
649 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100650 - tune.recv_enough
Olivier Houchard1599b802018-05-24 18:59:04 +0200651 - tune.runqueue-depth
Willy Tarreaue803de22010-01-21 17:43:04 +0100652 - tune.sndbuf.client
653 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100654 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100655 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200656 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100657 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200658 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200659 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER5bf77322017-02-25 12:45:22 +0100660 - tune.ssl.capture-cipherlist-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200661 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +0100662 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200663 - tune.vars.reqres-max-size
664 - tune.vars.sess-max-size
665 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100666 - tune.zlib.memlevel
667 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100668
Willy Tarreau6a06a402007-07-15 20:15:28 +0200669 * Debugging
670 - debug
671 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200672
673
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006743.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200675------------------------------------
676
Emeric Brunc8e8d122012-10-02 18:42:10 +0200677ca-base <dir>
678 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200679 relative path is used with "ca-file" or "crl-file" directives. Absolute
680 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200681
Willy Tarreau6a06a402007-07-15 20:15:28 +0200682chroot <jail dir>
683 Changes current directory to <jail dir> and performs a chroot() there before
684 dropping privileges. This increases the security level in case an unknown
685 vulnerability would be exploited, since it would make it very hard for the
686 attacker to exploit the system. This only works when the process is started
687 with superuser privileges. It is important to ensure that <jail_dir> is both
Davor Ocelice9ed2812017-12-25 17:49:28 +0100688 empty and non-writable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100689
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100690cpu-map [auto:]<process-set>[/<thread-set>] <cpu-set>...
691 On Linux 2.6 and above, it is possible to bind a process or a thread to a
692 specific CPU set. This means that the process or the thread will never run on
693 other CPUs. The "cpu-map" directive specifies CPU sets for process or thread
694 sets. The first argument is a process set, eventually followed by a thread
695 set. These sets have the format
696
697 all | odd | even | number[-[number]]
698
699 <number>> must be a number between 1 and 32 or 64, depending on the machine's
Davor Ocelice9ed2812017-12-25 17:49:28 +0100700 word size. Any process IDs above nbproc and any thread IDs above nbthread are
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100701 ignored. It is possible to specify a range with two such number delimited by
702 a dash ('-'). It also is possible to specify all processes at once using
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100703 "all", only odd numbers using "odd" or even numbers using "even", just like
704 with the "bind-process" directive. The second and forthcoming arguments are
Davor Ocelice9ed2812017-12-25 17:49:28 +0100705 CPU sets. Each CPU set is either a unique number between 0 and 31 or 63 or a
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100706 range with two such numbers delimited by a dash ('-'). Multiple CPU numbers
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100707 or ranges may be specified, and the processes or threads will be allowed to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100708 bind to all of them. Obviously, multiple "cpu-map" directives may be
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100709 specified. Each "cpu-map" directive will replace the previous ones when they
710 overlap. A thread will be bound on the intersection of its mapping and the
711 one of the process on which it is attached. If the intersection is null, no
712 specific binding will be set for the thread.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100713
Christopher Fauletff4121f2017-11-22 16:38:49 +0100714 Ranges can be partially defined. The higher bound can be omitted. In such
715 case, it is replaced by the corresponding maximum value, 32 or 64 depending
716 on the machine's word size.
717
Christopher Faulet26028f62017-11-22 15:01:51 +0100718 The prefix "auto:" can be added before the process set to let HAProxy
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100719 automatically bind a process or a thread to a CPU by incrementing
720 process/thread and CPU sets. To be valid, both sets must have the same
721 size. No matter the declaration order of the CPU sets, it will be bound from
722 the lowest to the highest bound. Having a process and a thread range with the
723 "auto:" prefix is not supported. Only one range is supported, the other one
724 must be a fixed number.
Christopher Faulet26028f62017-11-22 15:01:51 +0100725
726 Examples:
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100727 cpu-map 1-4 0-3 # bind processes 1 to 4 on the first 4 CPUs
728
729 cpu-map 1/all 0-3 # bind all threads of the first process on the
730 # first 4 CPUs
731
732 cpu-map 1- 0- # will be replaced by "cpu-map 1-64 0-63"
733 # or "cpu-map 1-32 0-31" depending on the machine's
734 # word size.
735
Christopher Faulet26028f62017-11-22 15:01:51 +0100736 # all these lines bind the process 1 to the cpu 0, the process 2 to cpu 1
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100737 # and so on.
Christopher Faulet26028f62017-11-22 15:01:51 +0100738 cpu-map auto:1-4 0-3
739 cpu-map auto:1-4 0-1 2-3
740 cpu-map auto:1-4 3 2 1 0
741
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100742 # all these lines bind the thread 1 to the cpu 0, the thread 2 to cpu 1
743 # and so on.
744 cpu-map auto:1/1-4 0-3
745 cpu-map auto:1/1-4 0-1 2-3
746 cpu-map auto:1/1-4 3 2 1 0
747
Davor Ocelice9ed2812017-12-25 17:49:28 +0100748 # bind each process to exactly one CPU using all/odd/even keyword
Christopher Faulet26028f62017-11-22 15:01:51 +0100749 cpu-map auto:all 0-63
750 cpu-map auto:even 0-31
751 cpu-map auto:odd 32-63
752
753 # invalid cpu-map because process and CPU sets have different sizes.
754 cpu-map auto:1-4 0 # invalid
755 cpu-map auto:1 0-3 # invalid
756
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100757 # invalid cpu-map because automatic binding is used with a process range
758 # and a thread range.
759 cpu-map auto:all/all 0 # invalid
760 cpu-map auto:all/1-4 0 # invalid
761 cpu-map auto:1-4/all 0 # invalid
762
Emeric Brunc8e8d122012-10-02 18:42:10 +0200763crt-base <dir>
764 Assigns a default directory to fetch SSL certificates from when a relative
765 path is used with "crtfile" directives. Absolute locations specified after
766 "crtfile" prevail and ignore "crt-base".
767
Willy Tarreau6a06a402007-07-15 20:15:28 +0200768daemon
769 Makes the process fork into background. This is the recommended mode of
770 operation. It is equivalent to the command line "-D" argument. It can be
Lukas Tribusf46bf952017-11-21 12:39:34 +0100771 disabled by the command line "-db" argument. This option is ignored in
772 systemd mode.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200773
David Carlier8167f302015-06-01 13:50:06 +0200774deviceatlas-json-file <path>
775 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
Davor Ocelice9ed2812017-12-25 17:49:28 +0100776 The path must be a valid JSON data file and accessible by HAProxy process.
David Carlier8167f302015-06-01 13:50:06 +0200777
778deviceatlas-log-level <value>
Davor Ocelice9ed2812017-12-25 17:49:28 +0100779 Sets the level of information returned by the API. This directive is
David Carlier8167f302015-06-01 13:50:06 +0200780 optional and set to 0 by default if not set.
781
782deviceatlas-separator <char>
783 Sets the character separator for the API properties results. This directive
784 is optional and set to | by default if not set.
785
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100786deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200787 Sets the client cookie's name used for the detection if the DeviceAtlas
788 Client-side component was used during the request. This directive is optional
789 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100790
Simon Horman98637e52014-06-20 12:30:16 +0900791external-check
792 Allows the use of an external agent to perform health checks.
793 This is disabled by default as a security precaution.
794 See "option external-check".
795
Willy Tarreau6a06a402007-07-15 20:15:28 +0200796gid <number>
797 Changes the process' group ID to <number>. It is recommended that the group
798 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
799 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100800 Note that if haproxy is started from a user having supplementary groups, it
801 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200802 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100803
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100804hard-stop-after <time>
805 Defines the maximum time allowed to perform a clean soft-stop.
806
807 Arguments :
808 <time> is the maximum time (by default in milliseconds) for which the
809 instance will remain alive when a soft-stop is received via the
810 SIGUSR1 signal.
811
812 This may be used to ensure that the instance will quit even if connections
813 remain opened during a soft-stop (for example with long timeouts for a proxy
814 in tcp mode). It applies both in TCP and HTTP mode.
815
816 Example:
817 global
818 hard-stop-after 30s
819
Willy Tarreau6a06a402007-07-15 20:15:28 +0200820group <group name>
821 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
822 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100823
Dragan Dosen7ad31542015-09-28 17:16:47 +0200824log <address> [len <length>] [format <format>] <facility> [max level [min level]]
Cyril Bonté3e954872018-03-20 23:30:27 +0100825 Adds a global syslog server. Several global servers can be defined. They
Davor Ocelice9ed2812017-12-25 17:49:28 +0100826 will receive logs for starts and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100827 configured with "log global".
828
829 <address> can be one of:
830
Willy Tarreau2769aa02007-12-27 18:26:09 +0100831 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100832 no port is specified, 514 is used by default (the standard syslog
833 port).
834
David du Colombier24bb5f52011-03-17 10:40:23 +0100835 - An IPv6 address followed by a colon and optionally a UDP port. If
836 no port is specified, 514 is used by default (the standard syslog
837 port).
838
Robert Tsai81ae1952007-12-05 10:47:29 +0100839 - A filesystem path to a UNIX domain socket, keeping in mind
840 considerations for chroot (be sure the path is accessible inside
841 the chroot) and uid/gid (be sure the path is appropriately
Davor Ocelice9ed2812017-12-25 17:49:28 +0100842 writable).
Robert Tsai81ae1952007-12-05 10:47:29 +0100843
William Lallemandb2f07452015-05-12 14:27:13 +0200844 You may want to reference some environment variables in the address
845 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100846
Willy Tarreau18324f52014-06-27 18:10:07 +0200847 <length> is an optional maximum line length. Log lines larger than this value
848 will be truncated before being sent. The reason is that syslog
849 servers act differently on log line length. All servers support the
850 default value of 1024, but some servers simply drop larger lines
851 while others do log them. If a server supports long lines, it may
852 make sense to set this value here in order to avoid truncating long
853 lines. Similarly, if a server drops long lines, it is preferable to
854 truncate them before sending them. Accepted values are 80 to 65535
855 inclusive. The default value of 1024 is generally fine for all
856 standard usages. Some specific cases of long captures or
Davor Ocelice9ed2812017-12-25 17:49:28 +0100857 JSON-formatted logs may require larger values. You may also need to
858 increase "tune.http.logurilen" if your request URIs are truncated.
Willy Tarreau18324f52014-06-27 18:10:07 +0200859
Dragan Dosen7ad31542015-09-28 17:16:47 +0200860 <format> is the log format used when generating syslog messages. It may be
861 one of the following :
862
863 rfc3164 The RFC3164 syslog message format. This is the default.
864 (https://tools.ietf.org/html/rfc3164)
865
866 rfc5424 The RFC5424 syslog message format.
867 (https://tools.ietf.org/html/rfc5424)
868
Robert Tsai81ae1952007-12-05 10:47:29 +0100869 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200870
871 kern user mail daemon auth syslog lpr news
872 uucp cron auth2 ftp ntp audit alert cron2
873 local0 local1 local2 local3 local4 local5 local6 local7
874
875 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200876 all messages are sent. If a maximum level is specified, only messages with a
877 severity at least as important as this level will be sent. An optional minimum
878 level can be specified. If it is set, logs emitted with a more severe level
879 than this one will be capped to this level. This is used to avoid sending
880 "emerg" messages on all terminals on some default syslog configurations.
881 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200882
Cyril Bontédc4d9032012-04-08 21:57:39 +0200883 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200884
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100885log-send-hostname [<string>]
886 Sets the hostname field in the syslog header. If optional "string" parameter
887 is set the header is set to the string contents, otherwise uses the hostname
888 of the system. Generally used if one is not relaying logs through an
889 intermediate syslog server or for simply customizing the hostname printed in
890 the logs.
891
Kevinm48936af2010-12-22 16:08:21 +0000892log-tag <string>
893 Sets the tag field in the syslog header to this string. It defaults to the
894 program name as launched from the command line, which usually is "haproxy".
895 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100896 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000897
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100898lua-load <file>
899 This global directive loads and executes a Lua file. This directive can be
900 used multiple times.
901
William Lallemand4cfede82017-11-24 22:02:34 +0100902master-worker [no-exit-on-failure]
William Lallemande202b1e2017-06-01 17:38:56 +0200903 Master-worker mode. It is equivalent to the command line "-W" argument.
904 This mode will launch a "master" which will monitor the "workers". Using
905 this mode, you can reload HAProxy directly by sending a SIGUSR2 signal to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100906 the master. The master-worker mode is compatible either with the foreground
William Lallemande202b1e2017-06-01 17:38:56 +0200907 or daemon mode. It is recommended to use this mode with multiprocess and
908 systemd.
William Lallemand4cfede82017-11-24 22:02:34 +0100909 By default, if a worker exits with a bad return code, in the case of a
910 segfault for example, all workers will be killed, and the master will leave.
911 It is convenient to combine this behavior with Restart=on-failure in a
912 systemd unit file in order to relaunch the whole process. If you don't want
913 this behavior, you must use the keyword "no-exit-on-failure".
William Lallemande202b1e2017-06-01 17:38:56 +0200914
William Lallemand4cfede82017-11-24 22:02:34 +0100915 See also "-W" in the management guide.
William Lallemande202b1e2017-06-01 17:38:56 +0200916
Willy Tarreau6a06a402007-07-15 20:15:28 +0200917nbproc <number>
918 Creates <number> processes when going daemon. This requires the "daemon"
919 mode. By default, only one process is created, which is the recommended mode
920 of operation. For systems limited to small sets of file descriptors per
921 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
922 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
923
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200924nbthread <number>
925 This setting is only available when support for threads was built in. It
926 creates <number> threads for each created processes. It means if HAProxy is
927 started in foreground, it only creates <number> threads for the first
928 process. FOR NOW, THREADS SUPPORT IN HAPROXY IS HIGHLY EXPERIMENTAL AND IT
929 MUST BE ENABLED WITH CAUTION AND AT YOUR OWN RISK. See also "nbproc".
930
Willy Tarreau6a06a402007-07-15 20:15:28 +0200931pidfile <pidfile>
Davor Ocelice9ed2812017-12-25 17:49:28 +0100932 Writes PIDs of all daemons into file <pidfile>. This option is equivalent to
Willy Tarreau6a06a402007-07-15 20:15:28 +0200933 the "-p" command line argument. The file must be accessible to the user
934 starting the process. See also "daemon".
935
Willy Tarreau1d549722016-02-16 12:41:57 +0100936presetenv <name> <value>
937 Sets environment variable <name> to value <value>. If the variable exists, it
938 is NOT overwritten. The changes immediately take effect so that the next line
939 in the configuration file sees the new value. See also "setenv", "resetenv",
940 and "unsetenv".
941
942resetenv [<name> ...]
943 Removes all environment variables except the ones specified in argument. It
944 allows to use a clean controlled environment before setting new values with
945 setenv or unsetenv. Please note that some internal functions may make use of
946 some environment variables, such as time manipulation functions, but also
947 OpenSSL or even external checks. This must be used with extreme care and only
948 after complete validation. The changes immediately take effect so that the
949 next line in the configuration file sees the new environment. See also
950 "setenv", "presetenv", and "unsetenv".
951
Christopher Fauletff4121f2017-11-22 16:38:49 +0100952stats bind-process [ all | odd | even | <process_num>[-[process_num>]] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200953 Limits the stats socket to a certain set of processes numbers. By default the
954 stats socket is bound to all processes, causing a warning to be emitted when
955 nbproc is greater than 1 because there is no way to select the target process
956 when connecting. However, by using this setting, it becomes possible to pin
957 the stats socket to a specific set of processes, typically the first one. The
958 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100959 the number of processes used. The maximum process ID depends on the machine's
Christopher Fauletff4121f2017-11-22 16:38:49 +0100960 word size (32 or 64). Ranges can be partially defined. The higher bound can
961 be omitted. In such case, it is replaced by the corresponding maximum
962 value. A better option consists in using the "process" setting of the "stats
963 socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +0200964
Baptiste Assmann5626f482015-08-23 10:00:10 +0200965server-state-base <directory>
966 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200967 file names which do not start with a '/'. See also "server-state-file",
968 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200969
970server-state-file <file>
971 Specifies the path to the file containing state of servers. If the path starts
972 with a slash ('/'), it is considered absolute, otherwise it is considered
973 relative to the directory specified using "server-state-base" (if set) or to
974 the current directory. Before reloading HAProxy, it is possible to save the
975 servers' current state using the stats command "show servers state". The
976 output of this command must be written in the file pointed by <file>. When
977 starting up, before handling traffic, HAProxy will read, load and apply state
978 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200979 configuration. See also "server-state-base" and "show servers state",
980 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +0200981
Willy Tarreau1d549722016-02-16 12:41:57 +0100982setenv <name> <value>
983 Sets environment variable <name> to value <value>. If the variable exists, it
984 is overwritten. The changes immediately take effect so that the next line in
985 the configuration file sees the new value. See also "presetenv", "resetenv",
986 and "unsetenv".
987
Willy Tarreau610f04b2014-02-13 11:36:41 +0100988ssl-default-bind-ciphers <ciphers>
989 This setting is only available when support for OpenSSL was built in. It sets
990 the default string describing the list of cipher algorithms ("cipher suite")
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300991 that are negotiated during the SSL/TLS handshake for all "bind" lines which
Willy Tarreau610f04b2014-02-13 11:36:41 +0100992 do not explicitly define theirs. The format of the string is defined in
993 "man 1 ciphers" from OpenSSL man pages, and can be for instance a string such
994 as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes). Please check the
995 "bind" keyword for more information.
996
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100997ssl-default-bind-options [<option>]...
998 This setting is only available when support for OpenSSL was built in. It sets
999 default ssl-options to force on all "bind" lines. Please check the "bind"
1000 keyword to see available options.
1001
1002 Example:
1003 global
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +02001004 ssl-default-bind-options ssl-min-ver TLSv1.0 no-tls-tickets
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001005
Willy Tarreau610f04b2014-02-13 11:36:41 +01001006ssl-default-server-ciphers <ciphers>
1007 This setting is only available when support for OpenSSL was built in. It
1008 sets the default string describing the list of cipher algorithms that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001009 negotiated during the SSL/TLS handshake with the server, for all "server"
Willy Tarreau610f04b2014-02-13 11:36:41 +01001010 lines which do not explicitly define theirs. The format of the string is
1011 defined in "man 1 ciphers". Please check the "server" keyword for more
1012 information.
1013
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001014ssl-default-server-options [<option>]...
1015 This setting is only available when support for OpenSSL was built in. It sets
1016 default ssl-options to force on all "server" lines. Please check the "server"
1017 keyword to see available options.
1018
Remi Gacogne47783ef2015-05-29 15:53:22 +02001019ssl-dh-param-file <file>
1020 This setting is only available when support for OpenSSL was built in. It sets
1021 the default DH parameters that are used during the SSL/TLS handshake when
1022 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
Davor Ocelice9ed2812017-12-25 17:49:28 +01001023 which do not explicitly define theirs. It will be overridden by custom DH
Remi Gacogne47783ef2015-05-29 15:53:22 +02001024 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001025 are not specified either by using ssl-dh-param-file or by setting them
1026 directly in the certificate file, pre-generated DH parameters of the size
1027 specified by tune.ssl.default-dh-param will be used. Custom parameters are
1028 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +02001029 Custom DH parameters may be generated by using the OpenSSL command
1030 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
1031 parameters should not be considered secure anymore.
1032
Emeric Brun850efd52014-01-29 12:24:34 +01001033ssl-server-verify [none|required]
1034 The default behavior for SSL verify on servers side. If specified to 'none',
1035 servers certificates are not verified. The default is 'required' except if
1036 forced using cmdline option '-dV'.
1037
Willy Tarreauabb175f2012-09-24 12:43:26 +02001038stats socket [<address:port>|<path>] [param*]
1039 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
1040 Connections to this socket will return various statistics outputs and even
1041 allow some commands to be issued to change some runtime settings. Please
Willy Tarreau1af20c72017-06-23 16:01:14 +02001042 consult section 9.3 "Unix Socket commands" of Management Guide for more
Kevin Decherf949c7202015-10-13 23:26:44 +02001043 details.
Willy Tarreau6162db22009-10-10 17:13:00 +02001044
Willy Tarreauabb175f2012-09-24 12:43:26 +02001045 All parameters supported by "bind" lines are supported, for instance to
1046 restrict access to some users or their access rights. Please consult
1047 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001048
1049stats timeout <timeout, in milliseconds>
1050 The default timeout on the stats socket is set to 10 seconds. It is possible
1051 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +01001052 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001053
1054stats maxconn <connections>
1055 By default, the stats socket is limited to 10 concurrent connections. It is
1056 possible to change this value with "stats maxconn".
1057
Willy Tarreau6a06a402007-07-15 20:15:28 +02001058uid <number>
1059 Changes the process' user ID to <number>. It is recommended that the user ID
1060 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
1061 be started with superuser privileges in order to be able to switch to another
1062 one. See also "gid" and "user".
1063
1064ulimit-n <number>
1065 Sets the maximum number of per-process file-descriptors to <number>. By
1066 default, it is automatically computed, so it is recommended not to use this
1067 option.
1068
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001069unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
1070 [ group <group> ] [ gid <gid> ]
1071
1072 Fixes common settings to UNIX listening sockets declared in "bind" statements.
1073 This is mainly used to simplify declaration of those UNIX sockets and reduce
1074 the risk of errors, since those settings are most commonly required but are
1075 also process-specific. The <prefix> setting can be used to force all socket
1076 path to be relative to that directory. This might be needed to access another
1077 component's chroot. Note that those paths are resolved before haproxy chroots
1078 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
1079 all have the same meaning as their homonyms used by the "bind" statement. If
1080 both are specified, the "bind" statement has priority, meaning that the
1081 "unix-bind" settings may be seen as process-wide default settings.
1082
Willy Tarreau1d549722016-02-16 12:41:57 +01001083unsetenv [<name> ...]
1084 Removes environment variables specified in arguments. This can be useful to
1085 hide some sensitive information that are occasionally inherited from the
1086 user's environment during some operations. Variables which did not exist are
1087 silently ignored so that after the operation, it is certain that none of
1088 these variables remain. The changes immediately take effect so that the next
1089 line in the configuration file will not see these variables. See also
1090 "setenv", "presetenv", and "resetenv".
1091
Willy Tarreau6a06a402007-07-15 20:15:28 +02001092user <user name>
1093 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
1094 See also "uid" and "group".
1095
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02001096node <name>
1097 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
1098
1099 This statement is useful in HA configurations where two or more processes or
1100 servers share the same IP address. By setting a different node-name on all
1101 nodes, it becomes easy to immediately spot what server is handling the
1102 traffic.
1103
1104description <text>
1105 Add a text that describes the instance.
1106
1107 Please note that it is required to escape certain characters (# for example)
1108 and this text is inserted into a html page so you should avoid using
1109 "<" and ">" characters.
1110
Thomas Holmesdb04f192015-05-18 13:21:39 +0100111151degrees-data-file <file path>
1112 The path of the 51Degrees data file to provide device detection services. The
Davor Ocelice9ed2812017-12-25 17:49:28 +01001113 file should be unzipped and accessible by HAProxy with relevant permissions.
Thomas Holmesdb04f192015-05-18 13:21:39 +01001114
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001115 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001116 compiled with USE_51DEGREES.
1117
Ben Shillitof25e8e52016-12-02 14:25:37 +0000111851degrees-property-name-list [<string> ...]
Thomas Holmesdb04f192015-05-18 13:21:39 +01001119 A list of 51Degrees property names to be load from the dataset. A full list
1120 of names is available on the 51Degrees website:
1121 https://51degrees.com/resources/property-dictionary
1122
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001123 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001124 compiled with USE_51DEGREES.
1125
Dragan Dosen93b38d92015-06-29 16:43:25 +0200112651degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +01001127 A char that will be appended to every property value in a response header
1128 containing 51Degrees results. If not set that will be set as ','.
1129
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001130 Please note that this option is only available when haproxy has been
1131 compiled with USE_51DEGREES.
1132
113351degrees-cache-size <number>
1134 Sets the size of the 51Degrees converter cache to <number> entries. This
1135 is an LRU cache which reminds previous device detections and their results.
1136 By default, this cache is disabled.
1137
1138 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001139 compiled with USE_51DEGREES.
1140
scientiamobiled0027ed2016-11-04 10:55:08 +01001141wurfl-data-file <file path>
1142 The path of the WURFL data file to provide device detection services. The
1143 file should be accessible by HAProxy with relevant permissions.
1144
1145 Please note that this option is only available when haproxy has been compiled
1146 with USE_WURFL=1.
1147
1148wurfl-information-list [<capability>]*
1149 A space-delimited list of WURFL capabilities, virtual capabilities, property
1150 names we plan to use in injected headers. A full list of capability and
1151 virtual capability names is available on the Scientiamobile website :
1152
1153 https://www.scientiamobile.com/wurflCapability
1154
1155 Valid WURFL properties are:
1156 - wurfl_id Contains the device ID of the matched device.
1157
1158 - wurfl_root_id Contains the device root ID of the matched
1159 device.
1160
1161 - wurfl_isdevroot Tells if the matched device is a root device.
1162 Possible values are "TRUE" or "FALSE".
1163
1164 - wurfl_useragent The original useragent coming with this
1165 particular web request.
1166
1167 - wurfl_api_version Contains a string representing the currently
1168 used Libwurfl API version.
1169
1170 - wurfl_engine_target Contains a string representing the currently
1171 set WURFL Engine Target. Possible values are
1172 "HIGH_ACCURACY", "HIGH_PERFORMANCE", "INVALID".
1173
1174 - wurfl_info A string containing information on the parsed
1175 wurfl.xml and its full path.
1176
1177 - wurfl_last_load_time Contains the UNIX timestamp of the last time
1178 WURFL has been loaded successfully.
1179
1180 - wurfl_normalized_useragent The normalized useragent.
1181
1182 - wurfl_useragent_priority The user agent priority used by WURFL.
1183
1184 Please note that this option is only available when haproxy has been compiled
1185 with USE_WURFL=1.
1186
1187wurfl-information-list-separator <char>
1188 A char that will be used to separate values in a response header containing
1189 WURFL results. If not set that a comma (',') will be used by default.
1190
1191 Please note that this option is only available when haproxy has been compiled
1192 with USE_WURFL=1.
1193
1194wurfl-patch-file [<file path>]
1195 A list of WURFL patch file paths. Note that patches are loaded during startup
1196 thus before the chroot.
1197
1198 Please note that this option is only available when haproxy has been compiled
1199 with USE_WURFL=1.
1200
1201wurfl-engine-mode { accuracy | performance }
1202 Sets the WURFL engine target. You can choose between 'accuracy' or
1203 'performance' targets. In performance mode, desktop web browser detection is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001204 done programmatically without referencing the WURFL data. As a result, most
scientiamobiled0027ed2016-11-04 10:55:08 +01001205 desktop web browsers are returned as generic_web_browser WURFL ID for
Davor Ocelice9ed2812017-12-25 17:49:28 +01001206 performance. If either performance or accuracy are not defined, performance
scientiamobiled0027ed2016-11-04 10:55:08 +01001207 mode is enabled by default.
1208
1209 Please note that this option is only available when haproxy has been compiled
1210 with USE_WURFL=1.
1211
1212wurfl-cache-size <U>[,<D>]
1213 Sets the WURFL caching strategy. Here <U> is the Useragent cache size, and
1214 <D> is the internal device cache size. There are three possibilities here :
1215 - "0" : no cache is used.
1216 - <U> : the Single LRU cache is used, the size is expressed in elements.
1217 - <U>,<D> : the Double LRU cache is used, both sizes are in elements. This is
1218 the highest performing option.
1219
1220 Please note that this option is only available when haproxy has been compiled
1221 with USE_WURFL=1.
1222
1223wurfl-useragent-priority { plain | sideloaded_browser }
1224 Tells WURFL if it should prioritize use of the plain user agent ('plain')
1225 over the default sideloaded browser user agent ('sideloaded_browser').
1226
1227 Please note that this option is only available when haproxy has been compiled
1228 with USE_WURFL=1.
1229
Willy Tarreau6a06a402007-07-15 20:15:28 +02001230
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012313.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001232-----------------------
1233
Willy Tarreau1746eec2014-04-25 10:46:47 +02001234max-spread-checks <delay in milliseconds>
1235 By default, haproxy tries to spread the start of health checks across the
1236 smallest health check interval of all the servers in a farm. The principle is
1237 to avoid hammering services running on the same server. But when using large
1238 check intervals (10 seconds or more), the last servers in the farm take some
1239 time before starting to be tested, which can be a problem. This parameter is
1240 used to enforce an upper bound on delay between the first and the last check,
1241 even if the servers' check intervals are larger. When servers run with
1242 shorter intervals, their intervals will be respected though.
1243
Willy Tarreau6a06a402007-07-15 20:15:28 +02001244maxconn <number>
1245 Sets the maximum per-process number of concurrent connections to <number>. It
1246 is equivalent to the command-line argument "-n". Proxies will stop accepting
1247 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001248 automatically adjusted according to this value. See also "ulimit-n". Note:
1249 the "select" poller cannot reliably use more than 1024 file descriptors on
1250 some platforms. If your platform only supports select and reports "select
1251 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaud0256482015-01-15 21:45:22 +01001252 below 500 in general). If this value is not set, it will default to the value
1253 set in DEFAULT_MAXCONN at build time (reported in haproxy -vv) if no memory
1254 limit is enforced, or will be computed based on the memory limit, the buffer
1255 size, memory allocated to compression, SSL cache size, and use or not of SSL
1256 and the associated maxsslconn (which can also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001257
Willy Tarreau81c25d02011-09-07 15:17:21 +02001258maxconnrate <number>
1259 Sets the maximum per-process number of connections per second to <number>.
1260 Proxies will stop accepting connections when this limit is reached. It can be
1261 used to limit the global capacity regardless of each frontend capacity. It is
1262 important to note that this can only be used as a service protection measure,
1263 as there will not necessarily be a fair share between frontends when the
1264 limit is reached, so it's a good idea to also limit each frontend to some
1265 value close to its expected share. Also, lowering tune.maxaccept can improve
1266 fairness.
1267
William Lallemandd85f9172012-11-09 17:05:39 +01001268maxcomprate <number>
1269 Sets the maximum per-process input compression rate to <number> kilobytes
Davor Ocelice9ed2812017-12-25 17:49:28 +01001270 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001271 level will be decreased during the session. If the maximum is reached at the
1272 beginning of a session, the session will not compress at all. If the maximum
1273 is not reached, the compression level will be increased up to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001274 tune.comp.maxlevel. A value of zero means there is no limit, this is the
William Lallemandd85f9172012-11-09 17:05:39 +01001275 default value.
1276
William Lallemand072a2bf2012-11-20 17:01:01 +01001277maxcompcpuusage <number>
1278 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1279 for new requests or decreasing the compression level of current requests.
1280 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1281 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1282 case of multiple processes (nbproc > 1), each process manages its individual
1283 usage. A value of 100 disable the limit. The default value is 100. Setting
1284 a lower value will prevent the compression work from slowing the whole
1285 process down and from introducing high latencies.
1286
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001287maxpipes <number>
1288 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1289 are only used by kernel-based tcp splicing. Since a pipe contains two file
1290 descriptors, the "ulimit-n" value will be increased accordingly. The default
1291 value is maxconn/4, which seems to be more than enough for most heavy usages.
1292 The splice code dynamically allocates and releases pipes, and can fall back
1293 to standard copy, so setting this value too low may only impact performance.
1294
Willy Tarreau93e7c002013-10-07 18:51:07 +02001295maxsessrate <number>
1296 Sets the maximum per-process number of sessions per second to <number>.
1297 Proxies will stop accepting connections when this limit is reached. It can be
1298 used to limit the global capacity regardless of each frontend capacity. It is
1299 important to note that this can only be used as a service protection measure,
1300 as there will not necessarily be a fair share between frontends when the
1301 limit is reached, so it's a good idea to also limit each frontend to some
1302 value close to its expected share. Also, lowering tune.maxaccept can improve
1303 fairness.
1304
Willy Tarreau403edff2012-09-06 11:58:37 +02001305maxsslconn <number>
1306 Sets the maximum per-process number of concurrent SSL connections to
1307 <number>. By default there is no SSL-specific limit, which means that the
1308 global maxconn setting will apply to all connections. Setting this limit
1309 avoids having openssl use too much memory and crash when malloc returns NULL
1310 (since it unfortunately does not reliably check for such conditions). Note
1311 that the limit applies both to incoming and outgoing connections, so one
1312 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001313 If this value is not set, but a memory limit is enforced, this value will be
1314 automatically computed based on the memory limit, maxconn, the buffer size,
1315 memory allocated to compression, SSL cache size, and use of SSL in either
1316 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1317 when there is a memory limit, haproxy will automatically adjust these values
1318 so that 100% of the connections can be made over SSL with no risk, and will
1319 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001320
Willy Tarreaue43d5322013-10-07 20:01:52 +02001321maxsslrate <number>
1322 Sets the maximum per-process number of SSL sessions per second to <number>.
1323 SSL listeners will stop accepting connections when this limit is reached. It
1324 can be used to limit the global SSL CPU usage regardless of each frontend
1325 capacity. It is important to note that this can only be used as a service
1326 protection measure, as there will not necessarily be a fair share between
1327 frontends when the limit is reached, so it's a good idea to also limit each
1328 frontend to some value close to its expected share. It is also important to
1329 note that the sessions are accounted before they enter the SSL stack and not
1330 after, which also protects the stack against bad handshakes. Also, lowering
1331 tune.maxaccept can improve fairness.
1332
William Lallemand9d5f5482012-11-07 16:12:57 +01001333maxzlibmem <number>
1334 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1335 When the maximum amount is reached, future sessions will not compress as long
1336 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001337 The default value is 0. The value is available in bytes on the UNIX socket
1338 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1339 "ZlibMemUsage" in bytes.
1340
Willy Tarreau6a06a402007-07-15 20:15:28 +02001341noepoll
1342 Disables the use of the "epoll" event polling system on Linux. It is
1343 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001344 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001345
1346nokqueue
1347 Disables the use of the "kqueue" event polling system on BSD. It is
1348 equivalent to the command-line argument "-dk". The next polling system
1349 used will generally be "poll". See also "nopoll".
1350
1351nopoll
1352 Disables the use of the "poll" event polling system. It is equivalent to the
1353 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001354 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001355 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001356
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001357nosplice
1358 Disables the use of kernel tcp splicing between sockets on Linux. It is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001359 equivalent to the command line argument "-dS". Data will then be copied
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001360 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001361 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001362 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1363 be used. This option makes it easier to globally disable kernel splicing in
1364 case of doubt. See also "option splice-auto", "option splice-request" and
1365 "option splice-response".
1366
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001367nogetaddrinfo
1368 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1369 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1370
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001371noreuseport
1372 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
1373 command line argument "-dR".
1374
Willy Tarreaufe255b72007-10-14 23:09:26 +02001375spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001376 Sometimes it is desirable to avoid sending agent and health checks to
1377 servers at exact intervals, for instance when many logical servers are
1378 located on the same physical server. With the help of this parameter, it
1379 becomes possible to add some randomness in the check interval between 0
1380 and +/- 50%. A value between 2 and 5 seems to show good results. The
1381 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001382
Davor Ocelice9ed2812017-12-25 17:49:28 +01001383ssl-engine <name> [algo <comma-separated list of algorithms>]
Grant Zhang872f9c22017-01-21 01:10:18 +00001384 Sets the OpenSSL engine to <name>. List of valid values for <name> may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01001385 obtained using the command "openssl engine". This statement may be used
Grant Zhang872f9c22017-01-21 01:10:18 +00001386 multiple times, it will simply enable multiple crypto engines. Referencing an
1387 unsupported engine will prevent haproxy from starting. Note that many engines
1388 will lead to lower HTTPS performance than pure software with recent
1389 processors. The optional command "algo" sets the default algorithms an ENGINE
1390 will supply using the OPENSSL function ENGINE_set_default_string(). A value
Davor Ocelice9ed2812017-12-25 17:49:28 +01001391 of "ALL" uses the engine for all cryptographic operations. If no list of
1392 algo is specified then the value of "ALL" is used. A comma-separated list
Grant Zhang872f9c22017-01-21 01:10:18 +00001393 of different algorithms may be specified, including: RSA, DSA, DH, EC, RAND,
1394 CIPHERS, DIGESTS, PKEY, PKEY_CRYPTO, PKEY_ASN1. This is the same format that
1395 openssl configuration file uses:
1396 https://www.openssl.org/docs/man1.0.2/apps/config.html
1397
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001398ssl-mode-async
1399 Adds SSL_MODE_ASYNC mode to the SSL context. This enables asynchronous TLS
Emeric Brun3854e012017-05-17 20:42:48 +02001400 I/O operations if asynchronous capable SSL engines are used. The current
Emeric Brunb5e42a82017-06-06 12:35:14 +00001401 implementation supports a maximum of 32 engines. The Openssl ASYNC API
1402 doesn't support moving read/write buffers and is not compliant with
1403 haproxy's buffer management. So the asynchronous mode is disabled on
1404 read/write operations (it is only enabled during initial and reneg
1405 handshakes).
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001406
Willy Tarreau33cb0652014-12-23 22:52:37 +01001407tune.buffers.limit <number>
1408 Sets a hard limit on the number of buffers which may be allocated per process.
1409 The default value is zero which means unlimited. The minimum non-zero value
1410 will always be greater than "tune.buffers.reserve" and should ideally always
1411 be about twice as large. Forcing this value can be particularly useful to
1412 limit the amount of memory a process may take, while retaining a sane
Davor Ocelice9ed2812017-12-25 17:49:28 +01001413 behavior. When this limit is reached, sessions which need a buffer wait for
Willy Tarreau33cb0652014-12-23 22:52:37 +01001414 another one to be released by another session. Since buffers are dynamically
1415 allocated and released, the waiting time is very short and not perceptible
1416 provided that limits remain reasonable. In fact sometimes reducing the limit
1417 may even increase performance by increasing the CPU cache's efficiency. Tests
1418 have shown good results on average HTTP traffic with a limit to 1/10 of the
1419 expected global maxconn setting, which also significantly reduces memory
1420 usage. The memory savings come from the fact that a number of connections
1421 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1422 advised to do so by an haproxy core developer.
1423
Willy Tarreau1058ae72014-12-23 22:40:40 +01001424tune.buffers.reserve <number>
1425 Sets the number of buffers which are pre-allocated and reserved for use only
1426 during memory shortage conditions resulting in failed memory allocations. The
1427 minimum value is 2 and is also the default. There is no reason a user would
1428 want to change this value, it's mostly aimed at haproxy core developers.
1429
Willy Tarreau27a674e2009-08-17 07:23:33 +02001430tune.bufsize <number>
1431 Sets the buffer size to this size (in bytes). Lower values allow more
1432 sessions to coexist in the same amount of RAM, and higher values allow some
1433 applications with very large cookies to work. The default value is 16384 and
1434 can be changed at build time. It is strongly recommended not to change this
1435 from the default value, as very low values will break some services such as
1436 statistics, and values larger than default size will increase memory usage,
1437 possibly causing the system to run out of memory. At least the global maxconn
Willy Tarreau45a66cc2017-11-24 11:28:00 +01001438 parameter should be decreased by the same factor as this one is increased. In
1439 addition, use of HTTP/2 mandates that this value must be 16384 or more. If an
1440 HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001441 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
1442 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +02001443
Willy Tarreau43961d52010-10-04 20:39:20 +02001444tune.chksize <number>
1445 Sets the check buffer size to this size (in bytes). Higher values may help
1446 find string or regex patterns in very large pages, though doing so may imply
1447 more memory and CPU usage. The default value is 16384 and can be changed at
1448 build time. It is not recommended to change this value, but to use better
1449 checks whenever possible.
1450
William Lallemandf3747832012-11-09 12:33:10 +01001451tune.comp.maxlevel <number>
1452 Sets the maximum compression level. The compression level affects CPU
1453 usage during compression. This value affects CPU usage during compression.
1454 Each session using compression initializes the compression algorithm with
1455 this value. The default value is 1.
1456
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02001457tune.h2.header-table-size <number>
1458 Sets the HTTP/2 dynamic header table size. It defaults to 4096 bytes and
1459 cannot be larger than 65536 bytes. A larger value may help certain clients
1460 send more compact requests, depending on their capabilities. This amount of
1461 memory is consumed for each HTTP/2 connection. It is recommended not to
1462 change it.
1463
Willy Tarreaue6baec02017-07-27 11:45:11 +02001464tune.h2.initial-window-size <number>
1465 Sets the HTTP/2 initial window size, which is the number of bytes the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01001466 can upload before waiting for an acknowledgment from haproxy. This setting
1467 only affects payload contents (i.e. the body of POST requests), not headers.
Willy Tarreaue6baec02017-07-27 11:45:11 +02001468 The default value is 65535, which roughly allows up to 5 Mbps of upload
1469 bandwidth per client over a network showing a 100 ms ping time, or 500 Mbps
1470 over a 1-ms local network. It can make sense to increase this value to allow
1471 faster uploads, or to reduce it to increase fairness when dealing with many
1472 clients. It doesn't affect resource usage.
1473
Willy Tarreau5242ef82017-07-27 11:47:28 +02001474tune.h2.max-concurrent-streams <number>
1475 Sets the HTTP/2 maximum number of concurrent streams per connection (ie the
1476 number of outstanding requests on a single connection). The default value is
1477 100. A larger one may slightly improve page load time for complex sites when
1478 visited over high latency networks, but increases the amount of resources a
1479 single client may allocate. A value of zero disables the limit so a single
1480 client may create as many streams as allocatable by haproxy. It is highly
1481 recommended not to change this value.
1482
Willy Tarreau193b8c62012-11-22 00:17:38 +01001483tune.http.cookielen <number>
1484 Sets the maximum length of captured cookies. This is the maximum value that
1485 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1486 will automatically be truncated to this one. It is important not to set too
1487 high a value because all cookie captures still allocate this size whatever
1488 their configured value (they share a same pool). This value is per request
1489 per response, so the memory allocated is twice this value per connection.
1490 When not specified, the limit is set to 63 characters. It is recommended not
1491 to change this value.
1492
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001493tune.http.logurilen <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001494 Sets the maximum length of request URI in logs. This prevents truncating long
1495 request URIs with valuable query strings in log lines. This is not related
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001496 to syslog limits. If you increase this limit, you may also increase the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001497 'log ... len yyy' parameter. Your syslog daemon may also need specific
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001498 configuration directives too.
1499 The default value is 1024.
1500
Willy Tarreauac1932d2011-10-24 19:14:41 +02001501tune.http.maxhdr <number>
1502 Sets the maximum number of headers in a request. When a request comes with a
1503 number of headers greater than this value (including the first line), it is
1504 rejected with a "400 Bad Request" status code. Similarly, too large responses
1505 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1506 for all usages, considering that the widely deployed Apache server uses the
1507 same limit. It can be useful to push this limit further to temporarily allow
Christopher Faulet50174f32017-06-21 16:31:35 +02001508 a buggy application to work by the time it gets fixed. The accepted range is
1509 1..32767. Keep in mind that each new header consumes 32bits of memory for
1510 each session, so don't push this limit too high.
Willy Tarreauac1932d2011-10-24 19:14:41 +02001511
Willy Tarreau7e312732014-02-12 16:35:14 +01001512tune.idletimer <timeout>
1513 Sets the duration after which haproxy will consider that an empty buffer is
1514 probably associated with an idle stream. This is used to optimally adjust
1515 some packet sizes while forwarding large and small data alternatively. The
1516 decision to use splice() or to send large buffers in SSL is modulated by this
1517 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1518 means that haproxy will not try to detect idle streams. The default is 1000,
Davor Ocelice9ed2812017-12-25 17:49:28 +01001519 which seems to correctly detect end user pauses (e.g. read a page before
Willy Tarreau7e312732014-02-12 16:35:14 +01001520 clicking). There should be not reason for changing this value. Please check
1521 tune.ssl.maxrecord below.
1522
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001523tune.lua.forced-yield <number>
1524 This directive forces the Lua engine to execute a yield each <number> of
Tim Düsterhus4896c442016-11-29 02:15:19 +01001525 instructions executed. This permits interrupting a long script and allows the
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001526 HAProxy scheduler to process other tasks like accepting connections or
1527 forwarding traffic. The default value is 10000 instructions. If HAProxy often
Davor Ocelice9ed2812017-12-25 17:49:28 +01001528 executes some Lua code but more responsiveness is required, this value can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001529 lowered. If the Lua code is quite long and its result is absolutely required
1530 to process the data, the <number> can be increased.
1531
Willy Tarreau32f61e22015-03-18 17:54:59 +01001532tune.lua.maxmem
1533 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1534 default it is zero which means unlimited. It is important to set a limit to
1535 ensure that a bug in a script will not result in the system running out of
1536 memory.
1537
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001538tune.lua.session-timeout <timeout>
1539 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001540 preventing infinite loops or spending too much time in Lua. This timeout
1541 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001542 not taken in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001543
1544tune.lua.task-timeout <timeout>
1545 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1546 dedicated to the tasks. By default, this timeout isn't set because a task may
1547 remain alive during of the lifetime of HAProxy. For example, a task used to
1548 check servers.
1549
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001550tune.lua.service-timeout <timeout>
1551 This is the execution timeout for the Lua services. This is useful for
1552 preventing infinite loops or spending too much time in Lua. This timeout
1553 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001554 not taken in account. The default timeout is 4s.
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001555
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001556tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001557 Sets the maximum number of consecutive connections a process may accept in a
1558 row before switching to other work. In single process mode, higher numbers
1559 give better performance at high connection rates. However in multi-process
1560 modes, keeping a bit of fairness between processes generally is better to
1561 increase performance. This value applies individually to each listener, so
1562 that the number of processes a listener is bound to is taken into account.
1563 This value defaults to 64. In multi-process mode, it is divided by twice
1564 the number of processes the listener is bound to. Setting this value to -1
1565 completely disables the limitation. It should normally not be needed to tweak
1566 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001567
1568tune.maxpollevents <number>
1569 Sets the maximum amount of events that can be processed at once in a call to
1570 the polling system. The default value is adapted to the operating system. It
1571 has been noticed that reducing it below 200 tends to slightly decrease
1572 latency at the expense of network bandwidth, and increasing it above 200
1573 tends to trade latency for slightly increased bandwidth.
1574
Willy Tarreau27a674e2009-08-17 07:23:33 +02001575tune.maxrewrite <number>
1576 Sets the reserved buffer space to this size in bytes. The reserved space is
1577 used for header rewriting or appending. The first reads on sockets will never
1578 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1579 bufsize, though that does not make much sense since there are rarely large
1580 numbers of headers to add. Setting it too high prevents processing of large
1581 requests or responses. Setting it too low prevents addition of new headers
1582 to already large requests or to POST requests. It is generally wise to set it
1583 to about 1024. It is automatically readjusted to half of bufsize if it is
1584 larger than that. This means you don't have to worry about it when changing
1585 bufsize.
1586
Willy Tarreauf3045d22015-04-29 16:24:50 +02001587tune.pattern.cache-size <number>
1588 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1589 cache which reminds previous lookups and their results. It is used by ACLs
1590 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1591 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1592 strings. It applies to pattern expressions which means that it will be able
1593 to memorize the result of a lookup among all the patterns specified on a
1594 configuration line (including all those loaded from files). It automatically
1595 invalidates entries which are updated using HTTP actions or on the CLI. The
1596 default cache size is set to 10000 entries, which limits its footprint to
1597 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1598 risk of collision in this cache, which is in the order of the size of the
1599 cache divided by 2^64. Typically, at 10000 requests per second with the
1600 default cache size of 10000 entries, there's 1% chance that a brute force
1601 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1602 This is considered much lower than the risk of a memory corruption caused by
1603 aging components. If this is not acceptable, the cache can be disabled by
1604 setting this parameter to 0.
1605
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001606tune.pipesize <number>
1607 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1608 are the default size for the system. But sometimes when using TCP splicing,
1609 it can improve performance to increase pipe sizes, especially if it is
1610 suspected that pipes are not filled and that many calls to splice() are
1611 performed. This has an impact on the kernel's memory footprint, so this must
1612 not be changed if impacts are not understood.
1613
Willy Tarreaue803de22010-01-21 17:43:04 +01001614tune.rcvbuf.client <number>
1615tune.rcvbuf.server <number>
1616 Forces the kernel socket receive buffer size on the client or the server side
1617 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1618 and backends. It should normally never be set, and the default size (0) lets
1619 the kernel autotune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001620 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001621 order to save kernel memory by preventing it from buffering too large amounts
1622 of received data. Lower values will significantly increase CPU usage though.
1623
Willy Tarreaub22fc302015-12-14 12:04:35 +01001624tune.recv_enough <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001625 HAProxy uses some hints to detect that a short read indicates the end of the
Willy Tarreaub22fc302015-12-14 12:04:35 +01001626 socket buffers. One of them is that a read returns more than <recv_enough>
1627 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1628 may be changed by this setting to better deal with workloads involving lots
1629 of short messages such as telnet or SSH sessions.
1630
Olivier Houchard1599b802018-05-24 18:59:04 +02001631tune.runqueue-depth <number>
1632 Sets the maxinum amount of task that can be processed at once when running
1633 tasks. The default value is 200. Increasing it may incur latency when
1634 dealing with I/Os, making it too small can incur extra overhead.
1635
Willy Tarreaue803de22010-01-21 17:43:04 +01001636tune.sndbuf.client <number>
1637tune.sndbuf.server <number>
1638 Forces the kernel socket send buffer size on the client or the server side to
1639 the specified value in bytes. This value applies to all TCP/HTTP frontends
1640 and backends. It should normally never be set, and the default size (0) lets
1641 the kernel autotune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001642 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001643 order to save kernel memory by preventing it from buffering too large amounts
1644 of received data. Lower values will significantly increase CPU usage though.
1645 Another use case is to prevent write timeouts with extremely slow clients due
1646 to the kernel waiting for a large part of the buffer to be read before
1647 notifying haproxy again.
1648
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001649tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001650 Sets the size of the global SSL session cache, in a number of blocks. A block
1651 is large enough to contain an encoded session without peer certificate.
1652 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001653 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001654 200 bytes of memory. The default value may be forced at build time, otherwise
Davor Ocelice9ed2812017-12-25 17:49:28 +01001655 defaults to 20000. When the cache is full, the most idle entries are purged
Emeric Brunaf9619d2012-11-28 18:47:52 +01001656 and reassigned. Higher values reduce the occurrence of such a purge, hence
1657 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1658 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001659 and are shared between all processes if "nbproc" is greater than 1. Setting
1660 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001661
Emeric Brun8dc60392014-05-09 13:52:00 +02001662tune.ssl.force-private-cache
1663 This boolean disables SSL session cache sharing between all processes. It
1664 should normally not be used since it will force many renegotiations due to
1665 clients hitting a random process. But it may be required on some operating
1666 systems where none of the SSL cache synchronization method may be used. In
1667 this case, adding a first layer of hash-based load balancing before the SSL
1668 layer might limit the impact of the lack of session sharing.
1669
Emeric Brun4f65bff2012-11-16 15:11:00 +01001670tune.ssl.lifetime <timeout>
1671 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001672 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001673 does not guarantee that sessions will last that long, because if the cache is
1674 full, the longest idle sessions will be purged despite their configured
1675 lifetime. The real usefulness of this setting is to prevent sessions from
1676 being used for too long.
1677
Willy Tarreaubfd59462013-02-21 07:46:09 +01001678tune.ssl.maxrecord <number>
1679 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1680 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1681 data only once it has received a full record. With large records, it means
1682 that clients might have to download up to 16kB of data before starting to
1683 process them. Limiting the value can improve page load times on browsers
1684 located over high latency or low bandwidth networks. It is suggested to find
1685 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1686 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1687 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1688 2859 gave good results during tests. Use "strace -e trace=write" to find the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001689 best value. HAProxy will automatically switch to this setting after an idle
Willy Tarreau7e312732014-02-12 16:35:14 +01001690 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001691
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001692tune.ssl.default-dh-param <number>
1693 Sets the maximum size of the Diffie-Hellman parameters used for generating
1694 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1695 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1696 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1697 this maximum value. Default value if 1024. Only 1024 or higher values are
1698 allowed. Higher values will increase the CPU load, and values greater than
1699 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001700 used if static Diffie-Hellman parameters are supplied either directly
1701 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001702
Christopher Faulet31af49d2015-06-09 17:29:50 +02001703tune.ssl.ssl-ctx-cache-size <number>
1704 Sets the size of the cache used to store generated certificates to <number>
1705 entries. This is a LRU cache. Because generating a SSL certificate
1706 dynamically is expensive, they are cached. The default cache size is set to
1707 1000 entries.
1708
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01001709tune.ssl.capture-cipherlist-size <number>
1710 Sets the maximum size of the buffer used for capturing client-hello cipher
1711 list. If the value is 0 (default value) the capture is disabled, otherwise
1712 a buffer is allocated for each SSL/TLS connection.
1713
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001714tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001715tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001716tune.vars.reqres-max-size <size>
1717tune.vars.sess-max-size <size>
1718tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001719 These five tunes help to manage the maximum amount of memory used by the
1720 variables system. "global" limits the overall amount of memory available for
1721 all scopes. "proc" limits the memory for the process scope, "sess" limits the
1722 memory for the session scope, "txn" for the transaction scope, and "reqres"
1723 limits the memory for each request or response processing.
1724 Memory accounting is hierarchical, meaning more coarse grained limits include
1725 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
1726 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001727
Daniel Schneller0b547052016-03-21 20:46:57 +01001728 For example, when "tune.vars.sess-max-size" is limited to 100,
1729 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
1730 100 either. If we create a variable "txn.var" that contains 100 bytes,
1731 all available space is consumed.
1732 Notice that exceeding the limits at runtime will not result in an error
1733 message, but values might be cut off or corrupted. So make sure to accurately
1734 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001735
William Lallemanda509e4c2012-11-07 16:54:34 +01001736tune.zlib.memlevel <number>
1737 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001738 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001739 state. A value of 1 uses minimum memory but is slow and reduces compression
Davor Ocelice9ed2812017-12-25 17:49:28 +01001740 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
William Lallemanda509e4c2012-11-07 16:54:34 +01001741 between 1 and 9. The default value is 8.
1742
1743tune.zlib.windowsize <number>
1744 Sets the window size (the size of the history buffer) as a parameter of the
1745 zlib initialization for each session. Larger values of this parameter result
Davor Ocelice9ed2812017-12-25 17:49:28 +01001746 in better compression at the expense of memory usage. Can be a value between
1747 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001748
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017493.3. Debugging
1750--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001751
1752debug
1753 Enables debug mode which dumps to stdout all exchanges, and disables forking
1754 into background. It is the equivalent of the command-line argument "-d". It
1755 should never be used in a production configuration since it may prevent full
1756 system startup.
1757
1758quiet
1759 Do not display any message during startup. It is equivalent to the command-
1760 line argument "-q".
1761
Emeric Brunf099e792010-09-27 12:05:28 +02001762
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010017633.4. Userlists
1764--------------
1765It is possible to control access to frontend/backend/listen sections or to
1766http stats by allowing only authenticated and authorized users. To do this,
1767it is required to create at least one userlist and to define users.
1768
1769userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001770 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001771 used to store authentication & authorization data for independent customers.
1772
1773group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001774 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001775 attach users to this group by using a comma separated list of names
1776 proceeded by "users" keyword.
1777
Cyril Bontéf0c60612010-02-06 14:44:47 +01001778user <username> [password|insecure-password <password>]
1779 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001780 Adds user <username> to the current userlist. Both secure (encrypted) and
1781 insecure (unencrypted) passwords can be used. Encrypted passwords are
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01001782 evaluated using the crypt(3) function, so depending on the system's
1783 capabilities, different algorithms are supported. For example, modern Glibc
1784 based Linux systems support MD5, SHA-256, SHA-512, and, of course, the
1785 classic DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001786
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01001787 Attention: Be aware that using encrypted passwords might cause significantly
1788 increased CPU usage, depending on the number of requests, and the algorithm
1789 used. For any of the hashed variants, the password for each request must
1790 be processed through the chosen algorithm, before it can be compared to the
1791 value specified in the config file. Most current algorithms are deliberately
1792 designed to be expensive to compute to achieve resistance against brute
1793 force attacks. They do not simply salt/hash the clear text password once,
1794 but thousands of times. This can quickly become a major factor in haproxy's
1795 overall CPU consumption!
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001796
1797 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001798 userlist L1
1799 group G1 users tiger,scott
1800 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001801
Cyril Bontéf0c60612010-02-06 14:44:47 +01001802 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1803 user scott insecure-password elgato
1804 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001805
Cyril Bontéf0c60612010-02-06 14:44:47 +01001806 userlist L2
1807 group G1
1808 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001809
Cyril Bontéf0c60612010-02-06 14:44:47 +01001810 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1811 user scott insecure-password elgato groups G1,G2
1812 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001813
1814 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001815
Emeric Brunf099e792010-09-27 12:05:28 +02001816
18173.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001818----------
Emeric Brun94900952015-06-11 18:25:54 +02001819It is possible to propagate entries of any data-types in stick-tables between
1820several haproxy instances over TCP connections in a multi-master fashion. Each
1821instance pushes its local updates and insertions to remote peers. The pushed
1822values overwrite remote ones without aggregation. Interrupted exchanges are
1823automatically detected and recovered from the last known point.
1824In addition, during a soft restart, the old process connects to the new one
1825using such a TCP connection to push all its entries before the new process
1826tries to connect to other peers. That ensures very fast replication during a
1827reload, it typically takes a fraction of a second even for large tables.
1828Note that Server IDs are used to identify servers remotely, so it is important
1829that configurations look similar or at least that the same IDs are forced on
1830each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02001831
1832peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001833 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001834 which is referenced by one or more stick-tables.
1835
Willy Tarreau77e4bd12015-05-01 20:02:17 +02001836disabled
1837 Disables a peers section. It disables both listening and any synchronization
1838 related to this section. This is provided to disable synchronization of stick
1839 tables without having to comment out all "peers" references.
1840
1841enable
1842 This re-enables a disabled peers section which was previously disabled.
1843
Emeric Brunf099e792010-09-27 12:05:28 +02001844peer <peername> <ip>:<port>
1845 Defines a peer inside a peers section.
1846 If <peername> is set to the local peer name (by default hostname, or forced
1847 using "-L" command line option), haproxy will listen for incoming remote peer
1848 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1849 to join the remote peer, and <peername> is used at the protocol level to
1850 identify and validate the remote peer on the server side.
1851
1852 During a soft restart, local peer <ip>:<port> is used by the old instance to
1853 connect the new one and initiate a complete replication (teaching process).
1854
1855 It is strongly recommended to have the exact same peers declaration on all
1856 peers and to only rely on the "-L" command line argument to change the local
1857 peer name. This makes it easier to maintain coherent configuration files
1858 across all peers.
1859
William Lallemandb2f07452015-05-12 14:27:13 +02001860 You may want to reference some environment variables in the address
1861 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001862
Cyril Bontédc4d9032012-04-08 21:57:39 +02001863 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001864 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001865 peer haproxy1 192.168.0.1:1024
1866 peer haproxy2 192.168.0.2:1024
1867 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001868
1869 backend mybackend
1870 mode tcp
1871 balance roundrobin
1872 stick-table type ip size 20k peers mypeers
1873 stick on src
1874
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001875 server srv1 192.168.0.30:80
1876 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001877
1878
Simon Horman51a1cf62015-02-03 13:00:44 +090018793.6. Mailers
1880------------
1881It is possible to send email alerts when the state of servers changes.
1882If configured email alerts are sent to each mailer that is configured
1883in a mailers section. Email is sent to mailers using SMTP.
1884
Pieter Baauw386a1272015-08-16 15:26:24 +02001885mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09001886 Creates a new mailer list with the name <mailersect>. It is an
1887 independent section which is referenced by one or more proxies.
1888
1889mailer <mailername> <ip>:<port>
1890 Defines a mailer inside a mailers section.
1891
1892 Example:
1893 mailers mymailers
1894 mailer smtp1 192.168.0.1:587
1895 mailer smtp2 192.168.0.2:587
1896
1897 backend mybackend
1898 mode tcp
1899 balance roundrobin
1900
1901 email-alert mailers mymailers
1902 email-alert from test1@horms.org
1903 email-alert to test2@horms.org
1904
1905 server srv1 192.168.0.30:80
1906 server srv2 192.168.0.31:80
1907
Pieter Baauw235fcfc2016-02-13 15:33:40 +01001908timeout mail <time>
1909 Defines the time available for a mail/connection to be made and send to
1910 the mail-server. If not defined the default value is 10 seconds. To allow
1911 for at least two SYN-ACK packets to be send during initial TCP handshake it
1912 is advised to keep this value above 4 seconds.
1913
1914 Example:
1915 mailers mymailers
1916 timeout mail 20s
1917 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09001918
Willy Tarreauc57f0e22009-05-10 13:12:33 +020019194. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001920----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001921
Willy Tarreau6a06a402007-07-15 20:15:28 +02001922Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02001923 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001924 - frontend <name>
1925 - backend <name>
1926 - listen <name>
1927
1928A "defaults" section sets default parameters for all other sections following
1929its declaration. Those default parameters are reset by the next "defaults"
1930section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001931section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001932
1933A "frontend" section describes a set of listening sockets accepting client
1934connections.
1935
1936A "backend" section describes a set of servers to which the proxy will connect
1937to forward incoming connections.
1938
1939A "listen" section defines a complete proxy with its frontend and backend
1940parts combined in one section. It is generally useful for TCP-only traffic.
1941
Willy Tarreau0ba27502007-12-24 16:55:16 +01001942All proxy names must be formed from upper and lower case letters, digits,
1943'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1944case-sensitive, which means that "www" and "WWW" are two different proxies.
1945
1946Historically, all proxy names could overlap, it just caused troubles in the
1947logs. Since the introduction of content switching, it is mandatory that two
1948proxies with overlapping capabilities (frontend/backend) have different names.
1949However, it is still permitted that a frontend and a backend share the same
1950name, as this configuration seems to be commonly encountered.
1951
1952Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1953and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001954bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001955protocol, and can interact with it by allowing, blocking, switching, adding,
1956modifying, or removing arbitrary contents in requests or responses, based on
1957arbitrary criteria.
1958
Willy Tarreau70dffda2014-01-30 03:07:23 +01001959In HTTP mode, the processing applied to requests and responses flowing over
1960a connection depends in the combination of the frontend's HTTP options and
1961the backend's. HAProxy supports 5 connection modes :
1962
1963 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
1964 requests and responses are processed, and connections remain open but idle
1965 between responses and new requests.
1966
1967 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
1968 1.0 to 1.5-dev21 : only the first request and response are processed, and
1969 everything else is forwarded with no analysis at all. This mode should not
1970 be used as it creates lots of trouble with logging and HTTP processing.
1971
1972 - PCL: passive close ("option httpclose") : exactly the same as tunnel mode,
1973 but with "Connection: close" appended in both directions to try to make
1974 both ends close after the first request/response exchange.
1975
1976 - SCL: server close ("option http-server-close") : the server-facing
1977 connection is closed after the end of the response is received, but the
1978 client-facing connection remains open.
1979
1980 - FCL: forced close ("option forceclose") : the connection is actively closed
1981 after the end of the response.
1982
1983The effective mode that will be applied to a connection passing through a
1984frontend and a backend can be determined by both proxy modes according to the
1985following matrix, but in short, the modes are symmetric, keep-alive is the
1986weakest option and force close is the strongest.
1987
1988 Backend mode
1989
1990 | KAL | TUN | PCL | SCL | FCL
1991 ----+-----+-----+-----+-----+----
1992 KAL | KAL | TUN | PCL | SCL | FCL
1993 ----+-----+-----+-----+-----+----
1994 TUN | TUN | TUN | PCL | SCL | FCL
1995 Frontend ----+-----+-----+-----+-----+----
1996 mode PCL | PCL | PCL | PCL | FCL | FCL
1997 ----+-----+-----+-----+-----+----
1998 SCL | SCL | SCL | FCL | SCL | FCL
1999 ----+-----+-----+-----+-----+----
2000 FCL | FCL | FCL | FCL | FCL | FCL
2001
Willy Tarreau0ba27502007-12-24 16:55:16 +01002002
Willy Tarreau70dffda2014-01-30 03:07:23 +01002003
Willy Tarreauc57f0e22009-05-10 13:12:33 +020020044.1. Proxy keywords matrix
2005--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002006
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002007The following list of keywords is supported. Most of them may only be used in a
2008limited set of section types. Some of them are marked as "deprecated" because
2009they are inherited from an old syntax which may be confusing or functionally
2010limited, and there are new recommended keywords to replace them. Keywords
Davor Ocelice9ed2812017-12-25 17:49:28 +01002011marked with "(*)" can be optionally inverted using the "no" prefix, e.g. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002012option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02002013and must be disabled for a specific instance. Such options may also be prefixed
2014with "default" in order to restore default settings regardless of what has been
2015specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002016
Willy Tarreau6a06a402007-07-15 20:15:28 +02002017
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002018 keyword defaults frontend listen backend
2019------------------------------------+----------+----------+---------+---------
2020acl - X X X
Willy Tarreau294d0f02015-08-10 19:40:12 +02002021appsession - - - -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002022backlog X X X -
2023balance X - X X
2024bind - X X -
2025bind-process X X X X
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002026block (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002027capture cookie - X X -
2028capture request header - X X -
2029capture response header - X X -
2030clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02002031compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002032contimeout (deprecated) X - X X
2033cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002034declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002035default-server X - X X
2036default_backend X X X -
2037description - X X X
2038disabled X X X X
2039dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002040email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09002041email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002042email-alert mailers X X X X
2043email-alert myhostname X X X X
2044email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002045enabled X X X X
2046errorfile X X X X
2047errorloc X X X X
2048errorloc302 X X X X
2049-- keyword -------------------------- defaults - frontend - listen -- backend -
2050errorloc303 X X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002051force-persist - - X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002052filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002053fullconn X - X X
2054grace X X X X
2055hash-type X - X X
2056http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01002057http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02002058http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002059http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002060http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02002061http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02002062http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002063id - X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002064ignore-persist - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002065load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02002066log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01002067log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02002068log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01002069log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02002070max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002071maxconn X X X -
2072mode X X X X
2073monitor fail - X X -
2074monitor-net X X X -
2075monitor-uri X X X -
2076option abortonclose (*) X - X X
2077option accept-invalid-http-request (*) X X X -
2078option accept-invalid-http-response (*) X - X X
2079option allbackups (*) X - X X
2080option checkcache (*) X - X X
2081option clitcpka (*) X X X -
2082option contstats (*) X X X -
2083option dontlog-normal (*) X X X -
2084option dontlognull (*) X X X -
2085option forceclose (*) X X X X
2086-- keyword -------------------------- defaults - frontend - listen -- backend -
2087option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02002088option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02002089option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01002090option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02002091option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02002092option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002093option http-server-close (*) X X X X
Willy Tarreau02bce8b2014-01-30 00:15:28 +01002094option http-tunnel (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002095option http-use-proxy-header (*) X X X -
2096option httpchk X - X X
2097option httpclose (*) X X X X
2098option httplog X X X X
2099option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002100option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02002101option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002102option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002103option log-health-checks (*) X - X X
2104option log-separate-errors (*) X X X -
2105option logasap (*) X X X -
2106option mysql-check X - X X
2107option nolinger (*) X X X X
2108option originalto X X X X
2109option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02002110option pgsql-check X - X X
2111option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002112option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02002113option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002114option smtpchk X - X X
2115option socket-stats (*) X X X -
2116option splice-auto (*) X X X X
2117option splice-request (*) X X X X
2118option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01002119option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002120option srvtcpka (*) X - X X
2121option ssl-hello-chk X - X X
2122-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01002123option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002124option tcp-smart-accept (*) X X X -
2125option tcp-smart-connect (*) X - X X
2126option tcpka X X X X
2127option tcplog X X X X
2128option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002129external-check command X - X X
2130external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002131persist rdp-cookie X - X X
2132rate-limit sessions X X X -
2133redirect - X X X
2134redisp (deprecated) X - X X
2135redispatch (deprecated) X - X X
2136reqadd - X X X
2137reqallow - X X X
2138reqdel - X X X
2139reqdeny - X X X
2140reqiallow - X X X
2141reqidel - X X X
2142reqideny - X X X
2143reqipass - X X X
2144reqirep - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002145reqitarpit - X X X
2146reqpass - X X X
2147reqrep - X X X
2148-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002149reqtarpit - X X X
2150retries X - X X
2151rspadd - X X X
2152rspdel - X X X
2153rspdeny - X X X
2154rspidel - X X X
2155rspideny - X X X
2156rspirep - X X X
2157rsprep - X X X
2158server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002159server-state-file-name X - X X
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02002160server-template - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002161source X - X X
2162srvtimeout (deprecated) X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02002163stats admin - X X X
2164stats auth X X X X
2165stats enable X X X X
2166stats hide-version X X X X
2167stats http-request - X X X
2168stats realm X X X X
2169stats refresh X X X X
2170stats scope X X X X
2171stats show-desc X X X X
2172stats show-legends X X X X
2173stats show-node X X X X
2174stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002175-- keyword -------------------------- defaults - frontend - listen -- backend -
2176stick match - - X X
2177stick on - - X X
2178stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02002179stick store-response - - X X
Adam Spiers68af3c12017-04-06 16:31:39 +01002180stick-table - X X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02002181tcp-check connect - - X X
2182tcp-check expect - - X X
2183tcp-check send - - X X
2184tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02002185tcp-request connection - X X -
2186tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02002187tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02002188tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02002189tcp-response content - - X X
2190tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002191timeout check X - X X
2192timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02002193timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002194timeout clitimeout (deprecated) X X X -
2195timeout connect X - X X
2196timeout contimeout (deprecated) X - X X
2197timeout http-keep-alive X X X X
2198timeout http-request X X X X
2199timeout queue X - X X
2200timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02002201timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002202timeout srvtimeout (deprecated) X - X X
2203timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02002204timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002205transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01002206unique-id-format X X X -
2207unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002208use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02002209use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002210------------------------------------+----------+----------+---------+---------
2211 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02002212
Willy Tarreau0ba27502007-12-24 16:55:16 +01002213
Willy Tarreauc57f0e22009-05-10 13:12:33 +020022144.2. Alphabetically sorted keywords reference
2215---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002216
2217This section provides a description of each keyword and its usage.
2218
2219
2220acl <aclname> <criterion> [flags] [operator] <value> ...
2221 Declare or complete an access list.
2222 May be used in sections : defaults | frontend | listen | backend
2223 no | yes | yes | yes
2224 Example:
2225 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2226 acl invalid_src src_port 0:1023
2227 acl local_dst hdr(host) -i localhost
2228
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002229 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002230
2231
Cyril Bontéb21570a2009-11-29 20:04:48 +01002232appsession <cookie> len <length> timeout <holdtime>
2233 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002234 Define session stickiness on an existing application cookie.
2235 May be used in sections : defaults | frontend | listen | backend
2236 no | no | yes | yes
2237 Arguments :
2238 <cookie> this is the name of the cookie used by the application and which
2239 HAProxy will have to learn for each new session.
2240
Cyril Bontéb21570a2009-11-29 20:04:48 +01002241 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01002242 checked in each cookie value.
2243
2244 <holdtime> this is the time after which the cookie will be removed from
2245 memory if unused. If no unit is specified, this time is in
2246 milliseconds.
2247
Cyril Bontébf47aeb2009-10-15 00:15:40 +02002248 request-learn
2249 If this option is specified, then haproxy will be able to learn
2250 the cookie found in the request in case the server does not
2251 specify any in response. This is typically what happens with
2252 PHPSESSID cookies, or when haproxy's session expires before
2253 the application's session and the correct server is selected.
2254 It is recommended to specify this option to improve reliability.
2255
Cyril Bontéb21570a2009-11-29 20:04:48 +01002256 prefix When this option is specified, haproxy will match on the cookie
2257 prefix (or URL parameter prefix). The appsession value is the
2258 data following this prefix.
2259
2260 Example :
2261 appsession ASPSESSIONID len 64 timeout 3h prefix
2262
Davor Ocelice9ed2812017-12-25 17:49:28 +01002263 This will match the cookie ASPSESSIONIDXXX=XXXX,
2264 the appsession value will be XXX=XXXX.
Cyril Bontéb21570a2009-11-29 20:04:48 +01002265
2266 mode This option allows to change the URL parser mode.
2267 2 modes are currently supported :
2268 - path-parameters :
2269 The parser looks for the appsession in the path parameters
2270 part (each parameter is separated by a semi-colon), which is
2271 convenient for JSESSIONID for example.
2272 This is the default mode if the option is not set.
2273 - query-string :
2274 In this mode, the parser will look for the appsession in the
2275 query string.
2276
Willy Tarreau294d0f02015-08-10 19:40:12 +02002277 As of version 1.6, appsessions was removed. It is more flexible and more
2278 convenient to use stick-tables instead, and stick-tables support multi-master
2279 replication and data conservation across reloads, which appsessions did not.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002280
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01002281 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
2282 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002283
2284
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002285backlog <conns>
2286 Give hints to the system about the approximate listen backlog desired size
2287 May be used in sections : defaults | frontend | listen | backend
2288 yes | yes | yes | no
2289 Arguments :
2290 <conns> is the number of pending connections. Depending on the operating
2291 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02002292 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002293
2294 In order to protect against SYN flood attacks, one solution is to increase
2295 the system's SYN backlog size. Depending on the system, sometimes it is just
2296 tunable via a system parameter, sometimes it is not adjustable at all, and
2297 sometimes the system relies on hints given by the application at the time of
2298 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
2299 to the listen() syscall. On systems which can make use of this value, it can
2300 sometimes be useful to be able to specify a different value, hence this
2301 backlog parameter.
2302
2303 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
2304 used as a hint and the system accepts up to the smallest greater power of
2305 two, and never more than some limits (usually 32768).
2306
2307 See also : "maxconn" and the target operating system's tuning guide.
2308
2309
Willy Tarreau0ba27502007-12-24 16:55:16 +01002310balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02002311balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002312 Define the load balancing algorithm to be used in a backend.
2313 May be used in sections : defaults | frontend | listen | backend
2314 yes | no | yes | yes
2315 Arguments :
2316 <algorithm> is the algorithm used to select a server when doing load
2317 balancing. This only applies when no persistence information
2318 is available, or when a connection is redispatched to another
2319 server. <algorithm> may be one of the following :
2320
2321 roundrobin Each server is used in turns, according to their weights.
2322 This is the smoothest and fairest algorithm when the server's
2323 processing time remains equally distributed. This algorithm
2324 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002325 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002326 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002327 large farms, when a server becomes up after having been down
2328 for a very short time, it may sometimes take a few hundreds
2329 requests for it to be re-integrated into the farm and start
2330 receiving traffic. This is normal, though very rare. It is
2331 indicated here in case you would have the chance to observe
2332 it, so that you don't worry.
2333
2334 static-rr Each server is used in turns, according to their weights.
2335 This algorithm is as similar to roundrobin except that it is
2336 static, which means that changing a server's weight on the
2337 fly will have no effect. On the other hand, it has no design
2338 limitation on the number of servers, and when a server goes
2339 up, it is always immediately reintroduced into the farm, once
2340 the full map is recomputed. It also uses slightly less CPU to
2341 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002342
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002343 leastconn The server with the lowest number of connections receives the
2344 connection. Round-robin is performed within groups of servers
2345 of the same load to ensure that all servers will be used. Use
2346 of this algorithm is recommended where very long sessions are
2347 expected, such as LDAP, SQL, TSE, etc... but is not very well
2348 suited for protocols using short sessions such as HTTP. This
2349 algorithm is dynamic, which means that server weights may be
2350 adjusted on the fly for slow starts for instance.
2351
Willy Tarreauf09c6602012-02-13 17:12:08 +01002352 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002353 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002354 identifier to the highest (see server parameter "id"), which
2355 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002356 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002357 not make sense to use this algorithm without setting maxconn.
2358 The purpose of this algorithm is to always use the smallest
2359 number of servers so that extra servers can be powered off
2360 during non-intensive hours. This algorithm ignores the server
2361 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002362 or IMAP than HTTP, though it can be useful there too. In
2363 order to use this algorithm efficiently, it is recommended
2364 that a cloud controller regularly checks server usage to turn
2365 them off when unused, and regularly checks backend queue to
2366 turn new servers on when the queue inflates. Alternatively,
2367 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002368
Willy Tarreau0ba27502007-12-24 16:55:16 +01002369 source The source IP address is hashed and divided by the total
2370 weight of the running servers to designate which server will
2371 receive the request. This ensures that the same client IP
2372 address will always reach the same server as long as no
2373 server goes down or up. If the hash result changes due to the
2374 number of running servers changing, many clients will be
2375 directed to a different server. This algorithm is generally
2376 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002377 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002378 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002379 static by default, which means that changing a server's
2380 weight on the fly will have no effect, but this can be
2381 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002382
Oskar Stolc8dc41842012-05-19 10:19:54 +01002383 uri This algorithm hashes either the left part of the URI (before
2384 the question mark) or the whole URI (if the "whole" parameter
2385 is present) and divides the hash value by the total weight of
2386 the running servers. The result designates which server will
2387 receive the request. This ensures that the same URI will
2388 always be directed to the same server as long as no server
2389 goes up or down. This is used with proxy caches and
2390 anti-virus proxies in order to maximize the cache hit rate.
2391 Note that this algorithm may only be used in an HTTP backend.
2392 This algorithm is static by default, which means that
2393 changing a server's weight on the fly will have no effect,
2394 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002395
Oskar Stolc8dc41842012-05-19 10:19:54 +01002396 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002397 "depth", both followed by a positive integer number. These
2398 options may be helpful when it is needed to balance servers
2399 based on the beginning of the URI only. The "len" parameter
2400 indicates that the algorithm should only consider that many
2401 characters at the beginning of the URI to compute the hash.
2402 Note that having "len" set to 1 rarely makes sense since most
2403 URIs start with a leading "/".
2404
2405 The "depth" parameter indicates the maximum directory depth
2406 to be used to compute the hash. One level is counted for each
2407 slash in the request. If both parameters are specified, the
2408 evaluation stops when either is reached.
2409
Willy Tarreau0ba27502007-12-24 16:55:16 +01002410 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002411 the query string of each HTTP GET request.
2412
2413 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002414 request entity will be searched for the parameter argument,
2415 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002416 ('?') in the URL. The message body will only start to be
2417 analyzed once either the advertised amount of data has been
2418 received or the request buffer is full. In the unlikely event
2419 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002420 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002421 be randomly balanced if at all. This keyword used to support
2422 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002423
2424 If the parameter is found followed by an equal sign ('=') and
2425 a value, then the value is hashed and divided by the total
2426 weight of the running servers. The result designates which
2427 server will receive the request.
2428
2429 This is used to track user identifiers in requests and ensure
2430 that a same user ID will always be sent to the same server as
2431 long as no server goes up or down. If no value is found or if
2432 the parameter is not found, then a round robin algorithm is
2433 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002434 backend. This algorithm is static by default, which means
2435 that changing a server's weight on the fly will have no
2436 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002437
Cyril Bontédc4d9032012-04-08 21:57:39 +02002438 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2439 request. Just as with the equivalent ACL 'hdr()' function,
2440 the header name in parenthesis is not case sensitive. If the
2441 header is absent or if it does not contain any value, the
2442 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002443
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002444 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002445 reducing the hash algorithm to the main domain part with some
2446 specific headers such as 'Host'. For instance, in the Host
2447 value "haproxy.1wt.eu", only "1wt" will be considered.
2448
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002449 This algorithm is static by default, which means that
2450 changing a server's weight on the fly will have no effect,
2451 but this can be changed using "hash-type".
2452
Willy Tarreau760e81d2018-05-03 07:20:40 +02002453 random A random number will be used as the key for the consistent
2454 hashing function. This means that the servers' weights are
2455 respected, dynamic weight changes immediately take effect, as
2456 well as new server additions. Random load balancing can be
2457 useful with large farms or when servers are frequently added
2458 or removed. The hash-balance-factor directive can be used to
2459 further improve fairness of the load balancing, especially
2460 in situations where servers show highly variable response
2461 times.
2462
Emeric Brun736aa232009-06-30 17:56:00 +02002463 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002464 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002465 The RDP cookie <name> (or "mstshash" if omitted) will be
2466 looked up and hashed for each incoming TCP request. Just as
2467 with the equivalent ACL 'req_rdp_cookie()' function, the name
2468 is not case-sensitive. This mechanism is useful as a degraded
2469 persistence mode, as it makes it possible to always send the
2470 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002471 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002472 used instead.
2473
2474 Note that for this to work, the frontend must ensure that an
2475 RDP cookie is already present in the request buffer. For this
2476 you must use 'tcp-request content accept' rule combined with
2477 a 'req_rdp_cookie_cnt' ACL.
2478
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002479 This algorithm is static by default, which means that
2480 changing a server's weight on the fly will have no effect,
2481 but this can be changed using "hash-type".
2482
Cyril Bontédc4d9032012-04-08 21:57:39 +02002483 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002484
Willy Tarreau0ba27502007-12-24 16:55:16 +01002485 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002486 algorithms. Right now, only "url_param" and "uri" support an
2487 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002488
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002489 The load balancing algorithm of a backend is set to roundrobin when no other
2490 algorithm, mode nor option have been set. The algorithm may only be set once
2491 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002492
2493 Examples :
2494 balance roundrobin
2495 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002496 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002497 balance hdr(User-Agent)
2498 balance hdr(host)
2499 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002500
2501 Note: the following caveats and limitations on using the "check_post"
2502 extension with "url_param" must be considered :
2503
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002504 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002505 to determine if the parameters will be found in the body or entity which
2506 may contain binary data. Therefore another method may be required to
2507 restrict consideration of POST requests that have no URL parameters in
2508 the body. (see acl reqideny http_end)
2509
2510 - using a <max_wait> value larger than the request buffer size does not
2511 make sense and is useless. The buffer size is set at build time, and
2512 defaults to 16 kB.
2513
2514 - Content-Encoding is not supported, the parameter search will probably
2515 fail; and load balancing will fall back to Round Robin.
2516
2517 - Expect: 100-continue is not supported, load balancing will fall back to
2518 Round Robin.
2519
Lukas Tribus23953682017-04-28 13:24:30 +00002520 - Transfer-Encoding (RFC7230 3.3.1) is only supported in the first chunk.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002521 If the entire parameter value is not present in the first chunk, the
2522 selection of server is undefined (actually, defined by how little
2523 actually appeared in the first chunk).
2524
2525 - This feature does not support generation of a 100, 411 or 501 response.
2526
2527 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002528 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002529 white space or control characters are found, indicating the end of what
2530 might be a URL parameter list. This is probably not a concern with SGML
2531 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002532
Willy Tarreau294d0f02015-08-10 19:40:12 +02002533 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002534
2535
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002536bind [<address>]:<port_range> [, ...] [param*]
2537bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002538 Define one or several listening addresses and/or ports in a frontend.
2539 May be used in sections : defaults | frontend | listen | backend
2540 no | yes | yes | no
2541 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002542 <address> is optional and can be a host name, an IPv4 address, an IPv6
2543 address, or '*'. It designates the address the frontend will
2544 listen on. If unset, all IPv4 addresses of the system will be
2545 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002546 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002547 Optionally, an address family prefix may be used before the
2548 address to force the family regardless of the address format,
2549 which can be useful to specify a path to a unix socket with
2550 no slash ('/'). Currently supported prefixes are :
2551 - 'ipv4@' -> address is always IPv4
2552 - 'ipv6@' -> address is always IPv6
2553 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002554 - 'abns@' -> address is in abstract namespace (Linux only).
2555 Note: since abstract sockets are not "rebindable", they
2556 do not cope well with multi-process mode during
2557 soft-restart, so it is better to avoid them if
2558 nbproc is greater than 1. The effect is that if the
2559 new process fails to start, only one of the old ones
2560 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002561 - 'fd@<n>' -> use file descriptor <n> inherited from the
2562 parent. The fd must be bound and may or may not already
2563 be listening.
William Lallemandb2f07452015-05-12 14:27:13 +02002564 You may want to reference some environment variables in the
2565 address parameter, see section 2.3 about environment
2566 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002567
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002568 <port_range> is either a unique TCP port, or a port range for which the
2569 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002570 above. The port is mandatory for TCP listeners. Note that in
2571 the case of an IPv6 address, the port is always the number
2572 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002573 - a numerical port (ex: '80')
2574 - a dash-delimited ports range explicitly stating the lower
2575 and upper bounds (ex: '2000-2100') which are included in
2576 the range.
2577
2578 Particular care must be taken against port ranges, because
2579 every <address:port> couple consumes one socket (= a file
2580 descriptor), so it's easy to consume lots of descriptors
2581 with a simple range, and to run out of sockets. Also, each
2582 <address:port> couple must be used only once among all
2583 instances running on a same system. Please note that binding
2584 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002585 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002586 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002587
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002588 <path> is a UNIX socket path beginning with a slash ('/'). This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002589 alternative to the TCP listening port. HAProxy will then
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002590 receive UNIX connections on the socket located at this place.
2591 The path must begin with a slash and by default is absolute.
2592 It can be relative to the prefix defined by "unix-bind" in
2593 the global section. Note that the total length of the prefix
2594 followed by the socket path cannot exceed some system limits
2595 for UNIX sockets, which commonly are set to 107 characters.
2596
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002597 <param*> is a list of parameters common to all sockets declared on the
2598 same line. These numerous parameters depend on OS and build
2599 options and have a complete section dedicated to them. Please
2600 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002601
Willy Tarreau0ba27502007-12-24 16:55:16 +01002602 It is possible to specify a list of address:port combinations delimited by
2603 commas. The frontend will then listen on all of these addresses. There is no
2604 fixed limit to the number of addresses and ports which can be listened on in
2605 a frontend, as well as there is no limit to the number of "bind" statements
2606 in a frontend.
2607
2608 Example :
2609 listen http_proxy
2610 bind :80,:443
2611 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002612 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002613
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002614 listen http_https_proxy
2615 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002616 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002617
Willy Tarreau24709282013-03-10 21:32:12 +01002618 listen http_https_proxy_explicit
2619 bind ipv6@:80
2620 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2621 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2622
Willy Tarreaudad36a32013-03-11 01:20:04 +01002623 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002624 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002625
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002626 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2627 sun_path length is used for the address length. Some other programs
2628 such as socat use the string length only by default. Pass the option
2629 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2630 make it compatible with HAProxy's.
2631
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002632 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002633 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002634
2635
Christopher Fauletff4121f2017-11-22 16:38:49 +01002636bind-process [ all | odd | even | <process_num>[-[<process_num>]] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002637 Limit visibility of an instance to a certain set of processes numbers.
2638 May be used in sections : defaults | frontend | listen | backend
2639 yes | yes | yes | yes
2640 Arguments :
2641 all All process will see this instance. This is the default. It
2642 may be used to override a default value.
2643
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002644 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002645 option may be combined with other numbers.
2646
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002647 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002648 option may be combined with other numbers. Do not use it
2649 with less than 2 processes otherwise some instances might be
2650 missing from all processes.
2651
Christopher Fauletff4121f2017-11-22 16:38:49 +01002652 process_num The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002653 whose values must all be between 1 and 32 or 64 depending on
Christopher Fauletff4121f2017-11-22 16:38:49 +01002654 the machine's word size. Ranges can be partially defined. The
2655 higher bound can be omitted. In such case, it is replaced by
2656 the corresponding maximum value. If a proxy is bound to
2657 process numbers greater than the configured global.nbproc, it
2658 will either be forced to process #1 if a single process was
Willy Tarreau102df612014-05-07 23:56:38 +02002659 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002660
2661 This keyword limits binding of certain instances to certain processes. This
2662 is useful in order not to have too many processes listening to the same
2663 ports. For instance, on a dual-core machine, it might make sense to set
2664 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2665 and 'even' instances.
2666
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002667 At the moment, it is not possible to reference more than 32 or 64 processes
2668 using this keyword, but this should be more than enough for most setups.
2669 Please note that 'all' really means all processes regardless of the machine's
2670 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002671
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002672 Each "bind" line may further be limited to a subset of the proxy's processes,
2673 please consult the "process" bind keyword in section 5.1.
2674
Willy Tarreaub369a042014-09-16 13:21:03 +02002675 When a frontend has no explicit "bind-process" line, it tries to bind to all
2676 the processes referenced by its "bind" lines. That means that frontends can
2677 easily adapt to their listeners' processes.
2678
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002679 If some backends are referenced by frontends bound to other processes, the
2680 backend automatically inherits the frontend's processes.
2681
2682 Example :
2683 listen app_ip1
2684 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002685 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002686
2687 listen app_ip2
2688 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002689 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002690
2691 listen management
2692 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002693 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002694
Willy Tarreau110ecc12012-11-15 17:50:01 +01002695 listen management
2696 bind 10.0.0.4:80
2697 bind-process 1-4
2698
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002699 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002700
2701
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002702block { if | unless } <condition> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002703 Block a layer 7 request if/unless a condition is matched
2704 May be used in sections : defaults | frontend | listen | backend
2705 no | yes | yes | yes
2706
2707 The HTTP request will be blocked very early in the layer 7 processing
2708 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002709 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002710 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002711 conditions are met or not met. There is no fixed limit to the number of
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03002712 "block" statements per instance. To block connections at layer 4 (without
2713 sending a 403 error) see "tcp-request connection reject" and
2714 "tcp-request content reject" rules.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002715
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002716 This form is deprecated, do not use it in any new configuration, use the new
2717 "http-request deny" instead.
2718
Willy Tarreau0ba27502007-12-24 16:55:16 +01002719 Example:
2720 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2721 acl invalid_src src_port 0:1023
2722 acl local_dst hdr(host) -i localhost
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +03002723 # block is deprecated. Use http-request deny instead:
2724 #block if invalid_src || local_dst
2725 http-request deny if invalid_src || local_dst
Willy Tarreau0ba27502007-12-24 16:55:16 +01002726
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03002727 See also : section 7 about ACL usage, "http-request deny",
2728 "http-response deny", "tcp-request connection reject" and
2729 "tcp-request content reject".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002730
2731capture cookie <name> len <length>
2732 Capture and log a cookie in the request and in the response.
2733 May be used in sections : defaults | frontend | listen | backend
2734 no | yes | yes | no
2735 Arguments :
2736 <name> is the beginning of the name of the cookie to capture. In order
2737 to match the exact name, simply suffix the name with an equal
2738 sign ('='). The full name will appear in the logs, which is
2739 useful with application servers which adjust both the cookie name
Davor Ocelice9ed2812017-12-25 17:49:28 +01002740 and value (e.g. ASPSESSIONXXX).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002741
2742 <length> is the maximum number of characters to report in the logs, which
2743 include the cookie name, the equal sign and the value, all in the
2744 standard "name=value" form. The string will be truncated on the
2745 right if it exceeds <length>.
2746
2747 Only the first cookie is captured. Both the "cookie" request headers and the
2748 "set-cookie" response headers are monitored. This is particularly useful to
2749 check for application bugs causing session crossing or stealing between
2750 users, because generally the user's cookies can only change on a login page.
2751
2752 When the cookie was not presented by the client, the associated log column
2753 will report "-". When a request does not cause a cookie to be assigned by the
2754 server, a "-" is reported in the response column.
2755
2756 The capture is performed in the frontend only because it is necessary that
2757 the log format does not change for a given frontend depending on the
2758 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01002759 "capture cookie" statement in a frontend. The maximum capture length is set
2760 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
2761 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002762
2763 Example:
2764 capture cookie ASPSESSION len 32
2765
2766 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002767 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002768
2769
2770capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002771 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002772 May be used in sections : defaults | frontend | listen | backend
2773 no | yes | yes | no
2774 Arguments :
2775 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002776 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002777 appear in the requests, with the first letter of each word in
2778 upper case. The header name will not appear in the logs, only the
2779 value is reported, but the position in the logs is respected.
2780
2781 <length> is the maximum number of characters to extract from the value and
2782 report in the logs. The string will be truncated on the right if
2783 it exceeds <length>.
2784
Willy Tarreau4460d032012-11-21 23:37:37 +01002785 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002786 value will be added to the logs between braces ('{}'). If multiple headers
2787 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002788 in the same order they were declared in the configuration. Non-existent
2789 headers will be logged just as an empty string. Common uses for request
2790 header captures include the "Host" field in virtual hosting environments, the
2791 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002792 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002793 environments to find where the request came from.
2794
2795 Note that when capturing headers such as "User-agent", some spaces may be
2796 logged, making the log analysis more difficult. Thus be careful about what
2797 you log if you know your log parser is not smart enough to rely on the
2798 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002799
Willy Tarreau0900abb2012-11-22 00:21:46 +01002800 There is no limit to the number of captured request headers nor to their
2801 length, though it is wise to keep them low to limit memory usage per session.
2802 In order to keep log format consistent for a same frontend, header captures
2803 can only be declared in a frontend. It is not possible to specify a capture
2804 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002805
2806 Example:
2807 capture request header Host len 15
2808 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01002809 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01002810
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002811 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002812 about logging.
2813
2814
2815capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002816 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002817 May be used in sections : defaults | frontend | listen | backend
2818 no | yes | yes | no
2819 Arguments :
2820 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002821 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002822 appear in the response, with the first letter of each word in
2823 upper case. The header name will not appear in the logs, only the
2824 value is reported, but the position in the logs is respected.
2825
2826 <length> is the maximum number of characters to extract from the value and
2827 report in the logs. The string will be truncated on the right if
2828 it exceeds <length>.
2829
Willy Tarreau4460d032012-11-21 23:37:37 +01002830 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002831 result will be added to the logs between braces ('{}') after the captured
2832 request headers. If multiple headers are captured, they will be delimited by
2833 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002834 the configuration. Non-existent headers will be logged just as an empty
2835 string. Common uses for response header captures include the "Content-length"
2836 header which indicates how many bytes are expected to be returned, the
2837 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002838
Willy Tarreau0900abb2012-11-22 00:21:46 +01002839 There is no limit to the number of captured response headers nor to their
2840 length, though it is wise to keep them low to limit memory usage per session.
2841 In order to keep log format consistent for a same frontend, header captures
2842 can only be declared in a frontend. It is not possible to specify a capture
2843 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002844
2845 Example:
2846 capture response header Content-length len 9
2847 capture response header Location len 15
2848
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002849 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002850 about logging.
2851
2852
Cyril Bontéf0c60612010-02-06 14:44:47 +01002853clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002854 Set the maximum inactivity time on the client side.
2855 May be used in sections : defaults | frontend | listen | backend
2856 yes | yes | yes | no
2857 Arguments :
2858 <timeout> is the timeout value is specified in milliseconds by default, but
2859 can be in any other unit if the number is suffixed by the unit,
2860 as explained at the top of this document.
2861
2862 The inactivity timeout applies when the client is expected to acknowledge or
2863 send data. In HTTP mode, this timeout is particularly important to consider
2864 during the first phase, when the client sends the request, and during the
2865 response while it is reading data sent by the server. The value is specified
2866 in milliseconds by default, but can be in any other unit if the number is
2867 suffixed by the unit, as specified at the top of this document. In TCP mode
2868 (and to a lesser extent, in HTTP mode), it is highly recommended that the
2869 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002870 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01002871 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +01002872 (e.g. 4 or 5 seconds).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002873
2874 This parameter is specific to frontends, but can be specified once for all in
2875 "defaults" sections. This is in fact one of the easiest solutions not to
2876 forget about it. An unspecified timeout results in an infinite timeout, which
2877 is not recommended. Such a usage is accepted and works but reports a warning
2878 during startup because it may results in accumulation of expired sessions in
2879 the system if the system's timeouts are not configured either.
2880
2881 This parameter is provided for compatibility but is currently deprecated.
2882 Please use "timeout client" instead.
2883
Willy Tarreau036fae02008-01-06 13:24:40 +01002884 See also : "timeout client", "timeout http-request", "timeout server", and
2885 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002886
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002887compression algo <algorithm> ...
2888compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02002889compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02002890 Enable HTTP compression.
2891 May be used in sections : defaults | frontend | listen | backend
2892 yes | yes | yes | yes
2893 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002894 algo is followed by the list of supported compression algorithms.
2895 type is followed by the list of MIME types that will be compressed.
2896 offload makes haproxy work as a compression offloader only (see notes).
2897
2898 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01002899 identity this is mostly for debugging, and it was useful for developing
2900 the compression feature. Identity does not apply any change on
2901 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002902
Willy Tarreauc91840a2015-03-28 17:00:39 +01002903 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01002904 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01002905
2906 deflate same as "gzip", but with deflate algorithm and zlib format.
2907 Note that this algorithm has ambiguous support on many
2908 browsers and no support at all from recent ones. It is
2909 strongly recommended not to use it for anything else than
2910 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01002911 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002912
Willy Tarreauc91840a2015-03-28 17:00:39 +01002913 raw-deflate same as "deflate" without the zlib wrapper, and used as an
2914 alternative when the browser wants "deflate". All major
2915 browsers understand it and despite violating the standards,
2916 it is known to work better than "deflate", at least on MSIE
2917 and some versions of Safari. Do not use it in conjunction
2918 with "deflate", use either one or the other since both react
2919 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01002920 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002921
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002922 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002923 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002924 If backend servers support HTTP compression, these directives
2925 will be no-op: haproxy will see the compressed response and will not
2926 compress again. If backend servers do not support HTTP compression and
2927 there is Accept-Encoding header in request, haproxy will compress the
2928 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02002929
2930 The "offload" setting makes haproxy remove the Accept-Encoding header to
2931 prevent backend servers from compressing responses. It is strongly
2932 recommended not to do this because this means that all the compression work
2933 will be done on the single point where haproxy is located. However in some
2934 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002935 with broken HTTP compression implementation which can't be turned off.
2936 In that case haproxy can be used to prevent that gateway from emitting
2937 invalid payloads. In this case, simply removing the header in the
2938 configuration does not work because it applies before the header is parsed,
2939 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02002940 then be used for such scenarios. Note: for now, the "offload" setting is
2941 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02002942
William Lallemand05097442012-11-20 12:14:28 +01002943 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002944 * the request does not advertise a supported compression algorithm in the
2945 "Accept-Encoding" header
2946 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01002947 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01002948 * response header "Transfer-Encoding" contains "chunked" (Temporary
2949 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002950 * response contain neither a "Content-Length" header nor a
2951 "Transfer-Encoding" whose last value is "chunked"
2952 * response contains a "Content-Type" header whose first value starts with
2953 "multipart"
2954 * the response contains the "no-transform" value in the "Cache-control"
2955 header
2956 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2957 and later
2958 * The response contains a "Content-Encoding" header, indicating that the
2959 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002960
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002961 Note: The compression does not rewrite Etag headers, and does not emit the
2962 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002963
William Lallemand82fe75c2012-10-23 10:25:10 +02002964 Examples :
2965 compression algo gzip
2966 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002967
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002968
Cyril Bontéf0c60612010-02-06 14:44:47 +01002969contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002970 Set the maximum time to wait for a connection attempt to a server to succeed.
2971 May be used in sections : defaults | frontend | listen | backend
2972 yes | no | yes | yes
2973 Arguments :
2974 <timeout> is the timeout value is specified in milliseconds by default, but
2975 can be in any other unit if the number is suffixed by the unit,
2976 as explained at the top of this document.
2977
2978 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002979 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002980 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +01002981 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002982 connect timeout also presets the queue timeout to the same value if this one
2983 has not been specified. Historically, the contimeout was also used to set the
2984 tarpit timeout in a listen section, which is not possible in a pure frontend.
2985
2986 This parameter is specific to backends, but can be specified once for all in
2987 "defaults" sections. This is in fact one of the easiest solutions not to
2988 forget about it. An unspecified timeout results in an infinite timeout, which
2989 is not recommended. Such a usage is accepted and works but reports a warning
2990 during startup because it may results in accumulation of failed sessions in
2991 the system if the system's timeouts are not configured either.
2992
2993 This parameter is provided for backwards compatibility but is currently
2994 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2995 instead.
2996
2997 See also : "timeout connect", "timeout queue", "timeout tarpit",
2998 "timeout server", "contimeout".
2999
3000
Willy Tarreau55165fe2009-05-10 12:02:55 +02003001cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02003002 [ postonly ] [ preserve ] [ httponly ] [ secure ]
3003 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Olivier Houchard4e694042017-03-14 20:01:29 +01003004 [ dynamic ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01003005 Enable cookie-based persistence in a backend.
3006 May be used in sections : defaults | frontend | listen | backend
3007 yes | no | yes | yes
3008 Arguments :
3009 <name> is the name of the cookie which will be monitored, modified or
3010 inserted in order to bring persistence. This cookie is sent to
3011 the client via a "Set-Cookie" header in the response, and is
3012 brought back by the client in a "Cookie" header in all requests.
3013 Special care should be taken to choose a name which does not
3014 conflict with any likely application cookie. Also, if the same
Davor Ocelice9ed2812017-12-25 17:49:28 +01003015 backends are subject to be used by the same clients (e.g.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003016 HTTP/HTTPS), care should be taken to use different cookie names
3017 between all backends if persistence between them is not desired.
3018
3019 rewrite This keyword indicates that the cookie will be provided by the
3020 server and that haproxy will have to modify its value to set the
3021 server's identifier in it. This mode is handy when the management
3022 of complex combinations of "Set-cookie" and "Cache-control"
3023 headers is left to the application. The application can then
3024 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003025 cookie. Since all responses should be monitored, this mode
3026 doesn't work in HTTP tunnel mode. Unless the application
Davor Ocelice9ed2812017-12-25 17:49:28 +01003027 behavior is very complex and/or broken, it is advised not to
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003028 start with this mode for new deployments. This keyword is
3029 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003030
3031 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02003032 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003033
Willy Tarreaua79094d2010-08-31 22:54:15 +02003034 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003035 server. When used without the "preserve" option, if the server
3036 emits a cookie with the same name, it will be remove before
Davor Ocelice9ed2812017-12-25 17:49:28 +01003037 processing. For this reason, this mode can be used to upgrade
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003038 existing configurations running in the "rewrite" mode. The cookie
3039 will only be a session cookie and will not be stored on the
3040 client's disk. By default, unless the "indirect" option is added,
3041 the server will see the cookies emitted by the client. Due to
3042 caching effects, it is generally wise to add the "nocache" or
3043 "postonly" keywords (see below). The "insert" keyword is not
3044 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003045
3046 prefix This keyword indicates that instead of relying on a dedicated
3047 cookie for the persistence, an existing one will be completed.
3048 This may be needed in some specific environments where the client
3049 does not support more than one single cookie and the application
3050 already needs it. In this case, whenever the server sets a cookie
3051 named <name>, it will be prefixed with the server's identifier
3052 and a delimiter. The prefix will be removed from all client
3053 requests so that the server still finds the cookie it emitted.
3054 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003055 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02003056 not compatible with "rewrite" and "insert". Note: it is highly
3057 recommended not to use "indirect" with "prefix", otherwise server
3058 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003059
Willy Tarreaua79094d2010-08-31 22:54:15 +02003060 indirect When this option is specified, no cookie will be emitted to a
3061 client which already has a valid one for the server which has
3062 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003063 it will be removed, unless the "preserve" option is also set. In
3064 "insert" mode, this will additionally remove cookies from the
3065 requests transmitted to the server, making the persistence
3066 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02003067 Note: it is highly recommended not to use "indirect" with
3068 "prefix", otherwise server cookie updates would not be sent to
3069 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003070
3071 nocache This option is recommended in conjunction with the insert mode
3072 when there is a cache between the client and HAProxy, as it
3073 ensures that a cacheable response will be tagged non-cacheable if
3074 a cookie needs to be inserted. This is important because if all
3075 persistence cookies are added on a cacheable home page for
3076 instance, then all customers will then fetch the page from an
3077 outer cache and will all share the same persistence cookie,
3078 leading to one server receiving much more traffic than others.
3079 See also the "insert" and "postonly" options.
3080
3081 postonly This option ensures that cookie insertion will only be performed
3082 on responses to POST requests. It is an alternative to the
3083 "nocache" option, because POST responses are not cacheable, so
3084 this ensures that the persistence cookie will never get cached.
3085 Since most sites do not need any sort of persistence before the
3086 first POST which generally is a login request, this is a very
3087 efficient method to optimize caching without risking to find a
3088 persistence cookie in the cache.
3089 See also the "insert" and "nocache" options.
3090
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003091 preserve This option may only be used with "insert" and/or "indirect". It
3092 allows the server to emit the persistence cookie itself. In this
3093 case, if a cookie is found in the response, haproxy will leave it
3094 untouched. This is useful in order to end persistence after a
3095 logout request for instance. For this, the server just has to
Davor Ocelice9ed2812017-12-25 17:49:28 +01003096 emit a cookie with an invalid value (e.g. empty) or with a date in
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003097 the past. By combining this mechanism with the "disable-on-404"
3098 check option, it is possible to perform a completely graceful
3099 shutdown because users will definitely leave the server after
3100 they logout.
3101
Willy Tarreau4992dd22012-05-31 21:02:17 +02003102 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
3103 when a cookie is inserted. This attribute is used so that a
3104 user agent doesn't share the cookie with non-HTTP components.
3105 Please check RFC6265 for more information on this attribute.
3106
3107 secure This option tells haproxy to add a "Secure" cookie attribute when
3108 a cookie is inserted. This attribute is used so that a user agent
3109 never emits this cookie over non-secure channels, which means
3110 that a cookie learned with this flag will be presented only over
3111 SSL/TLS connections. Please check RFC6265 for more information on
3112 this attribute.
3113
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003114 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003115 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01003116 name. If the domain begins with a dot, the browser is allowed to
3117 use it for any host ending with that name. It is also possible to
3118 specify several domain names by invoking this option multiple
3119 times. Some browsers might have small limits on the number of
3120 domains, so be careful when doing that. For the record, sending
3121 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003122
Willy Tarreau996a92c2010-10-13 19:30:47 +02003123 maxidle This option allows inserted cookies to be ignored after some idle
3124 time. It only works with insert-mode cookies. When a cookie is
3125 sent to the client, the date this cookie was emitted is sent too.
3126 Upon further presentations of this cookie, if the date is older
3127 than the delay indicated by the parameter (in seconds), it will
3128 be ignored. Otherwise, it will be refreshed if needed when the
3129 response is sent to the client. This is particularly useful to
3130 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003131 too long on the same server (e.g. after a farm size change). When
Willy Tarreau996a92c2010-10-13 19:30:47 +02003132 this option is set and a cookie has no date, it is always
3133 accepted, but gets refreshed in the response. This maintains the
3134 ability for admins to access their sites. Cookies that have a
3135 date in the future further than 24 hours are ignored. Doing so
3136 lets admins fix timezone issues without risking kicking users off
3137 the site.
3138
3139 maxlife This option allows inserted cookies to be ignored after some life
3140 time, whether they're in use or not. It only works with insert
3141 mode cookies. When a cookie is first sent to the client, the date
3142 this cookie was emitted is sent too. Upon further presentations
3143 of this cookie, if the date is older than the delay indicated by
3144 the parameter (in seconds), it will be ignored. If the cookie in
3145 the request has no date, it is accepted and a date will be set.
3146 Cookies that have a date in the future further than 24 hours are
3147 ignored. Doing so lets admins fix timezone issues without risking
3148 kicking users off the site. Contrary to maxidle, this value is
3149 not refreshed, only the first visit date counts. Both maxidle and
3150 maxlife may be used at the time. This is particularly useful to
3151 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003152 too long on the same server (e.g. after a farm size change). This
Willy Tarreau996a92c2010-10-13 19:30:47 +02003153 is stronger than the maxidle method in that it forces a
3154 redispatch after some absolute delay.
3155
Olivier Houchard4e694042017-03-14 20:01:29 +01003156 dynamic Activate dynamic cookies. When used, a session cookie is
3157 dynamically created for each server, based on the IP and port
3158 of the server, and a secret key, specified in the
3159 "dynamic-cookie-key" backend directive.
3160 The cookie will be regenerated each time the IP address change,
3161 and is only generated for IPv4/IPv6.
3162
Willy Tarreau0ba27502007-12-24 16:55:16 +01003163 There can be only one persistence cookie per HTTP backend, and it can be
3164 declared in a defaults section. The value of the cookie will be the value
3165 indicated after the "cookie" keyword in a "server" statement. If no cookie
3166 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02003167
Willy Tarreau0ba27502007-12-24 16:55:16 +01003168 Examples :
3169 cookie JSESSIONID prefix
3170 cookie SRV insert indirect nocache
3171 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02003172 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01003173
Willy Tarreau294d0f02015-08-10 19:40:12 +02003174 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003175
Willy Tarreau983e01e2010-01-11 18:42:06 +01003176
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003177declare capture [ request | response ] len <length>
3178 Declares a capture slot.
3179 May be used in sections : defaults | frontend | listen | backend
3180 no | yes | yes | no
3181 Arguments:
3182 <length> is the length allowed for the capture.
3183
3184 This declaration is only available in the frontend or listen section, but the
3185 reserved slot can be used in the backends. The "request" keyword allocates a
3186 capture slot for use in the request, and "response" allocates a capture slot
3187 for use in the response.
3188
3189 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02003190 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003191 "http-request capture" and "http-response capture".
3192
3193
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003194default-server [param*]
3195 Change default options for a server in a backend
3196 May be used in sections : defaults | frontend | listen | backend
3197 yes | no | yes | yes
3198 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01003199 <param*> is a list of parameters for this server. The "default-server"
3200 keyword accepts an important number of options and has a complete
3201 section dedicated to it. Please refer to section 5 for more
3202 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003203
Willy Tarreau983e01e2010-01-11 18:42:06 +01003204 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003205 default-server inter 1000 weight 13
3206
3207 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01003208
Willy Tarreau983e01e2010-01-11 18:42:06 +01003209
Willy Tarreau0ba27502007-12-24 16:55:16 +01003210default_backend <backend>
3211 Specify the backend to use when no "use_backend" rule has been matched.
3212 May be used in sections : defaults | frontend | listen | backend
3213 yes | yes | yes | no
3214 Arguments :
3215 <backend> is the name of the backend to use.
3216
3217 When doing content-switching between frontend and backends using the
3218 "use_backend" keyword, it is often useful to indicate which backend will be
3219 used when no rule has matched. It generally is the dynamic backend which
3220 will catch all undetermined requests.
3221
Willy Tarreau0ba27502007-12-24 16:55:16 +01003222 Example :
3223
3224 use_backend dynamic if url_dyn
3225 use_backend static if url_css url_img extension_img
3226 default_backend dynamic
3227
Willy Tarreau98d04852015-05-26 12:18:29 +02003228 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003229
Willy Tarreau0ba27502007-12-24 16:55:16 +01003230
Baptiste Assmann27f51342013-10-09 06:51:49 +02003231description <string>
3232 Describe a listen, frontend or backend.
3233 May be used in sections : defaults | frontend | listen | backend
3234 no | yes | yes | yes
3235 Arguments : string
3236
3237 Allows to add a sentence to describe the related object in the HAProxy HTML
3238 stats page. The description will be printed on the right of the object name
3239 it describes.
3240 No need to backslash spaces in the <string> arguments.
3241
3242
Willy Tarreau0ba27502007-12-24 16:55:16 +01003243disabled
3244 Disable a proxy, frontend or backend.
3245 May be used in sections : defaults | frontend | listen | backend
3246 yes | yes | yes | yes
3247 Arguments : none
3248
3249 The "disabled" keyword is used to disable an instance, mainly in order to
3250 liberate a listening port or to temporarily disable a service. The instance
3251 will still be created and its configuration will be checked, but it will be
3252 created in the "stopped" state and will appear as such in the statistics. It
3253 will not receive any traffic nor will it send any health-checks or logs. It
3254 is possible to disable many instances at once by adding the "disabled"
3255 keyword in a "defaults" section.
3256
3257 See also : "enabled"
3258
3259
Willy Tarreau5ce94572010-06-07 14:35:41 +02003260dispatch <address>:<port>
3261 Set a default server address
3262 May be used in sections : defaults | frontend | listen | backend
3263 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003264 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02003265
3266 <address> is the IPv4 address of the default server. Alternatively, a
3267 resolvable hostname is supported, but this name will be resolved
3268 during start-up.
3269
3270 <ports> is a mandatory port specification. All connections will be sent
3271 to this port, and it is not permitted to use port offsets as is
3272 possible with normal servers.
3273
Willy Tarreau787aed52011-04-15 06:45:37 +02003274 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02003275 server can take the connection. In the past it was used to forward non
3276 persistent connections to an auxiliary load balancer. Due to its simple
3277 syntax, it has also been used for simple TCP relays. It is recommended not to
3278 use it for more clarity, and to use the "server" directive instead.
3279
3280 See also : "server"
3281
Olivier Houchard4e694042017-03-14 20:01:29 +01003282
3283dynamic-cookie-key <string>
3284 Set the dynamic cookie secret key for a backend.
3285 May be used in sections : defaults | frontend | listen | backend
3286 yes | no | yes | yes
3287 Arguments : The secret key to be used.
3288
3289 When dynamic cookies are enabled (see the "dynamic" directive for cookie),
Davor Ocelice9ed2812017-12-25 17:49:28 +01003290 a dynamic cookie is created for each server (unless one is explicitly
Olivier Houchard4e694042017-03-14 20:01:29 +01003291 specified on the "server" line), using a hash of the IP address of the
3292 server, the TCP port, and the secret key.
Davor Ocelice9ed2812017-12-25 17:49:28 +01003293 That way, we can ensure session persistence across multiple load-balancers,
Olivier Houchard4e694042017-03-14 20:01:29 +01003294 even if servers are dynamically added or removed.
Willy Tarreau5ce94572010-06-07 14:35:41 +02003295
Willy Tarreau0ba27502007-12-24 16:55:16 +01003296enabled
3297 Enable a proxy, frontend or backend.
3298 May be used in sections : defaults | frontend | listen | backend
3299 yes | yes | yes | yes
3300 Arguments : none
3301
3302 The "enabled" keyword is used to explicitly enable an instance, when the
3303 defaults has been set to "disabled". This is very rarely used.
3304
3305 See also : "disabled"
3306
3307
3308errorfile <code> <file>
3309 Return a file contents instead of errors generated by HAProxy
3310 May be used in sections : defaults | frontend | listen | backend
3311 yes | yes | yes | yes
3312 Arguments :
3313 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003314 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3315 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003316
3317 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003318 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01003319 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01003320 error pages, and to use absolute paths, since files are read
3321 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003322
3323 It is important to understand that this keyword is not meant to rewrite
3324 errors returned by the server, but errors detected and returned by HAProxy.
3325 This is why the list of supported errors is limited to a small set.
3326
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003327 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3328
Willy Tarreau0ba27502007-12-24 16:55:16 +01003329 The files are returned verbatim on the TCP socket. This allows any trick such
3330 as redirections to another URL or site, as well as tricks to clean cookies,
3331 force enable or disable caching, etc... The package provides default error
3332 files returning the same contents as default errors.
3333
Willy Tarreau59140a22009-02-22 12:02:30 +01003334 The files should not exceed the configured buffer size (BUFSIZE), which
3335 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
Davor Ocelice9ed2812017-12-25 17:49:28 +01003336 not to put any reference to local contents (e.g. images) in order to avoid
Willy Tarreau59140a22009-02-22 12:02:30 +01003337 loops between the client and HAProxy when all servers are down, causing an
3338 error to be returned instead of an image. For better HTTP compliance, it is
3339 recommended that all header lines end with CR-LF and not LF alone.
3340
Willy Tarreau0ba27502007-12-24 16:55:16 +01003341 The files are read at the same time as the configuration and kept in memory.
3342 For this reason, the errors continue to be returned even when the process is
3343 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01003344 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003345 403 status code and interrogating a blocked URL.
3346
3347 See also : "errorloc", "errorloc302", "errorloc303"
3348
Willy Tarreau59140a22009-02-22 12:02:30 +01003349 Example :
3350 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01003351 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01003352 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
3353 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
3354
Willy Tarreau2769aa02007-12-27 18:26:09 +01003355
3356errorloc <code> <url>
3357errorloc302 <code> <url>
3358 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3359 May be used in sections : defaults | frontend | listen | backend
3360 yes | yes | yes | yes
3361 Arguments :
3362 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003363 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3364 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003365
3366 <url> it is the exact contents of the "Location" header. It may contain
3367 either a relative URI to an error page hosted on the same site,
3368 or an absolute URI designating an error page on another site.
3369 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003370 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003371
3372 It is important to understand that this keyword is not meant to rewrite
3373 errors returned by the server, but errors detected and returned by HAProxy.
3374 This is why the list of supported errors is limited to a small set.
3375
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003376 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3377
Willy Tarreau2769aa02007-12-27 18:26:09 +01003378 Note that both keyword return the HTTP 302 status code, which tells the
3379 client to fetch the designated URL using the same HTTP method. This can be
3380 quite problematic in case of non-GET methods such as POST, because the URL
3381 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003382 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003383 status code, indicating to the client that the URL must be fetched with a GET
3384 request.
3385
3386 See also : "errorfile", "errorloc303"
3387
3388
3389errorloc303 <code> <url>
3390 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3391 May be used in sections : defaults | frontend | listen | backend
3392 yes | yes | yes | yes
3393 Arguments :
3394 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003395 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3396 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003397
3398 <url> it is the exact contents of the "Location" header. It may contain
3399 either a relative URI to an error page hosted on the same site,
3400 or an absolute URI designating an error page on another site.
3401 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003402 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003403
3404 It is important to understand that this keyword is not meant to rewrite
3405 errors returned by the server, but errors detected and returned by HAProxy.
3406 This is why the list of supported errors is limited to a small set.
3407
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003408 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3409
Willy Tarreau2769aa02007-12-27 18:26:09 +01003410 Note that both keyword return the HTTP 303 status code, which tells the
3411 client to fetch the designated URL using the same HTTP GET method. This
3412 solves the usual problems associated with "errorloc" and the 302 code. It is
3413 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003414 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003415
3416 See also : "errorfile", "errorloc", "errorloc302"
3417
3418
Simon Horman51a1cf62015-02-03 13:00:44 +09003419email-alert from <emailaddr>
3420 Declare the from email address to be used in both the envelope and header
Davor Ocelice9ed2812017-12-25 17:49:28 +01003421 of email alerts. This is the address that email alerts are sent from.
Simon Horman51a1cf62015-02-03 13:00:44 +09003422 May be used in sections: defaults | frontend | listen | backend
3423 yes | yes | yes | yes
3424
3425 Arguments :
3426
3427 <emailaddr> is the from email address to use when sending email alerts
3428
3429 Also requires "email-alert mailers" and "email-alert to" to be set
3430 and if so sending email alerts is enabled for the proxy.
3431
Simon Horman64e34162015-02-06 11:11:57 +09003432 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003433 "email-alert myhostname", "email-alert to", section 3.6 about
3434 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003435
3436
3437email-alert level <level>
3438 Declare the maximum log level of messages for which email alerts will be
3439 sent. This acts as a filter on the sending of email alerts.
3440 May be used in sections: defaults | frontend | listen | backend
3441 yes | yes | yes | yes
3442
3443 Arguments :
3444
3445 <level> One of the 8 syslog levels:
3446 emerg alert crit err warning notice info debug
3447 The above syslog levels are ordered from lowest to highest.
3448
3449 By default level is alert
3450
3451 Also requires "email-alert from", "email-alert mailers" and
3452 "email-alert to" to be set and if so sending email alerts is enabled
3453 for the proxy.
3454
Simon Horman1421e212015-04-30 13:10:35 +09003455 Alerts are sent when :
3456
3457 * An un-paused server is marked as down and <level> is alert or lower
3458 * A paused server is marked as down and <level> is notice or lower
3459 * A server is marked as up or enters the drain state and <level>
3460 is notice or lower
3461 * "option log-health-checks" is enabled, <level> is info or lower,
3462 and a health check status update occurs
3463
Simon Horman64e34162015-02-06 11:11:57 +09003464 See also : "email-alert from", "email-alert mailers",
3465 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003466 section 3.6 about mailers.
3467
3468
3469email-alert mailers <mailersect>
3470 Declare the mailers to be used when sending email alerts
3471 May be used in sections: defaults | frontend | listen | backend
3472 yes | yes | yes | yes
3473
3474 Arguments :
3475
3476 <mailersect> is the name of the mailers section to send email alerts.
3477
3478 Also requires "email-alert from" and "email-alert to" to be set
3479 and if so sending email alerts is enabled for the proxy.
3480
Simon Horman64e34162015-02-06 11:11:57 +09003481 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3482 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003483
3484
3485email-alert myhostname <hostname>
3486 Declare the to hostname address to be used when communicating with
3487 mailers.
3488 May be used in sections: defaults | frontend | listen | backend
3489 yes | yes | yes | yes
3490
3491 Arguments :
3492
Baptiste Assmann738bad92015-12-21 15:27:53 +01003493 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003494
3495 By default the systems hostname is used.
3496
3497 Also requires "email-alert from", "email-alert mailers" and
3498 "email-alert to" to be set and if so sending email alerts is enabled
3499 for the proxy.
3500
Simon Horman64e34162015-02-06 11:11:57 +09003501 See also : "email-alert from", "email-alert level", "email-alert mailers",
3502 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003503
3504
3505email-alert to <emailaddr>
Davor Ocelice9ed2812017-12-25 17:49:28 +01003506 Declare both the recipient address in the envelope and to address in the
Simon Horman51a1cf62015-02-03 13:00:44 +09003507 header of email alerts. This is the address that email alerts are sent to.
3508 May be used in sections: defaults | frontend | listen | backend
3509 yes | yes | yes | yes
3510
3511 Arguments :
3512
3513 <emailaddr> is the to email address to use when sending email alerts
3514
3515 Also requires "email-alert mailers" and "email-alert to" to be set
3516 and if so sending email alerts is enabled for the proxy.
3517
Simon Horman64e34162015-02-06 11:11:57 +09003518 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003519 "email-alert myhostname", section 3.6 about mailers.
3520
3521
Willy Tarreau4de91492010-01-22 19:10:05 +01003522force-persist { if | unless } <condition>
3523 Declare a condition to force persistence on down servers
3524 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003525 no | no | yes | yes
Willy Tarreau4de91492010-01-22 19:10:05 +01003526
3527 By default, requests are not dispatched to down servers. It is possible to
3528 force this using "option persist", but it is unconditional and redispatches
3529 to a valid server if "option redispatch" is set. That leaves with very little
3530 possibilities to force some requests to reach a server which is artificially
3531 marked down for maintenance operations.
3532
3533 The "force-persist" statement allows one to declare various ACL-based
3534 conditions which, when met, will cause a request to ignore the down status of
3535 a server and still try to connect to it. That makes it possible to start a
3536 server, still replying an error to the health checks, and run a specially
3537 configured browser to test the service. Among the handy methods, one could
3538 use a specific source IP address, or a specific cookie. The cookie also has
3539 the advantage that it can easily be added/removed on the browser from a test
3540 page. Once the service is validated, it is then possible to open the service
3541 to the world by returning a valid response to health checks.
3542
3543 The forced persistence is enabled when an "if" condition is met, or unless an
3544 "unless" condition is met. The final redispatch is always disabled when this
3545 is used.
3546
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003547 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003548 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003549
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003550
3551filter <name> [param*]
3552 Add the filter <name> in the filter list attached to the proxy.
3553 May be used in sections : defaults | frontend | listen | backend
3554 no | yes | yes | yes
3555 Arguments :
3556 <name> is the name of the filter. Officially supported filters are
3557 referenced in section 9.
3558
Tim Düsterhus4896c442016-11-29 02:15:19 +01003559 <param*> is a list of parameters accepted by the filter <name>. The
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003560 parsing of these parameters are the responsibility of the
Tim Düsterhus4896c442016-11-29 02:15:19 +01003561 filter. Please refer to the documentation of the corresponding
3562 filter (section 9) for all details on the supported parameters.
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003563
3564 Multiple occurrences of the filter line can be used for the same proxy. The
3565 same filter can be referenced many times if needed.
3566
3567 Example:
3568 listen
3569 bind *:80
3570
3571 filter trace name BEFORE-HTTP-COMP
3572 filter compression
3573 filter trace name AFTER-HTTP-COMP
3574
3575 compression algo gzip
3576 compression offload
3577
3578 server srv1 192.168.0.1:80
3579
3580 See also : section 9.
3581
Willy Tarreau4de91492010-01-22 19:10:05 +01003582
Willy Tarreau2769aa02007-12-27 18:26:09 +01003583fullconn <conns>
3584 Specify at what backend load the servers will reach their maxconn
3585 May be used in sections : defaults | frontend | listen | backend
3586 yes | no | yes | yes
3587 Arguments :
3588 <conns> is the number of connections on the backend which will make the
3589 servers use the maximal number of connections.
3590
Willy Tarreau198a7442008-01-17 12:05:32 +01003591 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003592 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003593 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003594 load. The server will then always accept at least <minconn> connections,
3595 never more than <maxconn>, and the limit will be on the ramp between both
3596 values when the backend has less than <conns> concurrent connections. This
3597 makes it possible to limit the load on the servers during normal loads, but
3598 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003599 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003600
Willy Tarreaufbb78422011-06-05 15:38:35 +02003601 Since it's hard to get this value right, haproxy automatically sets it to
3602 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003603 backend (based on "use_backend" and "default_backend" rules). That way it's
3604 safe to leave it unset. However, "use_backend" involving dynamic names are
3605 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003606
Willy Tarreau2769aa02007-12-27 18:26:09 +01003607 Example :
3608 # The servers will accept between 100 and 1000 concurrent connections each
3609 # and the maximum of 1000 will be reached when the backend reaches 10000
3610 # connections.
3611 backend dynamic
3612 fullconn 10000
3613 server srv1 dyn1:80 minconn 100 maxconn 1000
3614 server srv2 dyn2:80 minconn 100 maxconn 1000
3615
3616 See also : "maxconn", "server"
3617
3618
3619grace <time>
3620 Maintain a proxy operational for some time after a soft stop
3621 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003622 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003623 Arguments :
3624 <time> is the time (by default in milliseconds) for which the instance
3625 will remain operational with the frontend sockets still listening
3626 when a soft-stop is received via the SIGUSR1 signal.
3627
3628 This may be used to ensure that the services disappear in a certain order.
3629 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003630 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003631 needed by the equipment to detect the failure.
3632
3633 Note that currently, there is very little benefit in using this parameter,
3634 and it may in fact complicate the soft-reconfiguration process more than
3635 simplify it.
3636
Willy Tarreau0ba27502007-12-24 16:55:16 +01003637
Andrew Rodland17be45e2016-10-25 17:04:12 -04003638hash-balance-factor <factor>
3639 Specify the balancing factor for bounded-load consistent hashing
3640 May be used in sections : defaults | frontend | listen | backend
3641 yes | no | no | yes
3642 Arguments :
3643 <factor> is the control for the maximum number of concurrent requests to
3644 send to a server, expressed as a percentage of the average number
3645 of concurrent requests across all of the active servers.
3646
3647 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
3648 enables an algorithm that prevents any one server from getting too many
3649 requests at once, even if some hash buckets receive many more requests than
3650 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
3651 <factor> is a percentage greater than 100. For example, if <factor> is 150,
3652 then no server will be allowed to have a load more than 1.5 times the average.
3653 If server weights are used, they will be respected.
3654
3655 If the first-choice server is disqualified, the algorithm will choose another
3656 server based on the request hash, until a server with additional capacity is
3657 found. A higher <factor> allows more imbalance between the servers, while a
3658 lower <factor> means that more servers will be checked on average, affecting
3659 performance. Reasonable values are from 125 to 200.
3660
Willy Tarreau760e81d2018-05-03 07:20:40 +02003661 This setting is also used by "balance random" which internally relies on the
3662 consistent hashing mechanism.
3663
Andrew Rodland17be45e2016-10-25 17:04:12 -04003664 See also : "balance" and "hash-type".
3665
3666
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003667hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003668 Specify a method to use for mapping hashes to servers
3669 May be used in sections : defaults | frontend | listen | backend
3670 yes | no | yes | yes
3671 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003672 <method> is the method used to select a server from the hash computed by
3673 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003674
Bhaskar98634f02013-10-29 23:30:51 -04003675 map-based the hash table is a static array containing all alive servers.
3676 The hashes will be very smooth, will consider weights, but
3677 will be static in that weight changes while a server is up
3678 will be ignored. This means that there will be no slow start.
3679 Also, since a server is selected by its position in the array,
3680 most mappings are changed when the server count changes. This
3681 means that when a server goes up or down, or when a server is
3682 added to a farm, most connections will be redistributed to
3683 different servers. This can be inconvenient with caches for
3684 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003685
Bhaskar98634f02013-10-29 23:30:51 -04003686 consistent the hash table is a tree filled with many occurrences of each
3687 server. The hash key is looked up in the tree and the closest
3688 server is chosen. This hash is dynamic, it supports changing
3689 weights while the servers are up, so it is compatible with the
3690 slow start feature. It has the advantage that when a server
3691 goes up or down, only its associations are moved. When a
3692 server is added to the farm, only a few part of the mappings
3693 are redistributed, making it an ideal method for caches.
3694 However, due to its principle, the distribution will never be
3695 very smooth and it may sometimes be necessary to adjust a
3696 server's weight or its ID to get a more balanced distribution.
3697 In order to get the same distribution on multiple load
3698 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003699 same IDs. Note: consistent hash uses sdbm and avalanche if no
3700 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003701
3702 <function> is the hash function to be used :
3703
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003704 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003705 reimplementation of ndbm) database library. It was found to do
3706 well in scrambling bits, causing better distribution of the keys
3707 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003708 function with good distribution, unless the total server weight
3709 is a multiple of 64, in which case applying the avalanche
3710 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04003711
3712 djb2 this function was first proposed by Dan Bernstein many years ago
3713 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003714 function provides a better distribution than sdbm. It generally
3715 works well with text-based inputs though it can perform extremely
3716 poorly with numeric-only input or when the total server weight is
3717 a multiple of 33, unless the avalanche modifier is also used.
3718
Willy Tarreaua0f42712013-11-14 14:30:35 +01003719 wt6 this function was designed for haproxy while testing other
3720 functions in the past. It is not as smooth as the other ones, but
3721 is much less sensible to the input data set or to the number of
3722 servers. It can make sense as an alternative to sdbm+avalanche or
3723 djb2+avalanche for consistent hashing or when hashing on numeric
3724 data such as a source IP address or a visitor identifier in a URL
3725 parameter.
3726
Willy Tarreau324f07f2015-01-20 19:44:50 +01003727 crc32 this is the most common CRC32 implementation as used in Ethernet,
3728 gzip, PNG, etc. It is slower than the other ones but may provide
3729 a better distribution or less predictable results especially when
3730 used on strings.
3731
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003732 <modifier> indicates an optional method applied after hashing the key :
3733
3734 avalanche This directive indicates that the result from the hash
3735 function above should not be used in its raw form but that
3736 a 4-byte full avalanche hash must be applied first. The
3737 purpose of this step is to mix the resulting bits from the
3738 previous hash in order to avoid any undesired effect when
3739 the input contains some limited values or when the number of
3740 servers is a multiple of one of the hash's components (64
3741 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
3742 result less predictable, but it's also not as smooth as when
3743 using the original function. Some testing might be needed
3744 with some workloads. This hash is one of the many proposed
3745 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003746
Bhaskar98634f02013-10-29 23:30:51 -04003747 The default hash type is "map-based" and is recommended for most usages. The
3748 default function is "sdbm", the selection of a function should be based on
3749 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003750
Andrew Rodland17be45e2016-10-25 17:04:12 -04003751 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003752
3753
Willy Tarreau0ba27502007-12-24 16:55:16 +01003754http-check disable-on-404
3755 Enable a maintenance mode upon HTTP/404 response to health-checks
3756 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01003757 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01003758 Arguments : none
3759
3760 When this option is set, a server which returns an HTTP code 404 will be
3761 excluded from further load-balancing, but will still receive persistent
3762 connections. This provides a very convenient method for Web administrators
3763 to perform a graceful shutdown of their servers. It is also important to note
3764 that a server which is detected as failed while it was in this mode will not
3765 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
3766 will immediately be reinserted into the farm. The status on the stats page
3767 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01003768 option only works in conjunction with the "httpchk" option. If this option
3769 is used with "http-check expect", then it has precedence over it so that 404
3770 responses will still be considered as soft-stop.
3771
3772 See also : "option httpchk", "http-check expect"
3773
3774
3775http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003776 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01003777 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02003778 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01003779 Arguments :
3780 <match> is a keyword indicating how to look for a specific pattern in the
3781 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003782 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01003783 exclamation mark ("!") to negate the match. Spaces are allowed
3784 between the exclamation mark and the keyword. See below for more
3785 details on the supported keywords.
3786
3787 <pattern> is the pattern to look for. It may be a string or a regular
3788 expression. If the pattern contains spaces, they must be escaped
3789 with the usual backslash ('\').
3790
3791 By default, "option httpchk" considers that response statuses 2xx and 3xx
3792 are valid, and that others are invalid. When "http-check expect" is used,
3793 it defines what is considered valid or invalid. Only one "http-check"
3794 statement is supported in a backend. If a server fails to respond or times
3795 out, the check obviously fails. The available matches are :
3796
3797 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003798 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003799 response's status code is exactly this string. If the
3800 "status" keyword is prefixed with "!", then the response
3801 will be considered invalid if the status code matches.
3802
3803 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003804 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003805 response's status code matches the expression. If the
3806 "rstatus" keyword is prefixed with "!", then the response
3807 will be considered invalid if the status code matches.
3808 This is mostly used to check for multiple codes.
3809
3810 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003811 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003812 response's body contains this exact string. If the
3813 "string" keyword is prefixed with "!", then the response
3814 will be considered invalid if the body contains this
3815 string. This can be used to look for a mandatory word at
3816 the end of a dynamic page, or to detect a failure when a
Davor Ocelice9ed2812017-12-25 17:49:28 +01003817 specific error appears on the check page (e.g. a stack
Willy Tarreaubd741542010-03-16 18:46:54 +01003818 trace).
3819
3820 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003821 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003822 response's body matches this expression. If the "rstring"
3823 keyword is prefixed with "!", then the response will be
3824 considered invalid if the body matches the expression.
3825 This can be used to look for a mandatory word at the end
3826 of a dynamic page, or to detect a failure when a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01003827 error appears on the check page (e.g. a stack trace).
Willy Tarreaubd741542010-03-16 18:46:54 +01003828
3829 It is important to note that the responses will be limited to a certain size
3830 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
3831 Thus, too large responses may not contain the mandatory pattern when using
3832 "string" or "rstring". If a large response is absolutely required, it is
3833 possible to change the default max size by setting the global variable.
3834 However, it is worth keeping in mind that parsing very large responses can
3835 waste some CPU cycles, especially when regular expressions are used, and that
3836 it is always better to focus the checks on smaller resources.
3837
Cyril Bonté32602d22015-01-30 00:07:07 +01003838 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
3839 will automatically append a "Connection: close" header, meaning that this
3840 header should not be present in the request provided by "option httpchk".
3841
Willy Tarreaubd741542010-03-16 18:46:54 +01003842 Last, if "http-check expect" is combined with "http-check disable-on-404",
3843 then this last one has precedence when the server responds with 404.
3844
3845 Examples :
3846 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003847 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01003848
3849 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003850 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01003851
3852 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003853 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01003854
3855 # check that we have a correct hexadecimal tag before /html
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03003856 http-check expect rstring <!--tag:[0-9a-f]*--></html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01003857
Willy Tarreaubd741542010-03-16 18:46:54 +01003858 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003859
3860
Willy Tarreauef781042010-01-27 11:53:01 +01003861http-check send-state
3862 Enable emission of a state header with HTTP health checks
3863 May be used in sections : defaults | frontend | listen | backend
3864 yes | no | yes | yes
3865 Arguments : none
3866
3867 When this option is set, haproxy will systematically send a special header
3868 "X-Haproxy-Server-State" with a list of parameters indicating to each server
3869 how they are seen by haproxy. This can be used for instance when a server is
3870 manipulated without access to haproxy and the operator needs to know whether
3871 haproxy still sees it up or not, or if the server is the last one in a farm.
3872
3873 The header is composed of fields delimited by semi-colons, the first of which
3874 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
3875 checks on the total number before transition, just as appears in the stats
3876 interface. Next headers are in the form "<variable>=<value>", indicating in
3877 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08003878 - a variable "address", containing the address of the backend server.
3879 This corresponds to the <address> field in the server declaration. For
3880 unix domain sockets, it will read "unix".
3881
3882 - a variable "port", containing the port of the backend server. This
3883 corresponds to the <port> field in the server declaration. For unix
3884 domain sockets, it will read "unix".
3885
Willy Tarreauef781042010-01-27 11:53:01 +01003886 - a variable "name", containing the name of the backend followed by a slash
3887 ("/") then the name of the server. This can be used when a server is
3888 checked in multiple backends.
3889
3890 - a variable "node" containing the name of the haproxy node, as set in the
3891 global "node" variable, otherwise the system's hostname if unspecified.
3892
3893 - a variable "weight" indicating the weight of the server, a slash ("/")
3894 and the total weight of the farm (just counting usable servers). This
3895 helps to know if other servers are available to handle the load when this
3896 one fails.
3897
3898 - a variable "scur" indicating the current number of concurrent connections
3899 on the server, followed by a slash ("/") then the total number of
3900 connections on all servers of the same backend.
3901
3902 - a variable "qcur" indicating the current number of requests in the
3903 server's queue.
3904
3905 Example of a header received by the application server :
3906 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
3907 scur=13/22; qcur=0
3908
3909 See also : "option httpchk", "http-check disable-on-404"
3910
Willy Tarreau53275e82017-11-24 07:52:01 +01003911http-request { allow | auth [realm <realm>] | redirect <rule> | reject |
Jarno Huuskonen800d1762017-03-06 14:56:36 +02003912 tarpit [deny_status <status>] | deny [deny_status <status>] |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003913 add-header <name> <fmt> | set-header <name> <fmt> |
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003914 capture <sample> [ len <length> | id <id> ] |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003915 del-header <name> | set-nice <nice> | set-log-level <level> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003916 replace-header <name> <match-regex> <replace-fmt> |
3917 replace-value <name> <match-regex> <replace-fmt> |
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003918 set-method <fmt> | set-path <fmt> | set-query <fmt> |
3919 set-uri <fmt> | set-tos <tos> | set-mark <mark> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003920 add-acl(<file name>) <key fmt> |
3921 del-acl(<file name>) <key fmt> |
3922 del-map(<file name>) <key fmt> |
Baptiste Assmannbb7e86a2014-09-03 18:29:47 +02003923 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003924 set-var(<var name>) <expr> |
Christopher Faulet85d79c92016-11-09 16:54:56 +01003925 unset-var(<var name>) |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003926 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003927 sc-inc-gpc0(<sc-id>) |
Frédéric Lécaille6778b272018-01-29 15:22:53 +01003928 sc-inc-gpc1(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003929 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003930 silent-drop |
William Lallemand86d0df02017-11-24 21:36:45 +01003931 send-spoe-group |
3932 cache-use
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003933 }
Cyril Bontéf0c60612010-02-06 14:44:47 +01003934 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003935 Access control for Layer 7 requests
3936
3937 May be used in sections: defaults | frontend | listen | backend
3938 no | yes | yes | yes
3939
Willy Tarreau20b0de52012-12-24 15:45:22 +01003940 The http-request statement defines a set of rules which apply to layer 7
3941 processing. The rules are evaluated in their declaration order when they are
3942 met in a frontend, listen or backend section. Any rule may optionally be
3943 followed by an ACL-based condition, in which case it will only be evaluated
3944 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003945
Willy Tarreau20b0de52012-12-24 15:45:22 +01003946 The first keyword is the rule's action. Currently supported actions include :
3947 - "allow" : this stops the evaluation of the rules and lets the request
3948 pass the check. No further "http-request" rules are evaluated.
3949
3950 - "deny" : this stops the evaluation of the rules and immediately rejects
Willy Tarreaube1d34d2016-06-26 19:37:59 +02003951 the request and emits an HTTP 403 error, or optionally the status code
3952 specified as an argument to "deny_status". The list of permitted status
3953 codes is limited to those that can be overridden by the "errorfile"
3954 directive. No further "http-request" rules are evaluated.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003955
Willy Tarreau53275e82017-11-24 07:52:01 +01003956 - "reject" : this stops the evaluation of the rules and immediately closes
3957 the connection without sending any response. It acts similarly to the
3958 "tcp-request content reject" rules. It can be useful to force an
3959 immediate connection closure on HTTP/2 connections.
3960
Willy Tarreauccbcc372012-12-27 12:37:57 +01003961 - "tarpit" : this stops the evaluation of the rules and immediately blocks
3962 the request without responding for a delay specified by "timeout tarpit"
3963 or "timeout connect" if the former is not set. After that delay, if the
Jarno Huuskonen800d1762017-03-06 14:56:36 +02003964 client is still connected, an HTTP error 500 (or optionally the status
3965 code specified as an argument to "deny_status") is returned so that the
Willy Tarreauccbcc372012-12-27 12:37:57 +01003966 client does not suspect it has been tarpitted. Logs will report the flags
3967 "PT". The goal of the tarpit rule is to slow down robots during an attack
3968 when they're limited on the number of concurrent requests. It can be very
3969 efficient against very dumb robots, and will significantly reduce the
3970 load on firewalls compared to a "deny" rule. But when facing "correctly"
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003971 developed robots, it can make things worse by forcing haproxy and the
Willy Tarreau2d392c22015-08-24 01:43:45 +02003972 front firewall to support insane number of concurrent connections. See
3973 also the "silent-drop" action below.
Willy Tarreauccbcc372012-12-27 12:37:57 +01003974
Willy Tarreau20b0de52012-12-24 15:45:22 +01003975 - "auth" : this stops the evaluation of the rules and immediately responds
3976 with an HTTP 401 or 407 error code to invite the user to present a valid
3977 user name and password. No further "http-request" rules are evaluated. An
3978 optional "realm" parameter is supported, it sets the authentication realm
3979 that is returned with the response (typically the application's name).
3980
Willy Tarreau81499eb2012-12-27 12:19:02 +01003981 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3982 This is exactly the same as the "redirect" statement except that it
3983 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01003984 "http-request" rules and that these rules use the "log-format" strings.
3985 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01003986
Willy Tarreau20b0de52012-12-24 15:45:22 +01003987 - "add-header" appends an HTTP header field whose name is specified in
3988 <name> and whose value is defined by <fmt> which follows the log-format
3989 rules (see Custom Log Format in section 8.2.4). This is particularly
Davor Ocelice9ed2812017-12-25 17:49:28 +01003990 useful to pass connection-specific information to the server (e.g. the
Willy Tarreau20b0de52012-12-24 15:45:22 +01003991 client's SSL certificate), or to combine several headers into one. This
3992 rule is not final, so it is possible to add other similar rules. Note
3993 that header addition is performed immediately, so one rule might reuse
3994 the resulting header from a previous rule.
3995
3996 - "set-header" does the same as "add-header" except that the header name
3997 is first removed if it existed. This is useful when passing security
3998 information to the server, where the header must not be manipulated by
Willy Tarreau85603282015-01-21 20:39:27 +01003999 external users. Note that the new value is computed before the removal so
4000 it is possible to concatenate a value to an existing header.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004001
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004002 - "del-header" removes all HTTP header fields whose name is specified in
4003 <name>.
4004
Sasha Pachev218f0642014-06-16 12:05:59 -06004005 - "replace-header" matches the regular expression in all occurrences of
4006 header field <name> according to <match-regex>, and replaces them with
4007 the <replace-fmt> argument. Format characters are allowed in replace-fmt
4008 and work like in <fmt> arguments in "add-header". The match is only
4009 case-sensitive. It is important to understand that this action only
4010 considers whole header lines, regardless of the number of values they
4011 may contain. This usage is suited to headers naturally containing commas
4012 in their value, such as If-Modified-Since and so on.
4013
4014 Example:
4015
4016 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
4017
4018 applied to:
4019
4020 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4021
4022 outputs:
4023
4024 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4025
4026 assuming the backend IP is 192.168.1.20
4027
4028 - "replace-value" works like "replace-header" except that it matches the
4029 regex against every comma-delimited value of the header field <name>
4030 instead of the entire header. This is suited for all headers which are
4031 allowed to carry more than one value. An example could be the Accept
4032 header.
4033
4034 Example:
4035
4036 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
4037
4038 applied to:
4039
4040 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
4041
4042 outputs:
4043
4044 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
4045
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004046 - "set-method" rewrites the request method with the result of the
4047 evaluation of format string <fmt>. There should be very few valid reasons
4048 for having to do so as this is more likely to break something than to fix
4049 it.
4050
4051 - "set-path" rewrites the request path with the result of the evaluation of
4052 format string <fmt>. The query string, if any, is left intact. If a
4053 scheme and authority is found before the path, they are left intact as
4054 well. If the request doesn't have a path ("*"), this one is replaced with
4055 the format. This can be used to prepend a directory component in front of
4056 a path for example. See also "set-query" and "set-uri".
4057
4058 Example :
4059 # prepend the host name before the path
4060 http-request set-path /%[hdr(host)]%[path]
4061
4062 - "set-query" rewrites the request's query string which appears after the
4063 first question mark ("?") with the result of the evaluation of format
4064 string <fmt>. The part prior to the question mark is left intact. If the
4065 request doesn't contain a question mark and the new value is not empty,
4066 then one is added at the end of the URI, followed by the new value. If
4067 a question mark was present, it will never be removed even if the value
4068 is empty. This can be used to add or remove parameters from the query
4069 string. See also "set-query" and "set-uri".
4070
4071 Example :
4072 # replace "%3D" with "=" in the query string
4073 http-request set-query %[query,regsub(%3D,=,g)]
4074
4075 - "set-uri" rewrites the request URI with the result of the evaluation of
4076 format string <fmt>. The scheme, authority, path and query string are all
4077 replaced at once. This can be used to rewrite hosts in front of proxies,
4078 or to perform complex modifications to the URI such as moving parts
4079 between the path and the query string. See also "set-path" and
4080 "set-query".
4081
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004082 - "set-nice" sets the "nice" factor of the current request being processed.
4083 It only has effect against the other requests being processed at the same
4084 time. The default value is 0, unless altered by the "nice" setting on the
4085 "bind" line. The accepted range is -1024..1024. The higher the value, the
4086 nicest the request will be. Lower values will make the request more
4087 important than other ones. This can be useful to improve the speed of
4088 some requests, or lower the priority of non-important requests. Using
4089 this setting without prior experimentation can cause some major slowdown.
4090
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004091 - "set-log-level" is used to change the log level of the current request
4092 when a certain condition is met. Valid levels are the 8 syslog levels
4093 (see the "log" keyword) plus the special level "silent" which disables
4094 logging for this request. This rule is not final so the last matching
4095 rule wins. This rule can be useful to disable health checks coming from
4096 another equipment.
4097
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004098 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
4099 the client to the value passed in <tos> on platforms which support this.
4100 This value represents the whole 8 bits of the IP TOS field, and can be
4101 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
4102 that only the 6 higher bits are used in DSCP or TOS, and the two lower
Davor Ocelice9ed2812017-12-25 17:49:28 +01004103 bits are always 0. This can be used to adjust some routing behavior on
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004104 border routers based on some information from the request. See RFC 2474,
4105 2597, 3260 and 4594 for more information.
4106
Willy Tarreau51347ed2013-06-11 19:34:13 +02004107 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
4108 client to the value passed in <mark> on platforms which support it. This
4109 value is an unsigned 32 bit value which can be matched by netfilter and
4110 by the routing table. It can be expressed both in decimal or hexadecimal
4111 format (prefixed by "0x"). This can be useful to force certain packets to
4112 take a different route (for example a cheaper network path for bulk
4113 downloads). This works on Linux kernels 2.6.32 and above and requires
4114 admin privileges.
4115
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004116 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
4117 from a file (even a dummy empty file). The file name of the ACL to be
4118 updated is passed between parentheses. It takes one argument: <key fmt>,
4119 which follows log-format rules, to collect content of the new entry. It
4120 performs a lookup in the ACL before insertion, to avoid duplicated (or
4121 more) values. This lookup is done by a linear search and can be expensive
4122 with large lists! It is the equivalent of the "add acl" command from the
4123 stats socket, but can be triggered by an HTTP request.
4124
4125 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
4126 from a file (even a dummy empty file). The file name of the ACL to be
4127 updated is passed between parentheses. It takes one argument: <key fmt>,
4128 which follows log-format rules, to collect content of the entry to delete.
4129 It is the equivalent of the "del acl" command from the stats socket, but
4130 can be triggered by an HTTP request.
4131
4132 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
4133 from a file (even a dummy empty file). The file name of the MAP to be
4134 updated is passed between parentheses. It takes one argument: <key fmt>,
4135 which follows log-format rules, to collect content of the entry to delete.
4136 It takes one argument: "file name" It is the equivalent of the "del map"
4137 command from the stats socket, but can be triggered by an HTTP request.
4138
4139 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
4140 from a file (even a dummy empty file). The file name of the MAP to be
4141 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
4142 which follows log-format rules, used to collect MAP key, and <value fmt>,
4143 which follows log-format rules, used to collect content for the new entry.
4144 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4145 more) values. This lookup is done by a linear search and can be expensive
4146 with large lists! It is the equivalent of the "set map" command from the
4147 stats socket, but can be triggered by an HTTP request.
4148
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004149 - capture <sample> [ len <length> | id <id> ] :
Willy Tarreaua9083d02015-05-08 15:27:59 +02004150 captures sample expression <sample> from the request buffer, and converts
4151 it to a string of at most <len> characters. The resulting string is
4152 stored into the next request "capture" slot, so it will possibly appear
4153 next to some captured HTTP headers. It will then automatically appear in
4154 the logs, and it will be possible to extract it using sample fetch rules
4155 to feed it into headers or anything. The length should be limited given
4156 that this size will be allocated for each capture during the whole
4157 session life. Please check section 7.3 (Fetching samples) and "capture
4158 request header" for more information.
4159
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004160 If the keyword "id" is used instead of "len", the action tries to store
4161 the captured string in a previously declared capture slot. This is useful
4162 to run captures in backends. The slot id can be declared by a previous
4163 directive "http-request capture" or with the "declare capture" keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01004164 If the slot <id> doesn't exist, then HAProxy fails parsing the
4165 configuration to prevent unexpected behavior at run time.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004166
William Lallemand86d0df02017-11-24 21:36:45 +01004167 - cache-use <name> :
4168 See section 10.2 about cache setup.
4169
Willy Tarreau09448f72014-06-25 18:12:15 +02004170 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
4171 enables tracking of sticky counters from current request. These rules
4172 do not stop evaluation and do not change default action. Three sets of
4173 counters may be simultaneously tracked by the same connection. The first
4174 "track-sc0" rule executed enables tracking of the counters of the
4175 specified table as the first set. The first "track-sc1" rule executed
4176 enables tracking of the counters of the specified table as the second
4177 set. The first "track-sc2" rule executed enables tracking of the
4178 counters of the specified table as the third set. It is a recommended
4179 practice to use the first set of counters for the per-frontend counters
4180 and the second set for the per-backend ones. But this is just a
4181 guideline, all may be used everywhere.
4182
4183 These actions take one or two arguments :
4184 <key> is mandatory, and is a sample expression rule as described
4185 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +01004186 request or connection will be analyzed, extracted, combined,
Willy Tarreau09448f72014-06-25 18:12:15 +02004187 and used to select which table entry to update the counters.
4188
4189 <table> is an optional table to be used instead of the default one,
4190 which is the stick-table declared in the current proxy. All
4191 the counters for the matches and updates for the key will
4192 then be performed in that table until the session ends.
4193
4194 Once a "track-sc*" rule is executed, the key is looked up in the table
4195 and if it is not found, an entry is allocated for it. Then a pointer to
4196 that entry is kept during all the session's life, and this entry's
4197 counters are updated as often as possible, every time the session's
4198 counters are updated, and also systematically when the session ends.
4199 Counters are only updated for events that happen after the tracking has
4200 been started. As an exception, connection counters and request counters
4201 are systematically updated so that they reflect useful information.
4202
4203 If the entry tracks concurrent connection counters, one connection is
4204 counted for as long as the entry is tracked, and the entry will not
4205 expire during that time. Tracking counters also provides a performance
4206 advantage over just checking the keys, because only one table lookup is
4207 performed for all ACL checks that make use of it.
4208
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004209 - sc-set-gpt0(<sc-id>) <int> :
4210 This action sets the GPT0 tag according to the sticky counter designated
4211 by <sc-id> and the value of <int>. The expected result is a boolean. If
4212 an error occurs, this action silently fails and the actions evaluation
4213 continues.
4214
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004215 - sc-inc-gpc0(<sc-id>):
4216 This action increments the GPC0 counter according with the sticky counter
4217 designated by <sc-id>. If an error occurs, this action silently fails and
4218 the actions evaluation continues.
4219
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004220 - sc-inc-gpc1(<sc-id>):
4221 Same as "sc-inc-gpc0" action above but for GPC1 counter.
4222
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004223 - set-var(<var-name>) <expr> :
4224 Is used to set the contents of a variable. The variable is declared
4225 inline.
4226
Daniel Schneller0b547052016-03-21 20:46:57 +01004227 <var-name> The name of the variable starts with an indication about
4228 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01004229 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01004230 "sess" : the variable is shared with the whole session
4231 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004232 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01004233 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004234 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01004235 "res" : the variable is shared only during response
4236 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004237 This prefix is followed by a name. The separator is a '.'.
Daniel Schneller0b547052016-03-21 20:46:57 +01004238 The name may only contain characters 'a-z', 'A-Z', '0-9'
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004239 and '_'.
4240
4241 <expr> Is a standard HAProxy expression formed by a sample-fetch
4242 followed by some converters.
4243
4244 Example:
4245
4246 http-request set-var(req.my_var) req.fhdr(user-agent),lower
4247
Christopher Faulet85d79c92016-11-09 16:54:56 +01004248 - unset-var(<var-name>) :
4249 Is used to unset a variable. See above for details about <var-name>.
4250
4251 Example:
4252
4253 http-request unset-var(req.my_var)
4254
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004255 - set-src <expr> :
4256 Is used to set the source IP address to the value of specified
4257 expression. Useful when a proxy in front of HAProxy rewrites source IP,
4258 but provides the correct IP in a HTTP header; or you want to mask
4259 source IP for privacy.
4260
4261 <expr> Is a standard HAProxy expression formed by a sample-fetch
4262 followed by some converters.
4263
4264 Example:
4265
4266 http-request set-src hdr(x-forwarded-for)
4267 http-request set-src src,ipmask(24)
4268
Willy Tarreau00005ce2016-10-21 15:07:45 +02004269 When possible, set-src preserves the original source port as long as the
4270 address family allows it, otherwise the source port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004271
William Lallemand44be6402016-05-25 01:51:35 +02004272 - set-src-port <expr> :
4273 Is used to set the source port address to the value of specified
4274 expression.
4275
4276 <expr> Is a standard HAProxy expression formed by a sample-fetch
4277 followed by some converters.
4278
4279 Example:
4280
4281 http-request set-src-port hdr(x-port)
4282 http-request set-src-port int(4000)
4283
Willy Tarreau00005ce2016-10-21 15:07:45 +02004284 When possible, set-src-port preserves the original source address as long
4285 as the address family supports a port, otherwise it forces the source
4286 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02004287
William Lallemand13e9b0c2016-05-25 02:34:07 +02004288 - set-dst <expr> :
4289 Is used to set the destination IP address to the value of specified
4290 expression. Useful when a proxy in front of HAProxy rewrites destination
4291 IP, but provides the correct IP in a HTTP header; or you want to mask
4292 the IP for privacy. If you want to connect to the new address/port, use
4293 '0.0.0.0:0' as a server address in the backend.
4294
4295 <expr> Is a standard HAProxy expression formed by a sample-fetch
4296 followed by some converters.
4297
4298 Example:
4299
4300 http-request set-dst hdr(x-dst)
4301 http-request set-dst dst,ipmask(24)
4302
Willy Tarreau00005ce2016-10-21 15:07:45 +02004303 When possible, set-dst preserves the original destination port as long as
4304 the address family allows it, otherwise the destination port is set to 0.
4305
William Lallemand13e9b0c2016-05-25 02:34:07 +02004306 - set-dst-port <expr> :
4307 Is used to set the destination port address to the value of specified
4308 expression. If you want to connect to the new address/port, use
4309 '0.0.0.0:0' as a server address in the backend.
4310
4311 <expr> Is a standard HAProxy expression formed by a sample-fetch
4312 followed by some converters.
4313
4314 Example:
4315
4316 http-request set-dst-port hdr(x-port)
4317 http-request set-dst-port int(4000)
4318
Willy Tarreau00005ce2016-10-21 15:07:45 +02004319 When possible, set-dst-port preserves the original destination address as
4320 long as the address family supports a port, otherwise it forces the
4321 destination address to IPv4 "0.0.0.0" before rewriting the port.
4322
Willy Tarreau2d392c22015-08-24 01:43:45 +02004323 - "silent-drop" : this stops the evaluation of the rules and makes the
Davor Ocelice9ed2812017-12-25 17:49:28 +01004324 client-facing connection suddenly disappear using a system-dependent way
Willy Tarreau2d392c22015-08-24 01:43:45 +02004325 that tries to prevent the client from being notified. The effect it then
4326 that the client still sees an established connection while there's none
4327 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
4328 except that it doesn't use any local resource at all on the machine
4329 running HAProxy. It can resist much higher loads than "tarpit", and slow
Davor Ocelice9ed2812017-12-25 17:49:28 +01004330 down stronger attackers. It is important to understand the impact of using
4331 this mechanism. All stateful equipment placed between the client and
Willy Tarreau2d392c22015-08-24 01:43:45 +02004332 HAProxy (firewalls, proxies, load balancers) will also keep the
4333 established connection for a long time and may suffer from this action.
4334 On modern Linux systems running with enough privileges, the TCP_REPAIR
4335 socket option is used to block the emission of a TCP reset. On other
4336 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
4337 pass the first router, though it's still delivered to local networks. Do
4338 not use it unless you fully understand how it works.
4339
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004340
Olivier Houchardccaa7de2017-10-02 11:51:03 +02004341 - "wait-for-handshake" : this will delay the processing of the request
4342 until the SSL handshake happened. This is mostly useful to delay
4343 processing early data until we're sure they are valid.
4344
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004345 - send-spoe-group <engine-name> <group-name> :
4346 This action is used to trigger sending of a group of SPOE messages. To do
4347 so, the SPOE engine used to send messages must be defined, as well as the
4348 SPOE group to send. Of course, the SPOE engine must refer to an existing
4349 SPOE filter. If not engine name is provided on the SPOE filter line, the
4350 SPOE agent name must be used.
4351
4352 <engine-name> The SPOE engine name.
4353
4354 <group-name> The SPOE group name as specified in the engine
4355 configuration.
4356
Willy Tarreau20b0de52012-12-24 15:45:22 +01004357 There is no limit to the number of http-request statements per instance.
4358
4359 It is important to know that http-request rules are processed very early in
4360 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004361 or "reqadd" rules. That way, headers added by "add-header"/"set-header" are
4362 visible by almost all further ACL rules.
4363
4364 Using "reqadd"/"reqdel"/"reqrep" to manipulate request headers is discouraged
4365 in newer versions (>= 1.5). But if you need to use regular expression to
4366 delete headers, you can still use "reqdel". Also please use
4367 "http-request deny/allow/tarpit" instead of "reqdeny"/"reqpass"/"reqtarpit".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004368
4369 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01004370 acl nagios src 192.168.129.3
4371 acl local_net src 192.168.0.0/16
4372 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004373
Cyril Bonté78caf842010-03-10 22:41:43 +01004374 http-request allow if nagios
4375 http-request allow if local_net auth_ok
4376 http-request auth realm Gimme if local_net auth_ok
4377 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004378
Cyril Bonté78caf842010-03-10 22:41:43 +01004379 Example:
4380 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01004381 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004382
Willy Tarreau20b0de52012-12-24 15:45:22 +01004383 Example:
4384 http-request set-header X-Haproxy-Current-Date %T
4385 http-request set-header X-SSL %[ssl_fc]
Willy Tarreaufca42612015-08-27 17:15:05 +02004386 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
Willy Tarreau20b0de52012-12-24 15:45:22 +01004387 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
4388 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
4389 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
4390 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
4391 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
4392 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
4393
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004394 Example:
4395 acl key req.hdr(X-Add-Acl-Key) -m found
4396 acl add path /addacl
4397 acl del path /delacl
4398
4399 acl myhost hdr(Host) -f myhost.lst
4400
4401 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
4402 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
4403
4404 Example:
4405 acl value req.hdr(X-Value) -m found
4406 acl setmap path /setmap
4407 acl delmap path /delmap
4408
4409 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
4410
4411 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
4412 http-request del-map(map.lst) %[src] if delmap
4413
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02004414 See also : "stats http-request", section 3.4 about userlists and section 7
4415 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01004416
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004417http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Willy Tarreau51d861a2015-05-22 17:30:48 +02004418 capture <sample> id <id> | redirect <rule> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004419 set-header <name> <fmt> | del-header <name> |
Sasha Pachev218f0642014-06-16 12:05:59 -06004420 replace-header <name> <regex-match> <replace-fmt> |
4421 replace-value <name> <regex-match> <replace-fmt> |
Robin H. Johnson52f5db22017-01-01 13:10:52 -08004422 set-status <status> [reason <str>] |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004423 set-log-level <level> | set-mark <mark> | set-tos <tos> |
4424 add-acl(<file name>) <key fmt> |
4425 del-acl(<file name>) <key fmt> |
4426 del-map(<file name>) <key fmt> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01004427 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004428 set-var(<var-name>) <expr> |
Christopher Faulet85d79c92016-11-09 16:54:56 +01004429 unset-var(<var-name>) |
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08004430 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004431 sc-inc-gpc0(<sc-id>) |
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004432 sc-inc-gpc1(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004433 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02004434 silent-drop |
William Lallemand86d0df02017-11-24 21:36:45 +01004435 send-spoe-group |
4436 cache-store
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004437 }
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02004438 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004439 Access control for Layer 7 responses
4440
4441 May be used in sections: defaults | frontend | listen | backend
4442 no | yes | yes | yes
4443
4444 The http-response statement defines a set of rules which apply to layer 7
4445 processing. The rules are evaluated in their declaration order when they are
4446 met in a frontend, listen or backend section. Any rule may optionally be
4447 followed by an ACL-based condition, in which case it will only be evaluated
4448 if the condition is true. Since these rules apply on responses, the backend
4449 rules are applied first, followed by the frontend's rules.
4450
4451 The first keyword is the rule's action. Currently supported actions include :
4452 - "allow" : this stops the evaluation of the rules and lets the response
4453 pass the check. No further "http-response" rules are evaluated for the
4454 current section.
4455
4456 - "deny" : this stops the evaluation of the rules and immediately rejects
4457 the response and emits an HTTP 502 error. No further "http-response"
4458 rules are evaluated.
4459
4460 - "add-header" appends an HTTP header field whose name is specified in
4461 <name> and whose value is defined by <fmt> which follows the log-format
4462 rules (see Custom Log Format in section 8.2.4). This may be used to send
4463 a cookie to a client for example, or to pass some internal information.
4464 This rule is not final, so it is possible to add other similar rules.
4465 Note that header addition is performed immediately, so one rule might
4466 reuse the resulting header from a previous rule.
4467
4468 - "set-header" does the same as "add-header" except that the header name
4469 is first removed if it existed. This is useful when passing security
4470 information to the server, where the header must not be manipulated by
4471 external users.
4472
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004473 - "del-header" removes all HTTP header fields whose name is specified in
4474 <name>.
4475
Sasha Pachev218f0642014-06-16 12:05:59 -06004476 - "replace-header" matches the regular expression in all occurrences of
4477 header field <name> according to <match-regex>, and replaces them with
4478 the <replace-fmt> argument. Format characters are allowed in replace-fmt
4479 and work like in <fmt> arguments in "add-header". The match is only
4480 case-sensitive. It is important to understand that this action only
4481 considers whole header lines, regardless of the number of values they
4482 may contain. This usage is suited to headers naturally containing commas
4483 in their value, such as Set-Cookie, Expires and so on.
4484
4485 Example:
4486
4487 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
4488
4489 applied to:
4490
4491 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
4492
4493 outputs:
4494
4495 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
4496
4497 assuming the backend IP is 192.168.1.20.
4498
4499 - "replace-value" works like "replace-header" except that it matches the
4500 regex against every comma-delimited value of the header field <name>
4501 instead of the entire header. This is suited for all headers which are
4502 allowed to carry more than one value. An example could be the Accept
4503 header.
4504
4505 Example:
4506
4507 http-response replace-value Cache-control ^public$ private
4508
4509 applied to:
4510
4511 Cache-Control: max-age=3600, public
4512
4513 outputs:
4514
4515 Cache-Control: max-age=3600, private
4516
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004517 - "set-status" replaces the response status code with <status> which must
Robin H. Johnson52f5db22017-01-01 13:10:52 -08004518 be an integer between 100 and 999. Optionally, a custom reason text can be
4519 provided defined by <str>, or the default reason for the specified code
4520 will be used as a fallback.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004521
4522 Example:
4523
4524 # return "431 Request Header Fields Too Large"
4525 http-response set-status 431
Robin H. Johnson52f5db22017-01-01 13:10:52 -08004526 # return "503 Slow Down", custom reason
4527 http-response set-status 503 reason "Slow Down".
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004528
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004529 - "set-nice" sets the "nice" factor of the current request being processed.
4530 It only has effect against the other requests being processed at the same
4531 time. The default value is 0, unless altered by the "nice" setting on the
4532 "bind" line. The accepted range is -1024..1024. The higher the value, the
4533 nicest the request will be. Lower values will make the request more
4534 important than other ones. This can be useful to improve the speed of
4535 some requests, or lower the priority of non-important requests. Using
4536 this setting without prior experimentation can cause some major slowdown.
4537
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004538 - "set-log-level" is used to change the log level of the current request
4539 when a certain condition is met. Valid levels are the 8 syslog levels
4540 (see the "log" keyword) plus the special level "silent" which disables
4541 logging for this request. This rule is not final so the last matching
4542 rule wins. This rule can be useful to disable health checks coming from
4543 another equipment.
4544
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004545 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
4546 the client to the value passed in <tos> on platforms which support this.
4547 This value represents the whole 8 bits of the IP TOS field, and can be
4548 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
4549 that only the 6 higher bits are used in DSCP or TOS, and the two lower
Davor Ocelice9ed2812017-12-25 17:49:28 +01004550 bits are always 0. This can be used to adjust some routing behavior on
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004551 border routers based on some information from the request. See RFC 2474,
4552 2597, 3260 and 4594 for more information.
4553
Willy Tarreau51347ed2013-06-11 19:34:13 +02004554 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
4555 client to the value passed in <mark> on platforms which support it. This
4556 value is an unsigned 32 bit value which can be matched by netfilter and
4557 by the routing table. It can be expressed both in decimal or hexadecimal
4558 format (prefixed by "0x"). This can be useful to force certain packets to
4559 take a different route (for example a cheaper network path for bulk
4560 downloads). This works on Linux kernels 2.6.32 and above and requires
4561 admin privileges.
4562
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004563 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
4564 from a file (even a dummy empty file). The file name of the ACL to be
4565 updated is passed between parentheses. It takes one argument: <key fmt>,
4566 which follows log-format rules, to collect content of the new entry. It
4567 performs a lookup in the ACL before insertion, to avoid duplicated (or
4568 more) values. This lookup is done by a linear search and can be expensive
4569 with large lists! It is the equivalent of the "add acl" command from the
4570 stats socket, but can be triggered by an HTTP response.
4571
4572 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
4573 from a file (even a dummy empty file). The file name of the ACL to be
4574 updated is passed between parentheses. It takes one argument: <key fmt>,
4575 which follows log-format rules, to collect content of the entry to delete.
4576 It is the equivalent of the "del acl" command from the stats socket, but
4577 can be triggered by an HTTP response.
4578
4579 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
4580 from a file (even a dummy empty file). The file name of the MAP to be
4581 updated is passed between parentheses. It takes one argument: <key fmt>,
4582 which follows log-format rules, to collect content of the entry to delete.
4583 It takes one argument: "file name" It is the equivalent of the "del map"
4584 command from the stats socket, but can be triggered by an HTTP response.
4585
4586 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
4587 from a file (even a dummy empty file). The file name of the MAP to be
4588 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
4589 which follows log-format rules, used to collect MAP key, and <value fmt>,
4590 which follows log-format rules, used to collect content for the new entry.
4591 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4592 more) values. This lookup is done by a linear search and can be expensive
4593 with large lists! It is the equivalent of the "set map" command from the
4594 stats socket, but can be triggered by an HTTP response.
4595
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004596 - capture <sample> id <id> :
4597 captures sample expression <sample> from the response buffer, and converts
4598 it to a string. The resulting string is stored into the next request
4599 "capture" slot, so it will possibly appear next to some captured HTTP
4600 headers. It will then automatically appear in the logs, and it will be
4601 possible to extract it using sample fetch rules to feed it into headers or
4602 anything. Please check section 7.3 (Fetching samples) and "capture
4603 response header" for more information.
4604
4605 The keyword "id" is the id of the capture slot which is used for storing
4606 the string. The capture slot must be defined in an associated frontend.
4607 This is useful to run captures in backends. The slot id can be declared by
4608 a previous directive "http-response capture" or with the "declare capture"
4609 keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01004610 If the slot <id> doesn't exist, then HAProxy fails parsing the
4611 configuration to prevent unexpected behavior at run time.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004612
William Lallemand86d0df02017-11-24 21:36:45 +01004613 - cache-store <name> :
4614 See section 10.2 about cache setup.
4615
Willy Tarreau51d861a2015-05-22 17:30:48 +02004616 - "redirect" : this performs an HTTP redirection based on a redirect rule.
4617 This supports a format string similarly to "http-request redirect" rules,
4618 with the exception that only the "location" type of redirect is possible
4619 on the response. See the "redirect" keyword for the rule's syntax. When
4620 a redirect rule is applied during a response, connections to the server
4621 are closed so that no data can be forwarded from the server to the client.
4622
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004623 - set-var(<var-name>) expr:
4624 Is used to set the contents of a variable. The variable is declared
4625 inline.
4626
Daniel Schneller0b547052016-03-21 20:46:57 +01004627 <var-name> The name of the variable starts with an indication about
4628 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01004629 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01004630 "sess" : the variable is shared with the whole session
4631 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004632 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01004633 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004634 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01004635 "res" : the variable is shared only during response
4636 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004637 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01004638 The name may only contain characters 'a-z', 'A-Z', '0-9',
4639 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004640
4641 <expr> Is a standard HAProxy expression formed by a sample-fetch
4642 followed by some converters.
4643
4644 Example:
4645
4646 http-response set-var(sess.last_redir) res.hdr(location)
4647
Christopher Faulet85d79c92016-11-09 16:54:56 +01004648 - unset-var(<var-name>) :
4649 Is used to unset a variable. See above for details about <var-name>.
4650
4651 Example:
4652
4653 http-response unset-var(sess.last_redir)
4654
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08004655 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
4656 enables tracking of sticky counters from current response. Please refer to
4657 "http-request track-sc" for a complete description. The only difference
4658 from "http-request track-sc" is the <key> sample expression can only make
Davor Ocelice9ed2812017-12-25 17:49:28 +01004659 use of samples in response (e.g. res.*, status etc.) and samples below
4660 Layer 6 (e.g. SSL-related samples, see section 7.3.4). If the sample is
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08004661 not supported, haproxy will fail and warn while parsing the config.
4662
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004663 - sc-set-gpt0(<sc-id>) <int> :
4664 This action sets the GPT0 tag according to the sticky counter designated
4665 by <sc-id> and the value of <int>. The expected result is a boolean. If
4666 an error occurs, this action silently fails and the actions evaluation
4667 continues.
4668
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004669 - sc-inc-gpc0(<sc-id>):
4670 This action increments the GPC0 counter according with the sticky counter
4671 designated by <sc-id>. If an error occurs, this action silently fails and
4672 the actions evaluation continues.
4673
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004674 - sc-inc-gpc1(<sc-id>):
4675 Same as "sc-inc-gpc0" action above but for GPC1 counter.
4676
Willy Tarreau2d392c22015-08-24 01:43:45 +02004677 - "silent-drop" : this stops the evaluation of the rules and makes the
Davor Ocelice9ed2812017-12-25 17:49:28 +01004678 client-facing connection suddenly disappear using a system-dependent way
Willy Tarreau2d392c22015-08-24 01:43:45 +02004679 that tries to prevent the client from being notified. The effect it then
4680 that the client still sees an established connection while there's none
4681 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
4682 except that it doesn't use any local resource at all on the machine
4683 running HAProxy. It can resist much higher loads than "tarpit", and slow
Davor Ocelice9ed2812017-12-25 17:49:28 +01004684 down stronger attackers. It is important to understand the impact of using
4685 this mechanism. All stateful equipment placed between the client and
Willy Tarreau2d392c22015-08-24 01:43:45 +02004686 HAProxy (firewalls, proxies, load balancers) will also keep the
4687 established connection for a long time and may suffer from this action.
4688 On modern Linux systems running with enough privileges, the TCP_REPAIR
4689 socket option is used to block the emission of a TCP reset. On other
4690 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
4691 pass the first router, though it's still delivered to local networks. Do
4692 not use it unless you fully understand how it works.
4693
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004694 - send-spoe-group <engine-name> <group-name> :
4695 This action is used to trigger sending of a group of SPOE messages. To do
4696 so, the SPOE engine used to send messages must be defined, as well as the
4697 SPOE group to send. Of course, the SPOE engine must refer to an existing
4698 SPOE filter. If not engine name is provided on the SPOE filter line, the
4699 SPOE agent name must be used.
4700
4701 <engine-name> The SPOE engine name.
4702
4703 <group-name> The SPOE group name as specified in the engine
4704 configuration.
4705
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004706 There is no limit to the number of http-response statements per instance.
4707
Godbach09250262013-07-02 01:19:15 +08004708 It is important to know that http-response rules are processed very early in
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004709 the HTTP processing, before "rspdel" or "rsprep" or "rspadd" rules. That way,
Davor Ocelice9ed2812017-12-25 17:49:28 +01004710 headers added by "add-header"/"set-header" are visible by almost all further
4711 ACL rules.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004712
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004713 Using "rspadd"/"rspdel"/"rsprep" to manipulate request headers is discouraged
4714 in newer versions (>= 1.5). But if you need to use regular expression to
4715 delete headers, you can still use "rspdel". Also please use
4716 "http-response deny" instead of "rspdeny".
4717
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004718 Example:
4719 acl key_acl res.hdr(X-Acl-Key) -m found
4720
4721 acl myhost hdr(Host) -f myhost.lst
4722
4723 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4724 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4725
4726 Example:
4727 acl value res.hdr(X-Value) -m found
4728
4729 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
4730
4731 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4732 http-response del-map(map.lst) %[src] if ! value
4733
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004734 See also : "http-request", section 3.4 about userlists and section 7 about
4735 ACL usage.
4736
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02004737
Willy Tarreau30631952015-08-06 15:05:24 +02004738http-reuse { never | safe | aggressive | always }
4739 Declare how idle HTTP connections may be shared between requests
4740
4741 May be used in sections: defaults | frontend | listen | backend
4742 yes | no | yes | yes
4743
4744 By default, a connection established between haproxy and the backend server
4745 belongs to the session that initiated it. The downside is that between the
4746 response and the next request, the connection remains idle and is not used.
4747 In many cases for performance reasons it is desirable to make it possible to
4748 reuse these idle connections to serve other requests from different sessions.
Davor Ocelice9ed2812017-12-25 17:49:28 +01004749 This directive allows to tune this behavior.
Willy Tarreau30631952015-08-06 15:05:24 +02004750
4751 The argument indicates the desired connection reuse strategy :
4752
4753 - "never" : idle connections are never shared between sessions. This is
4754 the default choice. It may be enforced to cancel a different
4755 strategy inherited from a defaults section or for
4756 troubleshooting. For example, if an old bogus application
4757 considers that multiple requests over the same connection come
4758 from the same client and it is not possible to fix the
4759 application, it may be desirable to disable connection sharing
4760 in a single backend. An example of such an application could
4761 be an old haproxy using cookie insertion in tunnel mode and
4762 not checking any request past the first one.
4763
4764 - "safe" : this is the recommended strategy. The first request of a
4765 session is always sent over its own connection, and only
4766 subsequent requests may be dispatched over other existing
4767 connections. This ensures that in case the server closes the
4768 connection when the request is being sent, the browser can
4769 decide to silently retry it. Since it is exactly equivalent to
4770 regular keep-alive, there should be no side effects.
4771
4772 - "aggressive" : this mode may be useful in webservices environments where
4773 all servers are not necessarily known and where it would be
4774 appreciable to deliver most first requests over existing
4775 connections. In this case, first requests are only delivered
4776 over existing connections that have been reused at least once,
4777 proving that the server correctly supports connection reuse.
4778 It should only be used when it's sure that the client can
4779 retry a failed request once in a while and where the benefit
4780 of aggressive connection reuse significantly outweights the
4781 downsides of rare connection failures.
4782
4783 - "always" : this mode is only recommended when the path to the server is
4784 known for never breaking existing connections quickly after
4785 releasing them. It allows the first request of a session to be
4786 sent to an existing connection. This can provide a significant
4787 performance increase over the "safe" strategy when the backend
4788 is a cache farm, since such components tend to show a
Davor Ocelice9ed2812017-12-25 17:49:28 +01004789 consistent behavior and will benefit from the connection
Willy Tarreau30631952015-08-06 15:05:24 +02004790 sharing. It is recommended that the "http-keep-alive" timeout
4791 remains low in this mode so that no dead connections remain
4792 usable. In most cases, this will lead to the same performance
4793 gains as "aggressive" but with more risks. It should only be
4794 used when it improves the situation over "aggressive".
4795
4796 When http connection sharing is enabled, a great care is taken to respect the
Davor Ocelice9ed2812017-12-25 17:49:28 +01004797 connection properties and compatibility. Specifically :
4798 - connections made with "usesrc" followed by a client-dependent value
4799 ("client", "clientip", "hdr_ip") are marked private and never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02004800
4801 - connections sent to a server with a TLS SNI extension are marked private
Davor Ocelice9ed2812017-12-25 17:49:28 +01004802 and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02004803
4804 - connections receiving a status code 401 or 407 expect some authentication
4805 to be sent in return. Due to certain bogus authentication schemes (such
4806 as NTLM) relying on the connection, these connections are marked private
Davor Ocelice9ed2812017-12-25 17:49:28 +01004807 and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02004808
4809 No connection pool is involved, once a session dies, the last idle connection
4810 it was attached to is deleted at the same time. This ensures that connections
4811 may not last after all sessions are closed.
4812
4813 Note: connection reuse improves the accuracy of the "server maxconn" setting,
4814 because almost no new connection will be established while idle connections
4815 remain available. This is particularly true with the "always" strategy.
4816
4817 See also : "option http-keep-alive", "server maxconn"
4818
4819
Mark Lamourinec2247f02012-01-04 13:02:01 -05004820http-send-name-header [<header>]
4821 Add the server name to a request. Use the header string given by <header>
4822
4823 May be used in sections: defaults | frontend | listen | backend
4824 yes | no | yes | yes
4825
4826 Arguments :
4827
4828 <header> The header string to use to send the server name
4829
4830 The "http-send-name-header" statement causes the name of the target
Davor Ocelice9ed2812017-12-25 17:49:28 +01004831 server to be added to the headers of an HTTP request. The name
Mark Lamourinec2247f02012-01-04 13:02:01 -05004832 is added with the header string proved.
4833
4834 See also : "server"
4835
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004836id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02004837 Set a persistent ID to a proxy.
4838 May be used in sections : defaults | frontend | listen | backend
4839 no | yes | yes | yes
4840 Arguments : none
4841
4842 Set a persistent ID for the proxy. This ID must be unique and positive.
4843 An unused ID will automatically be assigned if unset. The first assigned
4844 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004845
4846
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004847ignore-persist { if | unless } <condition>
4848 Declare a condition to ignore persistence
4849 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01004850 no | no | yes | yes
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004851
4852 By default, when cookie persistence is enabled, every requests containing
4853 the cookie are unconditionally persistent (assuming the target server is up
4854 and running).
4855
4856 The "ignore-persist" statement allows one to declare various ACL-based
4857 conditions which, when met, will cause a request to ignore persistence.
4858 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004859 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004860 persistence for a specific User-Agent (for example, some web crawler bots).
4861
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004862 The persistence is ignored when an "if" condition is met, or unless an
4863 "unless" condition is met.
4864
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03004865 Example:
4866 acl url_static path_beg /static /images /img /css
4867 acl url_static path_end .gif .png .jpg .css .js
4868 ignore-persist if url_static
4869
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004870 See also : "force-persist", "cookie", and section 7 about ACL usage.
4871
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004872load-server-state-from-file { global | local | none }
4873 Allow seamless reload of HAProxy
4874 May be used in sections: defaults | frontend | listen | backend
4875 yes | no | yes | yes
4876
4877 This directive points HAProxy to a file where server state from previous
4878 running process has been saved. That way, when starting up, before handling
4879 traffic, the new process can apply old states to servers exactly has if no
Davor Ocelice9ed2812017-12-25 17:49:28 +01004880 reload occurred. The purpose of the "load-server-state-from-file" directive is
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004881 to tell haproxy which file to use. For now, only 2 arguments to either prevent
4882 loading state or load states from a file containing all backends and servers.
4883 The state file can be generated by running the command "show servers state"
4884 over the stats socket and redirect output.
4885
Davor Ocelice9ed2812017-12-25 17:49:28 +01004886 The format of the file is versioned and is very specific. To understand it,
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004887 please read the documentation of the "show servers state" command (chapter
Willy Tarreau1af20c72017-06-23 16:01:14 +02004888 9.3 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004889
4890 Arguments:
4891 global load the content of the file pointed by the global directive
4892 named "server-state-file".
4893
4894 local load the content of the file pointed by the directive
4895 "server-state-file-name" if set. If not set, then the backend
4896 name is used as a file name.
4897
4898 none don't load any stat for this backend
4899
4900 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01004901 - server's IP address is preserved across reloads by default, but the
4902 order can be changed thanks to the server's "init-addr" setting. This
4903 means that an IP address change performed on the CLI at run time will
Davor Ocelice9ed2812017-12-25 17:49:28 +01004904 be preserved, and that any change to the local resolver (e.g. /etc/hosts)
Willy Tarreaue5a60682016-11-09 14:54:53 +01004905 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004906
4907 - server's weight is applied from previous running process unless it has
4908 has changed between previous and new configuration files.
4909
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004910 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004911
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004912 global
4913 stats socket /tmp/socket
4914 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004915
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004916 defaults
4917 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004918
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004919 backend bk
4920 server s1 127.0.0.1:22 check weight 11
4921 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004922
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004923
4924 Then one can run :
4925
4926 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
4927
4928 Content of the file /tmp/server_state would be like this:
4929
4930 1
4931 # <field names skipped for the doc example>
4932 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4933 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4934
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004935 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004936
4937 global
4938 stats socket /tmp/socket
4939 server-state-base /etc/haproxy/states
4940
4941 defaults
4942 load-server-state-from-file local
4943
4944 backend bk
4945 server s1 127.0.0.1:22 check weight 11
4946 server s2 127.0.0.1:22 check weight 12
4947
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004948
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004949 Then one can run :
4950
4951 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
4952
4953 Content of the file /etc/haproxy/states/bk would be like this:
4954
4955 1
4956 # <field names skipped for the doc example>
4957 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4958 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4959
4960 See also: "server-state-file", "server-state-file-name", and
4961 "show servers state"
4962
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004963
Willy Tarreau2769aa02007-12-27 18:26:09 +01004964log global
Willy Tarreau18324f52014-06-27 18:10:07 +02004965log <address> [len <length>] <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02004966no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01004967 Enable per-instance logging of events and traffic.
4968 May be used in sections : defaults | frontend | listen | backend
4969 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02004970
4971 Prefix :
4972 no should be used when the logger list must be flushed. For example,
4973 if you don't want to inherit from the default logger list. This
4974 prefix does not allow arguments.
4975
Willy Tarreau2769aa02007-12-27 18:26:09 +01004976 Arguments :
4977 global should be used when the instance's logging parameters are the
4978 same as the global ones. This is the most common usage. "global"
4979 replaces <address>, <facility> and <level> with those of the log
4980 entries found in the "global" section. Only one "log global"
4981 statement may be used per instance, and this form takes no other
4982 parameter.
4983
4984 <address> indicates where to send the logs. It takes the same format as
4985 for the "global" section's logs, and can be one of :
4986
4987 - An IPv4 address optionally followed by a colon (':') and a UDP
4988 port. If no port is specified, 514 is used by default (the
4989 standard syslog port).
4990
David du Colombier24bb5f52011-03-17 10:40:23 +01004991 - An IPv6 address followed by a colon (':') and optionally a UDP
4992 port. If no port is specified, 514 is used by default (the
4993 standard syslog port).
4994
Willy Tarreau2769aa02007-12-27 18:26:09 +01004995 - A filesystem path to a UNIX domain socket, keeping in mind
4996 considerations for chroot (be sure the path is accessible
4997 inside the chroot) and uid/gid (be sure the path is
Davor Ocelice9ed2812017-12-25 17:49:28 +01004998 appropriately writable).
Willy Tarreau2769aa02007-12-27 18:26:09 +01004999
William Lallemandb2f07452015-05-12 14:27:13 +02005000 You may want to reference some environment variables in the
5001 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01005002
Willy Tarreau18324f52014-06-27 18:10:07 +02005003 <length> is an optional maximum line length. Log lines larger than this
5004 value will be truncated before being sent. The reason is that
5005 syslog servers act differently on log line length. All servers
5006 support the default value of 1024, but some servers simply drop
5007 larger lines while others do log them. If a server supports long
5008 lines, it may make sense to set this value here in order to avoid
5009 truncating long lines. Similarly, if a server drops long lines,
5010 it is preferable to truncate them before sending them. Accepted
5011 values are 80 to 65535 inclusive. The default value of 1024 is
5012 generally fine for all standard usages. Some specific cases of
Davor Ocelice9ed2812017-12-25 17:49:28 +01005013 long captures or JSON-formatted logs may require larger values.
Willy Tarreau18324f52014-06-27 18:10:07 +02005014
Willy Tarreau2769aa02007-12-27 18:26:09 +01005015 <facility> must be one of the 24 standard syslog facilities :
5016
5017 kern user mail daemon auth syslog lpr news
5018 uucp cron auth2 ftp ntp audit alert cron2
5019 local0 local1 local2 local3 local4 local5 local6 local7
5020
5021 <level> is optional and can be specified to filter outgoing messages. By
5022 default, all messages are sent. If a level is specified, only
5023 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005024 will be sent. An optional minimum level can be specified. If it
5025 is set, logs emitted with a more severe level than this one will
5026 be capped to this level. This is used to avoid sending "emerg"
5027 messages on all terminals on some default syslog configurations.
5028 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005029
5030 emerg alert crit err warning notice info debug
5031
William Lallemand0f99e342011-10-12 17:50:54 +02005032 It is important to keep in mind that it is the frontend which decides what to
5033 log from a connection, and that in case of content switching, the log entries
5034 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005035
5036 However, backend log declaration define how and where servers status changes
5037 will be logged. Level "notice" will be used to indicate a server going up,
5038 "warning" will be used for termination signals and definitive service
5039 termination, and "alert" will be used for when a server goes down.
5040
5041 Note : According to RFC3164, messages are truncated to 1024 bytes before
5042 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005043
5044 Example :
5045 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005046 log 127.0.0.1:514 local0 notice # only send important events
5047 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02005048 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01005049
Willy Tarreau2769aa02007-12-27 18:26:09 +01005050
William Lallemand48940402012-01-30 16:47:22 +01005051log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005052 Specifies the log format string to use for traffic logs
5053 May be used in sections: defaults | frontend | listen | backend
5054 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01005055
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005056 This directive specifies the log format string that will be used for all logs
5057 resulting from traffic passing through the frontend using this line. If the
5058 directive is used in a defaults section, all subsequent frontends will use
5059 the same log format. Please see section 8.2.4 which covers the log format
5060 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01005061
Guillaume de Lafond29f45602017-03-31 19:52:15 +02005062 "log-format" directive overrides previous "option tcplog", "log-format" and
5063 "option httplog" directives.
5064
Dragan Dosen7ad31542015-09-28 17:16:47 +02005065log-format-sd <string>
5066 Specifies the RFC5424 structured-data log format string
5067 May be used in sections: defaults | frontend | listen | backend
5068 yes | yes | yes | no
5069
5070 This directive specifies the RFC5424 structured-data log format string that
5071 will be used for all logs resulting from traffic passing through the frontend
5072 using this line. If the directive is used in a defaults section, all
5073 subsequent frontends will use the same log format. Please see section 8.2.4
5074 which covers the log format string in depth.
5075
5076 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
5077 about the RFC5424 structured-data part.
5078
5079 Note : This log format string will be used only for loggers that have set
5080 log format to "rfc5424".
5081
5082 Example :
5083 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
5084
5085
Willy Tarreau094af4e2015-01-07 15:03:42 +01005086log-tag <string>
5087 Specifies the log tag to use for all outgoing logs
5088 May be used in sections: defaults | frontend | listen | backend
5089 yes | yes | yes | yes
5090
5091 Sets the tag field in the syslog header to this string. It defaults to the
5092 log-tag set in the global section, otherwise the program name as launched
5093 from the command line, which usually is "haproxy". Sometimes it can be useful
5094 to differentiate between multiple processes running on the same host, or to
5095 differentiate customer instances running in the same process. In the backend,
5096 logs about servers up/down will use this tag. As a hint, it can be convenient
5097 to set a log-tag related to a hosted customer in a defaults section then put
5098 all the frontends and backends for that customer, then start another customer
5099 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005100
Willy Tarreauc35362a2014-04-25 13:58:37 +02005101max-keep-alive-queue <value>
5102 Set the maximum server queue size for maintaining keep-alive connections
5103 May be used in sections: defaults | frontend | listen | backend
5104 yes | no | yes | yes
5105
5106 HTTP keep-alive tries to reuse the same server connection whenever possible,
5107 but sometimes it can be counter-productive, for example if a server has a lot
5108 of connections while other ones are idle. This is especially true for static
5109 servers.
5110
5111 The purpose of this setting is to set a threshold on the number of queued
5112 connections at which haproxy stops trying to reuse the same server and prefers
5113 to find another one. The default value, -1, means there is no limit. A value
5114 of zero means that keep-alive requests will never be queued. For very close
5115 servers which can be reached with a low latency and which are not sensible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01005116 breaking keep-alive, a low value is recommended (e.g. local static server can
Willy Tarreauc35362a2014-04-25 13:58:37 +02005117 use a value of 10 or less). For remote servers suffering from a high latency,
5118 higher values might be needed to cover for the latency and/or the cost of
5119 picking a different server.
5120
5121 Note that this has no impact on responses which are maintained to the same
5122 server consecutively to a 401 response. They will still go to the same server
5123 even if they have to be queued.
5124
5125 See also : "option http-server-close", "option prefer-last-server", server
5126 "maxconn" and cookie persistence.
5127
5128
Willy Tarreau2769aa02007-12-27 18:26:09 +01005129maxconn <conns>
5130 Fix the maximum number of concurrent connections on a frontend
5131 May be used in sections : defaults | frontend | listen | backend
5132 yes | yes | yes | no
5133 Arguments :
5134 <conns> is the maximum number of concurrent connections the frontend will
5135 accept to serve. Excess connections will be queued by the system
5136 in the socket's listen queue and will be served once a connection
5137 closes.
5138
5139 If the system supports it, it can be useful on big sites to raise this limit
5140 very high so that haproxy manages connection queues, instead of leaving the
5141 clients with unanswered connection attempts. This value should not exceed the
5142 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01005143 of tune.bufsize (16kB by default) each, as well as some other data resulting
5144 in about 33 kB of RAM being consumed per established connection. That means
5145 that a medium system equipped with 1GB of RAM can withstand around
5146 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005147
5148 Also, when <conns> is set to large values, it is possible that the servers
5149 are not sized to accept such loads, and for this reason it is generally wise
5150 to assign them some reasonable connection limits.
5151
Vincent Bernat6341be52012-06-27 17:18:30 +02005152 By default, this value is set to 2000.
5153
Willy Tarreau2769aa02007-12-27 18:26:09 +01005154 See also : "server", global section's "maxconn", "fullconn"
5155
5156
5157mode { tcp|http|health }
5158 Set the running mode or protocol of the instance
5159 May be used in sections : defaults | frontend | listen | backend
5160 yes | yes | yes | yes
5161 Arguments :
5162 tcp The instance will work in pure TCP mode. A full-duplex connection
5163 will be established between clients and servers, and no layer 7
5164 examination will be performed. This is the default mode. It
5165 should be used for SSL, SSH, SMTP, ...
5166
5167 http The instance will work in HTTP mode. The client request will be
5168 analyzed in depth before connecting to any server. Any request
5169 which is not RFC-compliant will be rejected. Layer 7 filtering,
5170 processing and switching will be possible. This is the mode which
5171 brings HAProxy most of its value.
5172
5173 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02005174 to incoming connections and close the connection. Alternatively,
5175 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
5176 instead. Nothing will be logged in either case. This mode is used
5177 to reply to external components health checks. This mode is
5178 deprecated and should not be used anymore as it is possible to do
5179 the same and even better by combining TCP or HTTP modes with the
5180 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005181
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005182 When doing content switching, it is mandatory that the frontend and the
5183 backend are in the same mode (generally HTTP), otherwise the configuration
5184 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005185
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005186 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005187 defaults http_instances
5188 mode http
5189
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005190 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005191
Willy Tarreau0ba27502007-12-24 16:55:16 +01005192
Cyril Bontéf0c60612010-02-06 14:44:47 +01005193monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01005194 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005195 May be used in sections : defaults | frontend | listen | backend
5196 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01005197 Arguments :
5198 if <cond> the monitor request will fail if the condition is satisfied,
5199 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005200 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01005201 are met, for instance a low number of servers both in a
5202 backend and its backup.
5203
5204 unless <cond> the monitor request will succeed only if the condition is
5205 satisfied, and will fail otherwise. Such a condition may be
5206 based on a test on the presence of a minimum number of active
5207 servers in a list of backends.
5208
5209 This statement adds a condition which can force the response to a monitor
5210 request to report a failure. By default, when an external component queries
5211 the URI dedicated to monitoring, a 200 response is returned. When one of the
5212 conditions above is met, haproxy will return 503 instead of 200. This is
5213 very useful to report a site failure to an external component which may base
5214 routing advertisements between multiple sites on the availability reported by
5215 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005216 criterion. Note that "monitor fail" only works in HTTP mode. Both status
5217 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005218
5219 Example:
5220 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01005221 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01005222 acl site_dead nbsrv(dynamic) lt 2
5223 acl site_dead nbsrv(static) lt 2
5224 monitor-uri /site_alive
5225 monitor fail if site_dead
5226
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005227 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005228
5229
5230monitor-net <source>
5231 Declare a source network which is limited to monitor requests
5232 May be used in sections : defaults | frontend | listen | backend
5233 yes | yes | yes | no
5234 Arguments :
5235 <source> is the source IPv4 address or network which will only be able to
5236 get monitor responses to any request. It can be either an IPv4
5237 address, a host name, or an address followed by a slash ('/')
5238 followed by a mask.
5239
5240 In TCP mode, any connection coming from a source matching <source> will cause
5241 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005242 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01005243 forwarding the connection to a remote server.
5244
5245 In HTTP mode, a connection coming from a source matching <source> will be
5246 accepted, the following response will be sent without waiting for a request,
5247 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
5248 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02005249 running without forwarding the request to a backend server. Note that this
5250 response is sent in raw format, without any transformation. This is important
5251 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005252
Willy Tarreau82569f92012-09-27 23:48:56 +02005253 Monitor requests are processed very early, just after tcp-request connection
5254 ACLs which are the only ones able to block them. These connections are short
5255 lived and never wait for any data from the client. They cannot be logged, and
5256 it is the intended purpose. They are only used to report HAProxy's health to
5257 an upper component, nothing more. Please note that "monitor fail" rules do
5258 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01005259
Willy Tarreau95cd2832010-03-04 23:36:33 +01005260 Last, please note that only one "monitor-net" statement can be specified in
5261 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005262
Willy Tarreau2769aa02007-12-27 18:26:09 +01005263 Example :
5264 # addresses .252 and .253 are just probing us.
5265 frontend www
5266 monitor-net 192.168.0.252/31
5267
5268 See also : "monitor fail", "monitor-uri"
5269
5270
5271monitor-uri <uri>
5272 Intercept a URI used by external components' monitor requests
5273 May be used in sections : defaults | frontend | listen | backend
5274 yes | yes | yes | no
5275 Arguments :
5276 <uri> is the exact URI which we want to intercept to return HAProxy's
5277 health status instead of forwarding the request.
5278
5279 When an HTTP request referencing <uri> will be received on a frontend,
5280 HAProxy will not forward it nor log it, but instead will return either
5281 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
5282 conditions defined with "monitor fail". This is normally enough for any
5283 front-end HTTP probe to detect that the service is UP and running without
5284 forwarding the request to a backend server. Note that the HTTP method, the
5285 version and all headers are ignored, but the request must at least be valid
5286 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
5287
Willy Tarreau721d8e02017-12-01 18:25:08 +01005288 Monitor requests are processed very early, just after the request is parsed
5289 and even before any "http-request" or "block" rulesets. The only rulesets
5290 applied before are the tcp-request ones. They cannot be logged either, and it
5291 is the intended purpose. They are only used to report HAProxy's health to an
5292 upper component, nothing more. However, it is possible to add any number of
5293 conditions using "monitor fail" and ACLs so that the result can be adjusted
5294 to whatever check can be imagined (most often the number of available servers
5295 in a backend).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005296
5297 Example :
5298 # Use /haproxy_test to report haproxy's status
5299 frontend www
5300 mode http
5301 monitor-uri /haproxy_test
5302
5303 See also : "monitor fail", "monitor-net"
5304
Willy Tarreau0ba27502007-12-24 16:55:16 +01005305
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005306option abortonclose
5307no option abortonclose
5308 Enable or disable early dropping of aborted requests pending in queues.
5309 May be used in sections : defaults | frontend | listen | backend
5310 yes | no | yes | yes
5311 Arguments : none
5312
5313 In presence of very high loads, the servers will take some time to respond.
5314 The per-instance connection queue will inflate, and the response time will
5315 increase respective to the size of the queue times the average per-session
5316 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01005317 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005318 the queue, and slowing down other users, and the servers as well, because the
5319 request will eventually be served, then aborted at the first error
5320 encountered while delivering the response.
5321
5322 As there is no way to distinguish between a full STOP and a simple output
5323 close on the client side, HTTP agents should be conservative and consider
5324 that the client might only have closed its output channel while waiting for
5325 the response. However, this introduces risks of congestion when lots of users
5326 do the same, and is completely useless nowadays because probably no client at
5327 all will close the session while waiting for the response. Some HTTP agents
Davor Ocelice9ed2812017-12-25 17:49:28 +01005328 support this behavior (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005329 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01005330 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005331 of being the single component to break rare but valid traffic is extremely
5332 low, which adds to the temptation to be able to abort a session early while
5333 still not served and not pollute the servers.
5334
Davor Ocelice9ed2812017-12-25 17:49:28 +01005335 In HAProxy, the user can choose the desired behavior using the option
5336 "abortonclose". By default (without the option) the behavior is HTTP
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005337 compliant and aborted requests will be served. But when the option is
5338 specified, a session with an incoming channel closed will be aborted while
5339 it is still possible, either pending in the queue for a connection slot, or
5340 during the connection establishment if the server has not yet acknowledged
5341 the connection request. This considerably reduces the queue size and the load
5342 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01005343 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005344
5345 If this option has been enabled in a "defaults" section, it can be disabled
5346 in a specific instance by prepending the "no" keyword before it.
5347
5348 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
5349
5350
Willy Tarreau4076a152009-04-02 15:18:36 +02005351option accept-invalid-http-request
5352no option accept-invalid-http-request
5353 Enable or disable relaxing of HTTP request parsing
5354 May be used in sections : defaults | frontend | listen | backend
5355 yes | yes | yes | no
5356 Arguments : none
5357
Willy Tarreau91852eb2015-05-01 13:26:00 +02005358 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005359 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005360 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005361 forbidden characters are essentially used to build attacks exploiting server
5362 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5363 server will emit invalid header names for whatever reason (configuration,
5364 implementation) and the issue will not be immediately fixed. In such a case,
5365 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01005366 even if that does not make sense, by specifying this option. Similarly, the
5367 list of characters allowed to appear in a URI is well defined by RFC3986, and
5368 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
5369 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
Davor Ocelice9ed2812017-12-25 17:49:28 +01005370 not allowed at all. HAProxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02005371 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02005372 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
5373 to pass through (no version specified) and multiple digits for both the major
5374 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005375
5376 This option should never be enabled by default as it hides application bugs
5377 and open security breaches. It should only be deployed after a problem has
5378 been confirmed.
5379
5380 When this option is enabled, erroneous header names will still be accepted in
5381 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01005382 analysis using the "show errors" request on the UNIX stats socket. Similarly,
5383 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02005384 also helps confirming that the issue has been solved.
5385
5386 If this option has been enabled in a "defaults" section, it can be disabled
5387 in a specific instance by prepending the "no" keyword before it.
5388
5389 See also : "option accept-invalid-http-response" and "show errors" on the
5390 stats socket.
5391
5392
5393option accept-invalid-http-response
5394no option accept-invalid-http-response
5395 Enable or disable relaxing of HTTP response parsing
5396 May be used in sections : defaults | frontend | listen | backend
5397 yes | no | yes | yes
5398 Arguments : none
5399
Willy Tarreau91852eb2015-05-01 13:26:00 +02005400 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005401 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005402 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005403 forbidden characters are essentially used to build attacks exploiting server
5404 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5405 server will emit invalid header names for whatever reason (configuration,
5406 implementation) and the issue will not be immediately fixed. In such a case,
5407 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02005408 even if that does not make sense, by specifying this option. This option also
5409 relaxes the test on the HTTP version format, it allows multiple digits for
5410 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005411
5412 This option should never be enabled by default as it hides application bugs
5413 and open security breaches. It should only be deployed after a problem has
5414 been confirmed.
5415
5416 When this option is enabled, erroneous header names will still be accepted in
5417 responses, but the complete response will be captured in order to permit
5418 later analysis using the "show errors" request on the UNIX stats socket.
5419 Doing this also helps confirming that the issue has been solved.
5420
5421 If this option has been enabled in a "defaults" section, it can be disabled
5422 in a specific instance by prepending the "no" keyword before it.
5423
5424 See also : "option accept-invalid-http-request" and "show errors" on the
5425 stats socket.
5426
5427
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005428option allbackups
5429no option allbackups
5430 Use either all backup servers at a time or only the first one
5431 May be used in sections : defaults | frontend | listen | backend
5432 yes | no | yes | yes
5433 Arguments : none
5434
5435 By default, the first operational backup server gets all traffic when normal
5436 servers are all down. Sometimes, it may be preferred to use multiple backups
5437 at once, because one will not be enough. When "option allbackups" is enabled,
5438 the load balancing will be performed among all backup servers when all normal
5439 ones are unavailable. The same load balancing algorithm will be used and the
5440 servers' weights will be respected. Thus, there will not be any priority
5441 order between the backup servers anymore.
5442
5443 This option is mostly used with static server farms dedicated to return a
5444 "sorry" page when an application is completely offline.
5445
5446 If this option has been enabled in a "defaults" section, it can be disabled
5447 in a specific instance by prepending the "no" keyword before it.
5448
5449
5450option checkcache
5451no option checkcache
Godbach7056a352013-12-11 20:01:07 +08005452 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005453 May be used in sections : defaults | frontend | listen | backend
5454 yes | no | yes | yes
5455 Arguments : none
5456
5457 Some high-level frameworks set application cookies everywhere and do not
5458 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005459 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005460 high risk of session crossing or stealing between users traversing the same
5461 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005462 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005463
5464 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005465 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01005466 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005467 response to check if there's a risk of caching a cookie on a client-side
5468 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01005469 to the client are :
Davor Ocelice9ed2812017-12-25 17:49:28 +01005470 - all those without "Set-Cookie" header;
Willy Tarreauc55ddce2017-12-21 11:41:38 +01005471 - all those with a return code other than 200, 203, 204, 206, 300, 301,
5472 404, 405, 410, 414, 501, provided that the server has not set a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005473 "Cache-control: public" header field;
Willy Tarreau24ea0bc2017-12-21 11:32:55 +01005474 - all those that result from a request using a method other than GET, HEAD,
5475 OPTIONS, TRACE, provided that the server has not set a 'Cache-Control:
Davor Ocelice9ed2812017-12-25 17:49:28 +01005476 public' header field;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005477 - those with a 'Pragma: no-cache' header
5478 - those with a 'Cache-control: private' header
5479 - those with a 'Cache-control: no-store' header
5480 - those with a 'Cache-control: max-age=0' header
5481 - those with a 'Cache-control: s-maxage=0' header
5482 - those with a 'Cache-control: no-cache' header
5483 - those with a 'Cache-control: no-cache="set-cookie"' header
5484 - those with a 'Cache-control: no-cache="set-cookie,' header
5485 (allowing other fields after set-cookie)
5486
5487 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01005488 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005489 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005490 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005491 that admins are informed that there's something to be fixed.
5492
5493 Due to the high impact on the application, the application should be tested
5494 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005495 good practice to always activate it during tests, even if it is not used in
Davor Ocelice9ed2812017-12-25 17:49:28 +01005496 production, as it will report potentially dangerous application behaviors.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005497
5498 If this option has been enabled in a "defaults" section, it can be disabled
5499 in a specific instance by prepending the "no" keyword before it.
5500
5501
5502option clitcpka
5503no option clitcpka
5504 Enable or disable the sending of TCP keepalive packets on the client side
5505 May be used in sections : defaults | frontend | listen | backend
5506 yes | yes | yes | no
5507 Arguments : none
5508
5509 When there is a firewall or any session-aware component between a client and
5510 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01005511 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005512 components decides to expire a session which has remained idle for too long.
5513
5514 Enabling socket-level TCP keep-alives makes the system regularly send packets
5515 to the other end of the connection, leaving it active. The delay between
5516 keep-alive probes is controlled by the system only and depends both on the
5517 operating system and its tuning parameters.
5518
5519 It is important to understand that keep-alive packets are neither emitted nor
5520 received at the application level. It is only the network stacks which sees
5521 them. For this reason, even if one side of the proxy already uses keep-alives
5522 to maintain its connection alive, those keep-alive packets will not be
5523 forwarded to the other side of the proxy.
5524
5525 Please note that this has nothing to do with HTTP keep-alive.
5526
5527 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
5528 client side of a connection, which should help when session expirations are
5529 noticed between HAProxy and a client.
5530
5531 If this option has been enabled in a "defaults" section, it can be disabled
5532 in a specific instance by prepending the "no" keyword before it.
5533
5534 See also : "option srvtcpka", "option tcpka"
5535
5536
Willy Tarreau0ba27502007-12-24 16:55:16 +01005537option contstats
5538 Enable continuous traffic statistics updates
5539 May be used in sections : defaults | frontend | listen | backend
5540 yes | yes | yes | no
5541 Arguments : none
5542
5543 By default, counters used for statistics calculation are incremented
5544 only when a session finishes. It works quite well when serving small
5545 objects, but with big ones (for example large images or archives) or
5546 with A/V streaming, a graph generated from haproxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01005547 a hedgehog. With this option enabled counters get incremented frequently
5548 along the session, typically every 5 seconds, which is often enough to
5549 produce clean graphs. Recounting touches a hotpath directly so it is not
5550 not enabled by default, as it can cause a lot of wakeups for very large
5551 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005552
5553
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005554option dontlog-normal
5555no option dontlog-normal
5556 Enable or disable logging of normal, successful connections
5557 May be used in sections : defaults | frontend | listen | backend
5558 yes | yes | yes | no
5559 Arguments : none
5560
5561 There are large sites dealing with several thousand connections per second
5562 and for which logging is a major pain. Some of them are even forced to turn
5563 logs off and cannot debug production issues. Setting this option ensures that
5564 normal connections, those which experience no error, no timeout, no retry nor
5565 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
5566 mode, the response status code is checked and return codes 5xx will still be
5567 logged.
5568
5569 It is strongly discouraged to use this option as most of the time, the key to
5570 complex issues is in the normal logs which will not be logged here. If you
5571 need to separate logs, see the "log-separate-errors" option instead.
5572
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005573 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005574 logging.
5575
5576
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005577option dontlognull
5578no option dontlognull
5579 Enable or disable logging of null connections
5580 May be used in sections : defaults | frontend | listen | backend
5581 yes | yes | yes | no
5582 Arguments : none
5583
5584 In certain environments, there are components which will regularly connect to
5585 various systems to ensure that they are still alive. It can be the case from
5586 another load balancer as well as from monitoring systems. By default, even a
5587 simple port probe or scan will produce a log. If those connections pollute
5588 the logs too much, it is possible to enable option "dontlognull" to indicate
5589 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02005590 which typically corresponds to those probes. Note that errors will still be
5591 returned to the client and accounted for in the stats. If this is not what is
5592 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005593
5594 It is generally recommended not to use this option in uncontrolled
Davor Ocelice9ed2812017-12-25 17:49:28 +01005595 environments (e.g. internet), otherwise scans and other malicious activities
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005596 would not be logged.
5597
5598 If this option has been enabled in a "defaults" section, it can be disabled
5599 in a specific instance by prepending the "no" keyword before it.
5600
Willy Tarreau0f228a02015-05-01 15:37:53 +02005601 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
5602 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005603
5604
5605option forceclose
5606no option forceclose
5607 Enable or disable active connection closing after response is transferred.
5608 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01005609 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005610 Arguments : none
5611
5612 Some HTTP servers do not necessarily close the connections when they receive
5613 the "Connection: close" set by "option httpclose", and if the client does not
5614 close either, then the connection remains open till the timeout expires. This
5615 causes high number of simultaneous connections on the servers and shows high
5616 global session times in the logs.
5617
5618 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01005619 actively close the outgoing server channel as soon as the server has finished
Cyril Bonté653dcd62014-02-20 00:13:15 +01005620 to respond and release some resources earlier than with "option httpclose".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005621
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005622 This option may also be combined with "option http-pretend-keepalive", which
5623 will disable sending of the "Connection: close" header, but will still cause
5624 the connection to be closed once the whole response is received.
5625
Cyril Bonté653dcd62014-02-20 00:13:15 +01005626 This option disables and replaces any previous "option httpclose", "option
5627 http-server-close", "option http-keep-alive", or "option http-tunnel".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005628
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005629 If this option has been enabled in a "defaults" section, it can be disabled
5630 in a specific instance by prepending the "no" keyword before it.
5631
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005632 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005633
5634
Willy Tarreau87cf5142011-08-19 22:57:24 +02005635option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01005636 Enable insertion of the X-Forwarded-For header to requests sent to servers
5637 May be used in sections : defaults | frontend | listen | backend
5638 yes | yes | yes | yes
5639 Arguments :
5640 <network> is an optional argument used to disable this option for sources
5641 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02005642 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01005643 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005644
5645 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
5646 their client address. This is sometimes annoying when the client's IP address
5647 is expected in server logs. To solve this problem, the well-known HTTP header
5648 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
5649 This header contains a value representing the client's IP address. Since this
5650 header is always appended at the end of the existing header list, the server
5651 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02005652 the server's manual to find how to enable use of this standard header. Note
5653 that only the last occurrence of the header must be used, since it is really
5654 possible that the client has already brought one.
5655
Willy Tarreaud72758d2010-01-12 10:42:19 +01005656 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02005657 the default "X-Forwarded-For". This can be useful where you might already
Davor Ocelice9ed2812017-12-25 17:49:28 +01005658 have a "X-Forwarded-For" header from a different application (e.g. stunnel),
Willy Tarreaud72758d2010-01-12 10:42:19 +01005659 and you need preserve it. Also if your backend server doesn't use the
Davor Ocelice9ed2812017-12-25 17:49:28 +01005660 "X-Forwarded-For" header and requires different one (e.g. Zeus Web Servers
Ross Westaf72a1d2008-08-03 10:51:45 +02005661 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01005662
5663 Sometimes, a same HAProxy instance may be shared between a direct client
5664 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5665 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5666 header for a known source address or network by adding the "except" keyword
5667 followed by the network address. In this case, any source IP matching the
5668 network will not cause an addition of this header. Most common uses are with
5669 private networks or 127.0.0.1.
5670
Willy Tarreau87cf5142011-08-19 22:57:24 +02005671 Alternatively, the keyword "if-none" states that the header will only be
5672 added if it is not present. This should only be used in perfectly trusted
5673 environment, as this might cause a security issue if headers reaching haproxy
5674 are under the control of the end-user.
5675
Willy Tarreauc27debf2008-01-06 08:57:02 +01005676 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02005677 least one of them uses it, the header will be added. Note that the backend's
5678 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02005679 both are defined. In the case of the "if-none" argument, if at least one of
5680 the frontend or the backend does not specify it, it wants the addition to be
5681 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005682
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005683 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01005684 # Public HTTP address also used by stunnel on the same machine
5685 frontend www
5686 mode http
5687 option forwardfor except 127.0.0.1 # stunnel already adds the header
5688
Ross Westaf72a1d2008-08-03 10:51:45 +02005689 # Those servers want the IP Address in X-Client
5690 backend www
5691 mode http
5692 option forwardfor header X-Client
5693
Willy Tarreau87cf5142011-08-19 22:57:24 +02005694 See also : "option httpclose", "option http-server-close",
Willy Tarreau70dffda2014-01-30 03:07:23 +01005695 "option forceclose", "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01005696
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005697
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005698option http-buffer-request
5699no option http-buffer-request
5700 Enable or disable waiting for whole HTTP request body before proceeding
5701 May be used in sections : defaults | frontend | listen | backend
5702 yes | yes | yes | yes
5703 Arguments : none
5704
5705 It is sometimes desirable to wait for the body of an HTTP request before
5706 taking a decision. This is what is being done by "balance url_param" for
5707 example. The first use case is to buffer requests from slow clients before
5708 connecting to the server. Another use case consists in taking the routing
5709 decision based on the request body's contents. This option placed in a
5710 frontend or backend forces the HTTP processing to wait until either the whole
5711 body is received, or the request buffer is full, or the first chunk is
5712 complete in case of chunked encoding. It can have undesired side effects with
Tim Düsterhus4896c442016-11-29 02:15:19 +01005713 some applications abusing HTTP by expecting unbuffered transmissions between
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005714 the frontend and the backend, so this should definitely not be used by
5715 default.
5716
Baptiste Assmanneccdf432015-10-28 13:49:01 +01005717 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005718
5719
Willy Tarreau0f228a02015-05-01 15:37:53 +02005720option http-ignore-probes
5721no option http-ignore-probes
5722 Enable or disable logging of null connections and request timeouts
5723 May be used in sections : defaults | frontend | listen | backend
5724 yes | yes | yes | no
5725 Arguments : none
5726
5727 Recently some browsers started to implement a "pre-connect" feature
5728 consisting in speculatively connecting to some recently visited web sites
5729 just in case the user would like to visit them. This results in many
5730 connections being established to web sites, which end up in 408 Request
5731 Timeout if the timeout strikes first, or 400 Bad Request when the browser
5732 decides to close them first. These ones pollute the log and feed the error
5733 counters. There was already "option dontlognull" but it's insufficient in
5734 this case. Instead, this option does the following things :
5735 - prevent any 400/408 message from being sent to the client if nothing
Davor Ocelice9ed2812017-12-25 17:49:28 +01005736 was received over a connection before it was closed;
5737 - prevent any log from being emitted in this situation;
Willy Tarreau0f228a02015-05-01 15:37:53 +02005738 - prevent any error counter from being incremented
5739
5740 That way the empty connection is silently ignored. Note that it is better
5741 not to use this unless it is clear that it is needed, because it will hide
5742 real problems. The most common reason for not receiving a request and seeing
5743 a 408 is due to an MTU inconsistency between the client and an intermediary
5744 element such as a VPN, which blocks too large packets. These issues are
5745 generally seen with POST requests as well as GET with large cookies. The logs
5746 are often the only way to detect them.
5747
5748 If this option has been enabled in a "defaults" section, it can be disabled
5749 in a specific instance by prepending the "no" keyword before it.
5750
5751 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
5752
5753
Willy Tarreau16bfb022010-01-16 19:48:41 +01005754option http-keep-alive
5755no option http-keep-alive
5756 Enable or disable HTTP keep-alive from client to server
5757 May be used in sections : defaults | frontend | listen | backend
5758 yes | yes | yes | yes
5759 Arguments : none
5760
Willy Tarreau70dffda2014-01-30 03:07:23 +01005761 By default HAProxy operates in keep-alive mode with regards to persistent
5762 connections: for each connection it processes each request and response, and
5763 leaves the connection idle on both sides between the end of a response and the
5764 start of a new request. This mode may be changed by several options such as
5765 "option http-server-close", "option forceclose", "option httpclose" or
5766 "option http-tunnel". This option allows to set back the keep-alive mode,
5767 which can be useful when another mode was used in a defaults section.
5768
5769 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
5770 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01005771 network) and the fastest session reuse on the server side at the expense
5772 of maintaining idle connections to the servers. In general, it is possible
5773 with this option to achieve approximately twice the request rate that the
5774 "http-server-close" option achieves on small objects. There are mainly two
5775 situations where this option may be useful :
5776
5777 - when the server is non-HTTP compliant and authenticates the connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01005778 instead of requests (e.g. NTLM authentication)
Willy Tarreau16bfb022010-01-16 19:48:41 +01005779
5780 - when the cost of establishing the connection to the server is significant
5781 compared to the cost of retrieving the associated object from the server.
5782
5783 This last case can happen when the server is a fast static server of cache.
5784 In this case, the server will need to be properly tuned to support high enough
5785 connection counts because connections will last until the client sends another
5786 request.
5787
5788 If the client request has to go to another backend or another server due to
5789 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01005790 immediately be closed and a new one re-opened. Option "prefer-last-server" is
5791 available to try optimize server selection so that if the server currently
5792 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01005793
Willy Tarreau16bfb022010-01-16 19:48:41 +01005794 At the moment, logs will not indicate whether requests came from the same
5795 session or not. The accept date reported in the logs corresponds to the end
5796 of the previous request, and the request time corresponds to the time spent
5797 waiting for a new request. The keep-alive request time is still bound to the
5798 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5799 not set.
5800
Cyril Bonté653dcd62014-02-20 00:13:15 +01005801 This option disables and replaces any previous "option httpclose", "option
5802 http-server-close", "option forceclose" or "option http-tunnel". When backend
Willy Tarreau70dffda2014-01-30 03:07:23 +01005803 and frontend options differ, all of these 4 options have precedence over
Cyril Bonté653dcd62014-02-20 00:13:15 +01005804 "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005805
5806 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01005807 "option prefer-last-server", "option http-pretend-keepalive",
5808 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005809
5810
Willy Tarreau96e31212011-05-30 18:10:30 +02005811option http-no-delay
5812no option http-no-delay
5813 Instruct the system to favor low interactive delays over performance in HTTP
5814 May be used in sections : defaults | frontend | listen | backend
5815 yes | yes | yes | yes
5816 Arguments : none
5817
5818 In HTTP, each payload is unidirectional and has no notion of interactivity.
5819 Any agent is expected to queue data somewhat for a reasonably low delay.
5820 There are some very rare server-to-server applications that abuse the HTTP
5821 protocol and expect the payload phase to be highly interactive, with many
5822 interleaved data chunks in both directions within a single request. This is
5823 absolutely not supported by the HTTP specification and will not work across
5824 most proxies or servers. When such applications attempt to do this through
5825 haproxy, it works but they will experience high delays due to the network
5826 optimizations which favor performance by instructing the system to wait for
5827 enough data to be available in order to only send full packets. Typical
5828 delays are around 200 ms per round trip. Note that this only happens with
5829 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
5830 affected.
5831
5832 When "option http-no-delay" is present in either the frontend or the backend
5833 used by a connection, all such optimizations will be disabled in order to
5834 make the exchanges as fast as possible. Of course this offers no guarantee on
5835 the functionality, as it may break at any other place. But if it works via
5836 HAProxy, it will work as fast as possible. This option should never be used
5837 by default, and should never be used at all unless such a buggy application
5838 is discovered. The impact of using this option is an increase of bandwidth
5839 usage and CPU usage, which may significantly lower performance in high
5840 latency environments.
5841
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005842 See also : "option http-buffer-request"
5843
Willy Tarreau96e31212011-05-30 18:10:30 +02005844
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005845option http-pretend-keepalive
5846no option http-pretend-keepalive
5847 Define whether haproxy will announce keepalive to the server or not
5848 May be used in sections : defaults | frontend | listen | backend
5849 yes | yes | yes | yes
5850 Arguments : none
5851
5852 When running with "option http-server-close" or "option forceclose", haproxy
5853 adds a "Connection: close" header to the request forwarded to the server.
5854 Unfortunately, when some servers see this header, they automatically refrain
5855 from using the chunked encoding for responses of unknown length, while this
5856 is totally unrelated. The immediate effect is that this prevents haproxy from
5857 maintaining the client connection alive. A second effect is that a client or
5858 a cache could receive an incomplete response without being aware of it, and
5859 consider the response complete.
5860
5861 By setting "option http-pretend-keepalive", haproxy will make the server
5862 believe it will keep the connection alive. The server will then not fall back
5863 to the abnormal undesired above. When haproxy gets the whole response, it
5864 will close the connection with the server just as it would do with the
5865 "forceclose" option. That way the client gets a normal response and the
5866 connection is correctly closed on the server side.
5867
5868 It is recommended not to enable this option by default, because most servers
5869 will more efficiently close the connection themselves after the last packet,
5870 and release its buffers slightly earlier. Also, the added packet on the
5871 network could slightly reduce the overall peak performance. However it is
5872 worth noting that when this option is enabled, haproxy will have slightly
5873 less work to do. So if haproxy is the bottleneck on the whole architecture,
5874 enabling this option might save a few CPU cycles.
5875
5876 This option may be set both in a frontend and in a backend. It is enabled if
5877 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005878 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02005879 keepalive to be announced to the server and close to be announced to the
5880 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005881
5882 If this option has been enabled in a "defaults" section, it can be disabled
5883 in a specific instance by prepending the "no" keyword before it.
5884
Willy Tarreau16bfb022010-01-16 19:48:41 +01005885 See also : "option forceclose", "option http-server-close", and
5886 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005887
Willy Tarreauc27debf2008-01-06 08:57:02 +01005888
Willy Tarreaub608feb2010-01-02 22:47:18 +01005889option http-server-close
5890no option http-server-close
5891 Enable or disable HTTP connection closing on the server side
5892 May be used in sections : defaults | frontend | listen | backend
5893 yes | yes | yes | yes
5894 Arguments : none
5895
Willy Tarreau70dffda2014-01-30 03:07:23 +01005896 By default HAProxy operates in keep-alive mode with regards to persistent
5897 connections: for each connection it processes each request and response, and
5898 leaves the connection idle on both sides between the end of a response and
5899 the start of a new request. This mode may be changed by several options such
5900 as "option http-server-close", "option forceclose", "option httpclose" or
5901 "option http-tunnel". Setting "option http-server-close" enables HTTP
5902 connection-close mode on the server side while keeping the ability to support
Davor Ocelice9ed2812017-12-25 17:49:28 +01005903 HTTP keep-alive and pipelining on the client side. This provides the lowest
Willy Tarreau70dffda2014-01-30 03:07:23 +01005904 latency on the client side (slow network) and the fastest session reuse on
5905 the server side to save server resources, similarly to "option forceclose".
5906 It also permits non-keepalive capable servers to be served in keep-alive mode
Lukas Tribus23953682017-04-28 13:24:30 +00005907 to the clients if they conform to the requirements of RFC7230. Please note
Willy Tarreau70dffda2014-01-30 03:07:23 +01005908 that some servers do not always conform to those requirements when they see
5909 "Connection: close" in the request. The effect will be that keep-alive will
5910 never be used. A workaround consists in enabling "option
5911 http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005912
5913 At the moment, logs will not indicate whether requests came from the same
5914 session or not. The accept date reported in the logs corresponds to the end
5915 of the previous request, and the request time corresponds to the time spent
5916 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01005917 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5918 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005919
5920 This option may be set both in a frontend and in a backend. It is enabled if
5921 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005922 It disables and replaces any previous "option httpclose", "option forceclose",
5923 "option http-tunnel" or "option http-keep-alive". Please check section 4
Willy Tarreau70dffda2014-01-30 03:07:23 +01005924 ("Proxies") to see how this option combines with others when frontend and
5925 backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005926
5927 If this option has been enabled in a "defaults" section, it can be disabled
5928 in a specific instance by prepending the "no" keyword before it.
5929
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005930 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01005931 "option httpclose", "option http-keep-alive", and
5932 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005933
5934
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005935option http-tunnel
5936no option http-tunnel
5937 Disable or enable HTTP connection processing after first transaction
5938 May be used in sections : defaults | frontend | listen | backend
5939 yes | yes | yes | yes
5940 Arguments : none
5941
Willy Tarreau70dffda2014-01-30 03:07:23 +01005942 By default HAProxy operates in keep-alive mode with regards to persistent
5943 connections: for each connection it processes each request and response, and
5944 leaves the connection idle on both sides between the end of a response and
5945 the start of a new request. This mode may be changed by several options such
5946 as "option http-server-close", "option forceclose", "option httpclose" or
5947 "option http-tunnel".
5948
5949 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005950 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01005951 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
5952 is normally not needed anymore unless in very specific cases such as when
5953 using an in-house protocol that looks like HTTP but is not compatible, or
5954 just to log one request per client in order to reduce log size. Note that
5955 everything which works at the HTTP level, including header parsing/addition,
5956 cookie processing or content switching will only work for the first request
5957 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005958
5959 If this option has been enabled in a "defaults" section, it can be disabled
5960 in a specific instance by prepending the "no" keyword before it.
5961
5962 See also : "option forceclose", "option http-server-close",
5963 "option httpclose", "option http-keep-alive", and
5964 "1.1. The HTTP transaction model".
5965
5966
Willy Tarreau88d349d2010-01-25 12:15:43 +01005967option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01005968no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01005969 Make use of non-standard Proxy-Connection header instead of Connection
5970 May be used in sections : defaults | frontend | listen | backend
5971 yes | yes | yes | no
5972 Arguments : none
5973
Lukas Tribus23953682017-04-28 13:24:30 +00005974 While RFC7230 explicitly states that HTTP/1.1 agents must use the
Willy Tarreau88d349d2010-01-25 12:15:43 +01005975 Connection header to indicate their wish of persistent or non-persistent
5976 connections, both browsers and proxies ignore this header for proxied
5977 connections and make use of the undocumented, non-standard Proxy-Connection
5978 header instead. The issue begins when trying to put a load balancer between
5979 browsers and such proxies, because there will be a difference between what
5980 haproxy understands and what the client and the proxy agree on.
5981
5982 By setting this option in a frontend, haproxy can automatically switch to use
5983 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01005984 defined here as one where the URI begins with neither a '/' nor a '*'. This
5985 is incompatible with the HTTP tunnel mode. Note that this option can only be
5986 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01005987
Willy Tarreau844a7e72010-01-31 21:46:18 +01005988 Also, when this option is set, a request which requires authentication will
5989 automatically switch to use proxy authentication headers if it is itself a
5990 proxied request. That makes it possible to check or enforce authentication in
5991 front of an existing proxy.
5992
Willy Tarreau88d349d2010-01-25 12:15:43 +01005993 This option should normally never be used, except in front of a proxy.
5994
5995 See also : "option httpclose", "option forceclose" and "option
5996 http-server-close".
5997
5998
Willy Tarreaud63335a2010-02-26 12:56:52 +01005999option httpchk
6000option httpchk <uri>
6001option httpchk <method> <uri>
6002option httpchk <method> <uri> <version>
6003 Enable HTTP protocol to check on the servers health
6004 May be used in sections : defaults | frontend | listen | backend
6005 yes | no | yes | yes
6006 Arguments :
6007 <method> is the optional HTTP method used with the requests. When not set,
6008 the "OPTIONS" method is used, as it generally requires low server
6009 processing and is easy to filter out from the logs. Any method
6010 may be used, though it is not recommended to invent non-standard
6011 ones.
6012
6013 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
6014 which is accessible by default on almost any server, but may be
6015 changed to any other URI. Query strings are permitted.
6016
6017 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
6018 but some servers might behave incorrectly in HTTP 1.0, so turning
6019 it to HTTP/1.1 may sometimes help. Note that the Host field is
6020 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
6021 after "\r\n" following the version string.
6022
6023 By default, server health checks only consist in trying to establish a TCP
6024 connection. When "option httpchk" is specified, a complete HTTP request is
6025 sent once the TCP connection is established, and responses 2xx and 3xx are
6026 considered valid, while all other ones indicate a server failure, including
6027 the lack of any response.
6028
6029 The port and interval are specified in the server configuration.
6030
6031 This option does not necessarily require an HTTP backend, it also works with
6032 plain TCP backends. This is particularly useful to check simple scripts bound
6033 to some dedicated ports using the inetd daemon.
6034
6035 Examples :
6036 # Relay HTTPS traffic to Apache instance and check service availability
6037 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
6038 backend https_relay
6039 mode tcp
6040 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
6041 server apache1 192.168.1.1:443 check port 80
6042
Simon Hormanafc47ee2013-11-25 10:46:35 +09006043 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
6044 "option pgsql-check", "http-check" and the "check", "port" and
6045 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006046
6047
Willy Tarreauc27debf2008-01-06 08:57:02 +01006048option httpclose
6049no option httpclose
6050 Enable or disable passive HTTP connection closing
6051 May be used in sections : defaults | frontend | listen | backend
6052 yes | yes | yes | yes
6053 Arguments : none
6054
Willy Tarreau70dffda2014-01-30 03:07:23 +01006055 By default HAProxy operates in keep-alive mode with regards to persistent
6056 connections: for each connection it processes each request and response, and
6057 leaves the connection idle on both sides between the end of a response and
6058 the start of a new request. This mode may be changed by several options such
Cyril Bonté653dcd62014-02-20 00:13:15 +01006059 as "option http-server-close", "option forceclose", "option httpclose" or
Willy Tarreau70dffda2014-01-30 03:07:23 +01006060 "option http-tunnel".
6061
6062 If "option httpclose" is set, HAProxy will work in HTTP tunnel mode and check
6063 if a "Connection: close" header is already set in each direction, and will
6064 add one if missing. Each end should react to this by actively closing the TCP
6065 connection after each transfer, thus resulting in a switch to the HTTP close
6066 mode. Any "Connection" header different from "close" will also be removed.
6067 Note that this option is deprecated since what it does is very cheap but not
6068 reliable. Using "option http-server-close" or "option forceclose" is strongly
6069 recommended instead.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006070
6071 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006072 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01006073 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
6074 it is possible to use the "option forceclose" which actively closes the
6075 request connection once the server responds. Option "forceclose" also
6076 releases the server connection earlier because it does not have to wait for
6077 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006078
6079 This option may be set both in a frontend and in a backend. It is enabled if
6080 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01006081 It disables and replaces any previous "option http-server-close",
6082 "option forceclose", "option http-keep-alive" or "option http-tunnel". Please
Willy Tarreau70dffda2014-01-30 03:07:23 +01006083 check section 4 ("Proxies") to see how this option combines with others when
6084 frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006085
6086 If this option has been enabled in a "defaults" section, it can be disabled
6087 in a specific instance by prepending the "no" keyword before it.
6088
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02006089 See also : "option forceclose", "option http-server-close" and
6090 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01006091
6092
Emeric Brun3a058f32009-06-30 18:26:00 +02006093option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006094 Enable logging of HTTP request, session state and timers
6095 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01006096 yes | yes | yes | no
Emeric Brun3a058f32009-06-30 18:26:00 +02006097 Arguments :
6098 clf if the "clf" argument is added, then the output format will be
6099 the CLF format instead of HAProxy's default HTTP format. You can
6100 use this when you need to feed HAProxy's logs through a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01006101 log analyzer which only support the CLF format and which is not
Emeric Brun3a058f32009-06-30 18:26:00 +02006102 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006103
6104 By default, the log output format is very poor, as it only contains the
6105 source and destination addresses, and the instance name. By specifying
6106 "option httplog", each log line turns into a much richer format including,
6107 but not limited to, the HTTP request, the connection timers, the session
6108 status, the connections numbers, the captured headers and cookies, the
6109 frontend, backend and server name, and of course the source address and
6110 ports.
6111
PiBa-NLbd556bf2014-12-11 21:31:54 +01006112 Specifying only "option httplog" will automatically clear the 'clf' mode
6113 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02006114
Guillaume de Lafond29f45602017-03-31 19:52:15 +02006115 "option httplog" overrides any previous "log-format" directive.
6116
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006117 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006118
Willy Tarreau55165fe2009-05-10 12:02:55 +02006119
6120option http_proxy
6121no option http_proxy
6122 Enable or disable plain HTTP proxy mode
6123 May be used in sections : defaults | frontend | listen | backend
6124 yes | yes | yes | yes
6125 Arguments : none
6126
6127 It sometimes happens that people need a pure HTTP proxy which understands
6128 basic proxy requests without caching nor any fancy feature. In this case,
6129 it may be worth setting up an HAProxy instance with the "option http_proxy"
6130 set. In this mode, no server is declared, and the connection is forwarded to
6131 the IP address and port found in the URL after the "http://" scheme.
6132
6133 No host address resolution is performed, so this only works when pure IP
6134 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006135 it will probably be used only by experts who know they need exactly it. This
6136 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02006137
6138 If this option has been enabled in a "defaults" section, it can be disabled
6139 in a specific instance by prepending the "no" keyword before it.
6140
6141 Example :
6142 # this backend understands HTTP proxy requests and forwards them directly.
6143 backend direct_forward
6144 option httpclose
6145 option http_proxy
6146
6147 See also : "option httpclose"
6148
Willy Tarreau211ad242009-10-03 21:45:07 +02006149
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006150option independent-streams
6151no option independent-streams
6152 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006153 May be used in sections : defaults | frontend | listen | backend
6154 yes | yes | yes | yes
6155 Arguments : none
6156
6157 By default, when data is sent over a socket, both the write timeout and the
6158 read timeout for that socket are refreshed, because we consider that there is
6159 activity on that socket, and we have no other means of guessing if we should
6160 receive data or not.
6161
Davor Ocelice9ed2812017-12-25 17:49:28 +01006162 While this default behavior is desirable for almost all applications, there
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006163 exists a situation where it is desirable to disable it, and only refresh the
6164 read timeout if there are incoming data. This happens on sessions with large
6165 timeouts and low amounts of exchanged data such as telnet session. If the
6166 server suddenly disappears, the output data accumulates in the system's
6167 socket buffers, both timeouts are correctly refreshed, and there is no way
6168 to know the server does not receive them, so we don't timeout. However, when
6169 the underlying protocol always echoes sent data, it would be enough by itself
6170 to detect the issue using the read timeout. Note that this problem does not
6171 happen with more verbose protocols because data won't accumulate long in the
6172 socket buffers.
6173
6174 When this option is set on the frontend, it will disable read timeout updates
6175 on data sent to the client. There probably is little use of this case. When
6176 the option is set on the backend, it will disable read timeout updates on
6177 data sent to the server. Doing so will typically break large HTTP posts from
6178 slow lines, so use it with caution.
6179
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006180 Note: older versions used to call this setting "option independent-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006181 with a spelling mistake. This spelling is still supported but
6182 deprecated.
6183
Willy Tarreauce887fd2012-05-12 12:50:00 +02006184 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006185
6186
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02006187option ldap-check
6188 Use LDAPv3 health checks for server testing
6189 May be used in sections : defaults | frontend | listen | backend
6190 yes | no | yes | yes
6191 Arguments : none
6192
6193 It is possible to test that the server correctly talks LDAPv3 instead of just
6194 testing that it accepts the TCP connection. When this option is set, an
6195 LDAPv3 anonymous simple bind message is sent to the server, and the response
6196 is analyzed to find an LDAPv3 bind response message.
6197
6198 The server is considered valid only when the LDAP response contains success
6199 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
6200
6201 Logging of bind requests is server dependent see your documentation how to
6202 configure it.
6203
6204 Example :
6205 option ldap-check
6206
6207 See also : "option httpchk"
6208
6209
Simon Horman98637e52014-06-20 12:30:16 +09006210option external-check
6211 Use external processes for server health checks
6212 May be used in sections : defaults | frontend | listen | backend
6213 yes | no | yes | yes
6214
6215 It is possible to test the health of a server using an external command.
6216 This is achieved by running the executable set using "external-check
6217 command".
6218
6219 Requires the "external-check" global to be set.
6220
6221 See also : "external-check", "external-check command", "external-check path"
6222
6223
Willy Tarreau211ad242009-10-03 21:45:07 +02006224option log-health-checks
6225no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02006226 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02006227 May be used in sections : defaults | frontend | listen | backend
6228 yes | no | yes | yes
6229 Arguments : none
6230
Willy Tarreaubef1b322014-05-13 21:01:39 +02006231 By default, failed health check are logged if server is UP and successful
6232 health checks are logged if server is DOWN, so the amount of additional
6233 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02006234
Willy Tarreaubef1b322014-05-13 21:01:39 +02006235 When this option is enabled, any change of the health check status or to
6236 the server's health will be logged, so that it becomes possible to know
6237 that a server was failing occasional checks before crashing, or exactly when
6238 it failed to respond a valid HTTP status, then when the port started to
6239 reject connections, then when the server stopped responding at all.
6240
Davor Ocelice9ed2812017-12-25 17:49:28 +01006241 Note that status changes not caused by health checks (e.g. enable/disable on
Willy Tarreaubef1b322014-05-13 21:01:39 +02006242 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02006243
Willy Tarreaubef1b322014-05-13 21:01:39 +02006244 See also: "option httpchk", "option ldap-check", "option mysql-check",
6245 "option pgsql-check", "option redis-check", "option smtpchk",
6246 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02006247
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006248
6249option log-separate-errors
6250no option log-separate-errors
6251 Change log level for non-completely successful connections
6252 May be used in sections : defaults | frontend | listen | backend
6253 yes | yes | yes | no
6254 Arguments : none
6255
6256 Sometimes looking for errors in logs is not easy. This option makes haproxy
6257 raise the level of logs containing potentially interesting information such
6258 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
6259 level changes from "info" to "err". This makes it possible to log them
6260 separately to a different file with most syslog daemons. Be careful not to
6261 remove them from the original file, otherwise you would lose ordering which
6262 provides very important information.
6263
6264 Using this option, large sites dealing with several thousand connections per
6265 second may log normal traffic to a rotating buffer and only archive smaller
6266 error logs.
6267
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006268 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006269 logging.
6270
Willy Tarreauc27debf2008-01-06 08:57:02 +01006271
6272option logasap
6273no option logasap
6274 Enable or disable early logging of HTTP requests
6275 May be used in sections : defaults | frontend | listen | backend
6276 yes | yes | yes | no
6277 Arguments : none
6278
6279 By default, HTTP requests are logged upon termination so that the total
6280 transfer time and the number of bytes appear in the logs. When large objects
6281 are being transferred, it may take a while before the request appears in the
6282 logs. Using "option logasap", the request gets logged as soon as the server
6283 sends the complete headers. The only missing information in the logs will be
6284 the total number of bytes which will indicate everything except the amount
6285 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006286 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01006287 "Content-Length" response header so that the logs at least indicate how many
6288 bytes are expected to be transferred.
6289
Willy Tarreaucc6c8912009-02-22 10:53:55 +01006290 Examples :
6291 listen http_proxy 0.0.0.0:80
6292 mode http
6293 option httplog
6294 option logasap
6295 log 192.168.2.200 local3
6296
6297 >>> Feb 6 12:14:14 localhost \
6298 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
6299 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
6300 "GET /image.iso HTTP/1.0"
6301
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006302 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01006303 logging.
6304
6305
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006306option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006307 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006308 May be used in sections : defaults | frontend | listen | backend
6309 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006310 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006311 <username> This is the username which will be used when connecting to MySQL
6312 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006313 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006314
6315 If you specify a username, the check consists of sending two MySQL packet,
6316 one Client Authentication packet, and one QUIT packet, to correctly close
Davor Ocelice9ed2812017-12-25 17:49:28 +01006317 MySQL session. We then parse the MySQL Handshake Initialization packet and/or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006318 Error packet. It is a basic but useful test which does not produce error nor
6319 aborted connect on the server. However, it requires adding an authorization
6320 in the MySQL table, like this :
6321
6322 USE mysql;
6323 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
6324 FLUSH PRIVILEGES;
6325
6326 If you don't specify a username (it is deprecated and not recommended), the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006327 check only consists in parsing the Mysql Handshake Initialization packet or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006328 Error packet, we don't send anything in this mode. It was reported that it
6329 can generate lockout if check is too frequent and/or if there is not enough
6330 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
6331 value as if a connection is established successfully within fewer than MySQL
6332 "max_connect_errors" attempts after a previous connection was interrupted,
6333 the error count for the host is cleared to zero. If HAProxy's server get
6334 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
6335
6336 Remember that this does not check database presence nor database consistency.
6337 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006338
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02006339 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006340
6341 Most often, an incoming MySQL server needs to see the client's IP address for
6342 various purposes, including IP privilege matching and connection logging.
6343 When possible, it is often wise to masquerade the client's IP address when
6344 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006345 which requires the transparent proxy feature to be compiled in, and the MySQL
6346 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006347
6348 See also: "option httpchk"
6349
6350
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006351option nolinger
6352no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006353 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006354 May be used in sections: defaults | frontend | listen | backend
6355 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006356 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006357
Davor Ocelice9ed2812017-12-25 17:49:28 +01006358 When clients or servers abort connections in a dirty way (e.g. they are
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006359 physically disconnected), the session timeouts triggers and the session is
6360 closed. But it will remain in FIN_WAIT1 state for some time in the system,
6361 using some resources and possibly limiting the ability to establish newer
6362 connections.
6363
6364 When this happens, it is possible to activate "option nolinger" which forces
6365 the system to immediately remove any socket's pending data on close. Thus,
6366 the session is instantly purged from the system's tables. This usually has
6367 side effects such as increased number of TCP resets due to old retransmits
6368 getting immediately rejected. Some firewalls may sometimes complain about
6369 this too.
6370
6371 For this reason, it is not recommended to use this option when not absolutely
6372 needed. You know that you need it when you have thousands of FIN_WAIT1
6373 sessions on your system (TIME_WAIT ones do not count).
6374
6375 This option may be used both on frontends and backends, depending on the side
6376 where it is required. Use it on the frontend for clients, and on the backend
6377 for servers.
6378
6379 If this option has been enabled in a "defaults" section, it can be disabled
6380 in a specific instance by prepending the "no" keyword before it.
6381
6382
Willy Tarreau55165fe2009-05-10 12:02:55 +02006383option originalto [ except <network> ] [ header <name> ]
6384 Enable insertion of the X-Original-To header to requests sent to servers
6385 May be used in sections : defaults | frontend | listen | backend
6386 yes | yes | yes | yes
6387 Arguments :
6388 <network> is an optional argument used to disable this option for sources
6389 matching <network>
6390 <name> an optional argument to specify a different "X-Original-To"
6391 header name.
6392
6393 Since HAProxy can work in transparent mode, every request from a client can
6394 be redirected to the proxy and HAProxy itself can proxy every request to a
6395 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
6396 be lost. This is annoying when you want access rules based on destination ip
6397 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
6398 added by HAProxy to all requests sent to the server. This header contains a
6399 value representing the original destination IP address. Since this must be
6400 configured to always use the last occurrence of this header only. Note that
6401 only the last occurrence of the header must be used, since it is really
6402 possible that the client has already brought one.
6403
6404 The keyword "header" may be used to supply a different header name to replace
6405 the default "X-Original-To". This can be useful where you might already
6406 have a "X-Original-To" header from a different application, and you need
6407 preserve it. Also if your backend server doesn't use the "X-Original-To"
6408 header and requires different one.
6409
6410 Sometimes, a same HAProxy instance may be shared between a direct client
6411 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6412 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6413 header for a known source address or network by adding the "except" keyword
6414 followed by the network address. In this case, any source IP matching the
6415 network will not cause an addition of this header. Most common uses are with
6416 private networks or 127.0.0.1.
6417
6418 This option may be specified either in the frontend or in the backend. If at
6419 least one of them uses it, the header will be added. Note that the backend's
6420 setting of the header subargument takes precedence over the frontend's if
6421 both are defined.
6422
Willy Tarreau55165fe2009-05-10 12:02:55 +02006423 Examples :
6424 # Original Destination address
6425 frontend www
6426 mode http
6427 option originalto except 127.0.0.1
6428
6429 # Those servers want the IP Address in X-Client-Dst
6430 backend www
6431 mode http
6432 option originalto header X-Client-Dst
6433
Willy Tarreau87cf5142011-08-19 22:57:24 +02006434 See also : "option httpclose", "option http-server-close",
6435 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02006436
6437
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006438option persist
6439no option persist
6440 Enable or disable forced persistence on down servers
6441 May be used in sections: defaults | frontend | listen | backend
6442 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006443 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006444
6445 When an HTTP request reaches a backend with a cookie which references a dead
6446 server, by default it is redispatched to another server. It is possible to
6447 force the request to be sent to the dead server first using "option persist"
6448 if absolutely needed. A common use case is when servers are under extreme
6449 load and spend their time flapping. In this case, the users would still be
6450 directed to the server they opened the session on, in the hope they would be
6451 correctly served. It is recommended to use "option redispatch" in conjunction
6452 with this option so that in the event it would not be possible to connect to
6453 the server at all (server definitely dead), the client would finally be
6454 redirected to another valid server.
6455
6456 If this option has been enabled in a "defaults" section, it can be disabled
6457 in a specific instance by prepending the "no" keyword before it.
6458
Willy Tarreau4de91492010-01-22 19:10:05 +01006459 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006460
6461
Willy Tarreau0c122822013-12-15 18:49:01 +01006462option pgsql-check [ user <username> ]
6463 Use PostgreSQL health checks for server testing
6464 May be used in sections : defaults | frontend | listen | backend
6465 yes | no | yes | yes
6466 Arguments :
6467 <username> This is the username which will be used when connecting to
6468 PostgreSQL server.
6469
6470 The check sends a PostgreSQL StartupMessage and waits for either
6471 Authentication request or ErrorResponse message. It is a basic but useful
6472 test which does not produce error nor aborted connect on the server.
6473 This check is identical with the "mysql-check".
6474
6475 See also: "option httpchk"
6476
6477
Willy Tarreau9420b122013-12-15 18:58:25 +01006478option prefer-last-server
6479no option prefer-last-server
6480 Allow multiple load balanced requests to remain on the same server
6481 May be used in sections: defaults | frontend | listen | backend
6482 yes | no | yes | yes
6483 Arguments : none
6484
6485 When the load balancing algorithm in use is not deterministic, and a previous
6486 request was sent to a server to which haproxy still holds a connection, it is
6487 sometimes desirable that subsequent requests on a same session go to the same
6488 server as much as possible. Note that this is different from persistence, as
6489 we only indicate a preference which haproxy tries to apply without any form
6490 of warranty. The real use is for keep-alive connections sent to servers. When
6491 this option is used, haproxy will try to reuse the same connection that is
6492 attached to the server instead of rebalancing to another server, causing a
6493 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01006494 not make much sense to use this in combination with hashing algorithms. Note,
6495 haproxy already automatically tries to stick to a server which sends a 401 or
6496 to a proxy which sends a 407 (authentication required). This is mandatory for
6497 use with the broken NTLM authentication challenge, and significantly helps in
6498 troubleshooting some faulty applications. Option prefer-last-server might be
6499 desirable in these environments as well, to avoid redistributing the traffic
6500 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01006501
6502 If this option has been enabled in a "defaults" section, it can be disabled
6503 in a specific instance by prepending the "no" keyword before it.
6504
6505 See also: "option http-keep-alive"
6506
6507
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006508option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07006509option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006510no option redispatch
6511 Enable or disable session redistribution in case of connection failure
6512 May be used in sections: defaults | frontend | listen | backend
6513 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07006514 Arguments :
6515 <interval> The optional integer value that controls how often redispatches
6516 occur when retrying connections. Positive value P indicates a
6517 redispatch is desired on every Pth retry, and negative value
Davor Ocelice9ed2812017-12-25 17:49:28 +01006518 N indicate a redispatch is desired on the Nth retry prior to the
Joseph Lynch726ab712015-05-11 23:25:34 -07006519 last retry. For example, the default of -1 preserves the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006520 historical behavior of redispatching on the last retry, a
Joseph Lynch726ab712015-05-11 23:25:34 -07006521 positive value of 1 would indicate a redispatch on every retry,
6522 and a positive value of 3 would indicate a redispatch on every
6523 third retry. You can disable redispatches with a value of 0.
6524
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006525
6526 In HTTP mode, if a server designated by a cookie is down, clients may
6527 definitely stick to it because they cannot flush the cookie, so they will not
6528 be able to access the service anymore.
6529
6530 Specifying "option redispatch" will allow the proxy to break their
6531 persistence and redistribute them to a working server.
6532
Joseph Lynch726ab712015-05-11 23:25:34 -07006533 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006534 connection failures. Of course, it requires having "retries" set to a nonzero
6535 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006536
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006537 This form is the preferred form, which replaces both the "redispatch" and
6538 "redisp" keywords.
6539
6540 If this option has been enabled in a "defaults" section, it can be disabled
6541 in a specific instance by prepending the "no" keyword before it.
6542
Willy Tarreau4de91492010-01-22 19:10:05 +01006543 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006544
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006545
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02006546option redis-check
6547 Use redis health checks for server testing
6548 May be used in sections : defaults | frontend | listen | backend
6549 yes | no | yes | yes
6550 Arguments : none
6551
6552 It is possible to test that the server correctly talks REDIS protocol instead
6553 of just testing that it accepts the TCP connection. When this option is set,
6554 a PING redis command is sent to the server, and the response is analyzed to
6555 find the "+PONG" response message.
6556
6557 Example :
6558 option redis-check
6559
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03006560 See also : "option httpchk", "option tcp-check", "tcp-check expect"
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02006561
6562
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006563option smtpchk
6564option smtpchk <hello> <domain>
6565 Use SMTP health checks for server testing
6566 May be used in sections : defaults | frontend | listen | backend
6567 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01006568 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006569 <hello> is an optional argument. It is the "hello" command to use. It can
6570 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
6571 values will be turned into the default command ("HELO").
6572
6573 <domain> is the domain name to present to the server. It may only be
6574 specified (and is mandatory) if the hello command has been
6575 specified. By default, "localhost" is used.
6576
6577 When "option smtpchk" is set, the health checks will consist in TCP
6578 connections followed by an SMTP command. By default, this command is
6579 "HELO localhost". The server's return code is analyzed and only return codes
6580 starting with a "2" will be considered as valid. All other responses,
6581 including a lack of response will constitute an error and will indicate a
6582 dead server.
6583
6584 This test is meant to be used with SMTP servers or relays. Depending on the
6585 request, it is possible that some servers do not log each connection attempt,
Davor Ocelice9ed2812017-12-25 17:49:28 +01006586 so you may want to experiment to improve the behavior. Using telnet on port
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006587 25 is often easier than adjusting the configuration.
6588
6589 Most often, an incoming SMTP server needs to see the client's IP address for
6590 various purposes, including spam filtering, anti-spoofing and logging. When
6591 possible, it is often wise to masquerade the client's IP address when
6592 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006593 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006594
6595 Example :
6596 option smtpchk HELO mydomain.org
6597
6598 See also : "option httpchk", "source"
6599
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006600
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02006601option socket-stats
6602no option socket-stats
6603
6604 Enable or disable collecting & providing separate statistics for each socket.
6605 May be used in sections : defaults | frontend | listen | backend
6606 yes | yes | yes | no
6607
6608 Arguments : none
6609
6610
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006611option splice-auto
6612no option splice-auto
6613 Enable or disable automatic kernel acceleration on sockets in both directions
6614 May be used in sections : defaults | frontend | listen | backend
6615 yes | yes | yes | yes
6616 Arguments : none
6617
6618 When this option is enabled either on a frontend or on a backend, haproxy
6619 will automatically evaluate the opportunity to use kernel tcp splicing to
Davor Ocelice9ed2812017-12-25 17:49:28 +01006620 forward data between the client and the server, in either direction. HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006621 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006622 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006623 are not much aggressive in order to limit excessive use of splicing. This
6624 option requires splicing to be enabled at compile time, and may be globally
6625 disabled with the global option "nosplice". Since splice uses pipes, using it
6626 requires that there are enough spare pipes.
6627
6628 Important note: kernel-based TCP splicing is a Linux-specific feature which
6629 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
6630 transfer data between sockets without copying these data to user-space, thus
6631 providing noticeable performance gains and CPU cycles savings. Since many
6632 early implementations are buggy, corrupt data and/or are inefficient, this
6633 feature is not enabled by default, and it should be used with extreme care.
6634 While it is not possible to detect the correctness of an implementation,
6635 2.6.29 is the first version offering a properly working implementation. In
6636 case of doubt, splicing may be globally disabled using the global "nosplice"
6637 keyword.
6638
6639 Example :
6640 option splice-auto
6641
6642 If this option has been enabled in a "defaults" section, it can be disabled
6643 in a specific instance by prepending the "no" keyword before it.
6644
6645 See also : "option splice-request", "option splice-response", and global
6646 options "nosplice" and "maxpipes"
6647
6648
6649option splice-request
6650no option splice-request
6651 Enable or disable automatic kernel acceleration on sockets for requests
6652 May be used in sections : defaults | frontend | listen | backend
6653 yes | yes | yes | yes
6654 Arguments : none
6655
6656 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006657 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006658 the client to the server. It might still use the recv/send scheme if there
6659 are no spare pipes left. This option requires splicing to be enabled at
6660 compile time, and may be globally disabled with the global option "nosplice".
6661 Since splice uses pipes, using it requires that there are enough spare pipes.
6662
6663 Important note: see "option splice-auto" for usage limitations.
6664
6665 Example :
6666 option splice-request
6667
6668 If this option has been enabled in a "defaults" section, it can be disabled
6669 in a specific instance by prepending the "no" keyword before it.
6670
6671 See also : "option splice-auto", "option splice-response", and global options
6672 "nosplice" and "maxpipes"
6673
6674
6675option splice-response
6676no option splice-response
6677 Enable or disable automatic kernel acceleration on sockets for responses
6678 May be used in sections : defaults | frontend | listen | backend
6679 yes | yes | yes | yes
6680 Arguments : none
6681
6682 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006683 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006684 the server to the client. It might still use the recv/send scheme if there
6685 are no spare pipes left. This option requires splicing to be enabled at
6686 compile time, and may be globally disabled with the global option "nosplice".
6687 Since splice uses pipes, using it requires that there are enough spare pipes.
6688
6689 Important note: see "option splice-auto" for usage limitations.
6690
6691 Example :
6692 option splice-response
6693
6694 If this option has been enabled in a "defaults" section, it can be disabled
6695 in a specific instance by prepending the "no" keyword before it.
6696
6697 See also : "option splice-auto", "option splice-request", and global options
6698 "nosplice" and "maxpipes"
6699
6700
Christopher Fauletba7bc162016-11-07 21:07:38 +01006701option spop-check
6702 Use SPOP health checks for server testing
6703 May be used in sections : defaults | frontend | listen | backend
6704 no | no | no | yes
6705 Arguments : none
6706
6707 It is possible to test that the server correctly talks SPOP protocol instead
6708 of just testing that it accepts the TCP connection. When this option is set,
6709 a HELLO handshake is performed between HAProxy and the server, and the
6710 response is analyzed to check no error is reported.
6711
6712 Example :
6713 option spop-check
6714
6715 See also : "option httpchk"
6716
6717
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006718option srvtcpka
6719no option srvtcpka
6720 Enable or disable the sending of TCP keepalive packets on the server side
6721 May be used in sections : defaults | frontend | listen | backend
6722 yes | no | yes | yes
6723 Arguments : none
6724
6725 When there is a firewall or any session-aware component between a client and
6726 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01006727 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006728 components decides to expire a session which has remained idle for too long.
6729
6730 Enabling socket-level TCP keep-alives makes the system regularly send packets
6731 to the other end of the connection, leaving it active. The delay between
6732 keep-alive probes is controlled by the system only and depends both on the
6733 operating system and its tuning parameters.
6734
6735 It is important to understand that keep-alive packets are neither emitted nor
6736 received at the application level. It is only the network stacks which sees
6737 them. For this reason, even if one side of the proxy already uses keep-alives
6738 to maintain its connection alive, those keep-alive packets will not be
6739 forwarded to the other side of the proxy.
6740
6741 Please note that this has nothing to do with HTTP keep-alive.
6742
6743 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
6744 server side of a connection, which should help when session expirations are
6745 noticed between HAProxy and a server.
6746
6747 If this option has been enabled in a "defaults" section, it can be disabled
6748 in a specific instance by prepending the "no" keyword before it.
6749
6750 See also : "option clitcpka", "option tcpka"
6751
6752
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006753option ssl-hello-chk
6754 Use SSLv3 client hello health checks for server testing
6755 May be used in sections : defaults | frontend | listen | backend
6756 yes | no | yes | yes
6757 Arguments : none
6758
6759 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
6760 possible to test that the server correctly talks SSL instead of just testing
6761 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
6762 SSLv3 client hello messages are sent once the connection is established to
6763 the server, and the response is analyzed to find an SSL server hello message.
6764 The server is considered valid only when the response contains this server
6765 hello message.
6766
6767 All servers tested till there correctly reply to SSLv3 client hello messages,
6768 and most servers tested do not even log the requests containing only hello
6769 messages, which is appreciable.
6770
Willy Tarreau763a95b2012-10-04 23:15:39 +02006771 Note that this check works even when SSL support was not built into haproxy
6772 because it forges the SSL message. When SSL support is available, it is best
6773 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006774
Willy Tarreau763a95b2012-10-04 23:15:39 +02006775 See also: "option httpchk", "check-ssl"
6776
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006777
Willy Tarreaued179852013-12-16 01:07:00 +01006778option tcp-check
6779 Perform health checks using tcp-check send/expect sequences
6780 May be used in sections: defaults | frontend | listen | backend
6781 yes | no | yes | yes
6782
6783 This health check method is intended to be combined with "tcp-check" command
6784 lists in order to support send/expect types of health check sequences.
6785
6786 TCP checks currently support 4 modes of operations :
6787 - no "tcp-check" directive : the health check only consists in a connection
6788 attempt, which remains the default mode.
6789
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006790 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006791 used to send a string along with a connection opening. With some
6792 protocols, it helps sending a "QUIT" message for example that prevents
6793 the server from logging a connection error for each health check. The
6794 check result will still be based on the ability to open the connection
6795 only.
6796
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006797 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01006798 The connection is opened and haproxy waits for the server to present some
6799 contents which must validate some rules. The check result will be based
6800 on the matching between the contents and the rules. This is suited for
6801 POP, IMAP, SMTP, FTP, SSH, TELNET.
6802
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006803 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Davor Ocelice9ed2812017-12-25 17:49:28 +01006804 used to test a hello-type protocol. HAProxy sends a message, the server
6805 responds and its response is analyzed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006806 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01006807 suited for protocols which require a binding or a request/response model.
6808 LDAP, MySQL, Redis and SSL are example of such protocols, though they
6809 already all have their dedicated checks with a deeper understanding of
6810 the respective protocols.
6811 In this mode, many questions may be sent and many answers may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01006812 analyzed.
Willy Tarreaued179852013-12-16 01:07:00 +01006813
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006814 A fifth mode can be used to insert comments in different steps of the
6815 script.
6816
6817 For each tcp-check rule you create, you can add a "comment" directive,
6818 followed by a string. This string will be reported in the log and stderr
6819 in debug mode. It is useful to make user-friendly error reporting.
6820 The "comment" is of course optional.
6821
6822
Willy Tarreaued179852013-12-16 01:07:00 +01006823 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01006824 # perform a POP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01006825 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006826 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006827
Davor Ocelice9ed2812017-12-25 17:49:28 +01006828 # perform an IMAP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01006829 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006830 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006831
6832 # look for the redis master server after ensuring it speaks well
6833 # redis protocol, then it exits properly.
Davor Ocelice9ed2812017-12-25 17:49:28 +01006834 # (send a command then analyze the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01006835 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006836 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006837 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02006838 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006839 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01006840 tcp-check send info\ replication\r\n
6841 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006842 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006843 tcp-check send QUIT\r\n
6844 tcp-check expect string +OK
6845
Davor Ocelice9ed2812017-12-25 17:49:28 +01006846 forge a HTTP request, then analyze the response
Willy Tarreaued179852013-12-16 01:07:00 +01006847 (send many headers before analyzing)
6848 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006849 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01006850 tcp-check send HEAD\ /\ HTTP/1.1\r\n
6851 tcp-check send Host:\ www.mydomain.com\r\n
6852 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
6853 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006854 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01006855
6856
6857 See also : "tcp-check expect", "tcp-check send"
6858
6859
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006860option tcp-smart-accept
6861no option tcp-smart-accept
6862 Enable or disable the saving of one ACK packet during the accept sequence
6863 May be used in sections : defaults | frontend | listen | backend
6864 yes | yes | yes | no
6865 Arguments : none
6866
6867 When an HTTP connection request comes in, the system acknowledges it on
6868 behalf of HAProxy, then the client immediately sends its request, and the
6869 system acknowledges it too while it is notifying HAProxy about the new
6870 connection. HAProxy then reads the request and responds. This means that we
6871 have one TCP ACK sent by the system for nothing, because the request could
6872 very well be acknowledged by HAProxy when it sends its response.
6873
6874 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
6875 sending this useless ACK on platforms which support it (currently at least
6876 Linux). It must not cause any problem, because the system will send it anyway
6877 after 40 ms if the response takes more time than expected to come.
6878
6879 During complex network debugging sessions, it may be desirable to disable
6880 this optimization because delayed ACKs can make troubleshooting more complex
6881 when trying to identify where packets are delayed. It is then possible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01006882 fall back to normal behavior by specifying "no option tcp-smart-accept".
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006883
6884 It is also possible to force it for non-HTTP proxies by simply specifying
6885 "option tcp-smart-accept". For instance, it can make sense with some services
6886 such as SMTP where the server speaks first.
6887
6888 It is recommended to avoid forcing this option in a defaults section. In case
6889 of doubt, consider setting it back to automatic values by prepending the
6890 "default" keyword before it, or disabling it using the "no" keyword.
6891
Willy Tarreaud88edf22009-06-14 15:48:17 +02006892 See also : "option tcp-smart-connect"
6893
6894
6895option tcp-smart-connect
6896no option tcp-smart-connect
6897 Enable or disable the saving of one ACK packet during the connect sequence
6898 May be used in sections : defaults | frontend | listen | backend
6899 yes | no | yes | yes
6900 Arguments : none
6901
6902 On certain systems (at least Linux), HAProxy can ask the kernel not to
6903 immediately send an empty ACK upon a connection request, but to directly
6904 send the buffer request instead. This saves one packet on the network and
6905 thus boosts performance. It can also be useful for some servers, because they
6906 immediately get the request along with the incoming connection.
6907
6908 This feature is enabled when "option tcp-smart-connect" is set in a backend.
6909 It is not enabled by default because it makes network troubleshooting more
6910 complex.
6911
6912 It only makes sense to enable it with protocols where the client speaks first
6913 such as HTTP. In other situations, if there is no data to send in place of
6914 the ACK, a normal ACK is sent.
6915
6916 If this option has been enabled in a "defaults" section, it can be disabled
6917 in a specific instance by prepending the "no" keyword before it.
6918
6919 See also : "option tcp-smart-accept"
6920
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006921
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006922option tcpka
6923 Enable or disable the sending of TCP keepalive packets on both sides
6924 May be used in sections : defaults | frontend | listen | backend
6925 yes | yes | yes | yes
6926 Arguments : none
6927
6928 When there is a firewall or any session-aware component between a client and
6929 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01006930 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006931 components decides to expire a session which has remained idle for too long.
6932
6933 Enabling socket-level TCP keep-alives makes the system regularly send packets
6934 to the other end of the connection, leaving it active. The delay between
6935 keep-alive probes is controlled by the system only and depends both on the
6936 operating system and its tuning parameters.
6937
6938 It is important to understand that keep-alive packets are neither emitted nor
6939 received at the application level. It is only the network stacks which sees
6940 them. For this reason, even if one side of the proxy already uses keep-alives
6941 to maintain its connection alive, those keep-alive packets will not be
6942 forwarded to the other side of the proxy.
6943
6944 Please note that this has nothing to do with HTTP keep-alive.
6945
6946 Using option "tcpka" enables the emission of TCP keep-alive probes on both
6947 the client and server sides of a connection. Note that this is meaningful
6948 only in "defaults" or "listen" sections. If this option is used in a
6949 frontend, only the client side will get keep-alives, and if this option is
6950 used in a backend, only the server side will get keep-alives. For this
6951 reason, it is strongly recommended to explicitly use "option clitcpka" and
6952 "option srvtcpka" when the configuration is split between frontends and
6953 backends.
6954
6955 See also : "option clitcpka", "option srvtcpka"
6956
Willy Tarreau844e3c52008-01-11 16:28:18 +01006957
6958option tcplog
6959 Enable advanced logging of TCP connections with session state and timers
6960 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01006961 yes | yes | yes | no
Willy Tarreau844e3c52008-01-11 16:28:18 +01006962 Arguments : none
6963
6964 By default, the log output format is very poor, as it only contains the
6965 source and destination addresses, and the instance name. By specifying
6966 "option tcplog", each log line turns into a much richer format including, but
6967 not limited to, the connection timers, the session status, the connections
6968 numbers, the frontend, backend and server name, and of course the source
6969 address and ports. This option is useful for pure TCP proxies in order to
6970 find which of the client or server disconnects or times out. For normal HTTP
6971 proxies, it's better to use "option httplog" which is even more complete.
6972
Guillaume de Lafond29f45602017-03-31 19:52:15 +02006973 "option tcplog" overrides any previous "log-format" directive.
6974
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006975 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006976
6977
Willy Tarreau844e3c52008-01-11 16:28:18 +01006978option transparent
6979no option transparent
6980 Enable client-side transparent proxying
6981 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01006982 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01006983 Arguments : none
6984
6985 This option was introduced in order to provide layer 7 persistence to layer 3
6986 load balancers. The idea is to use the OS's ability to redirect an incoming
6987 connection for a remote address to a local process (here HAProxy), and let
6988 this process know what address was initially requested. When this option is
6989 used, sessions without cookies will be forwarded to the original destination
6990 IP address of the incoming request (which should match that of another
6991 equipment), while requests with cookies will still be forwarded to the
6992 appropriate server.
6993
6994 Note that contrary to a common belief, this option does NOT make HAProxy
6995 present the client's IP to the server when establishing the connection.
6996
Willy Tarreaua1146052011-03-01 09:51:54 +01006997 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006998 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006999
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007000
Simon Horman98637e52014-06-20 12:30:16 +09007001external-check command <command>
7002 Executable to run when performing an external-check
7003 May be used in sections : defaults | frontend | listen | backend
7004 yes | no | yes | yes
7005
7006 Arguments :
7007 <command> is the external command to run
7008
Simon Horman98637e52014-06-20 12:30:16 +09007009 The arguments passed to the to the command are:
7010
Cyril Bonté777be862014-12-02 21:21:35 +01007011 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09007012
Cyril Bonté777be862014-12-02 21:21:35 +01007013 The <proxy_address> and <proxy_port> are derived from the first listener
7014 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
7015 listener the proxy_address will be the path of the socket and the
7016 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
7017 possible to determine a listener, and both <proxy_address> and <proxy_port>
7018 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09007019
Cyril Bonté72cda2a2014-12-27 22:28:39 +01007020 Some values are also provided through environment variables.
7021
7022 Environment variables :
7023 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
7024 applicable, for example in a "backend" section).
7025
7026 HAPROXY_PROXY_ID The backend id.
7027
7028 HAPROXY_PROXY_NAME The backend name.
7029
7030 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
7031 applicable, for example in a "backend" section or
7032 for a UNIX socket).
7033
7034 HAPROXY_SERVER_ADDR The server address.
7035
7036 HAPROXY_SERVER_CURCONN The current number of connections on the server.
7037
7038 HAPROXY_SERVER_ID The server id.
7039
7040 HAPROXY_SERVER_MAXCONN The server max connections.
7041
7042 HAPROXY_SERVER_NAME The server name.
7043
7044 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
7045 socket).
7046
7047 PATH The PATH environment variable used when executing
7048 the command may be set using "external-check path".
7049
Simon Horman98637e52014-06-20 12:30:16 +09007050 If the command executed and exits with a zero status then the check is
7051 considered to have passed, otherwise the check is considered to have
7052 failed.
7053
7054 Example :
7055 external-check command /bin/true
7056
7057 See also : "external-check", "option external-check", "external-check path"
7058
7059
7060external-check path <path>
7061 The value of the PATH environment variable used when running an external-check
7062 May be used in sections : defaults | frontend | listen | backend
7063 yes | no | yes | yes
7064
7065 Arguments :
7066 <path> is the path used when executing external command to run
7067
7068 The default path is "".
7069
7070 Example :
7071 external-check path "/usr/bin:/bin"
7072
7073 See also : "external-check", "option external-check",
7074 "external-check command"
7075
7076
Emeric Brun647caf12009-06-30 17:57:00 +02007077persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02007078persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02007079 Enable RDP cookie-based persistence
7080 May be used in sections : defaults | frontend | listen | backend
7081 yes | no | yes | yes
7082 Arguments :
7083 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02007084 default cookie name "msts" will be used. There currently is no
7085 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02007086
7087 This statement enables persistence based on an RDP cookie. The RDP cookie
7088 contains all information required to find the server in the list of known
Davor Ocelice9ed2812017-12-25 17:49:28 +01007089 servers. So when this option is set in the backend, the request is analyzed
Emeric Brun647caf12009-06-30 17:57:00 +02007090 and if an RDP cookie is found, it is decoded. If it matches a known server
7091 which is still UP (or if "option persist" is set), then the connection is
7092 forwarded to this server.
7093
7094 Note that this only makes sense in a TCP backend, but for this to work, the
7095 frontend must have waited long enough to ensure that an RDP cookie is present
7096 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007097 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02007098 a single "listen" section.
7099
Willy Tarreau61e28f22010-05-16 22:31:05 +02007100 Also, it is important to understand that the terminal server will emit this
7101 RDP cookie only if it is configured for "token redirection mode", which means
7102 that the "IP address redirection" option is disabled.
7103
Emeric Brun647caf12009-06-30 17:57:00 +02007104 Example :
7105 listen tse-farm
7106 bind :3389
7107 # wait up to 5s for an RDP cookie in the request
7108 tcp-request inspect-delay 5s
7109 tcp-request content accept if RDP_COOKIE
7110 # apply RDP cookie persistence
7111 persist rdp-cookie
7112 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02007113 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02007114 balance rdp-cookie
7115 server srv1 1.1.1.1:3389
7116 server srv2 1.1.1.2:3389
7117
Simon Hormanab814e02011-06-24 14:50:20 +09007118 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
7119 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02007120
7121
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007122rate-limit sessions <rate>
7123 Set a limit on the number of new sessions accepted per second on a frontend
7124 May be used in sections : defaults | frontend | listen | backend
7125 yes | yes | yes | no
7126 Arguments :
7127 <rate> The <rate> parameter is an integer designating the maximum number
7128 of new sessions per second to accept on the frontend.
7129
7130 When the frontend reaches the specified number of new sessions per second, it
7131 stops accepting new connections until the rate drops below the limit again.
7132 During this time, the pending sessions will be kept in the socket's backlog
7133 (in system buffers) and haproxy will not even be aware that sessions are
7134 pending. When applying very low limit on a highly loaded service, it may make
7135 sense to increase the socket's backlog using the "backlog" keyword.
7136
7137 This feature is particularly efficient at blocking connection-based attacks
7138 or service abuse on fragile servers. Since the session rate is measured every
7139 millisecond, it is extremely accurate. Also, the limit applies immediately,
7140 no delay is needed at all to detect the threshold.
7141
7142 Example : limit the connection rate on SMTP to 10 per second max
7143 listen smtp
7144 mode tcp
7145 bind :25
7146 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02007147 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007148
Willy Tarreaua17c2d92011-07-25 08:16:20 +02007149 Note : when the maximum rate is reached, the frontend's status is not changed
7150 but its sockets appear as "WAITING" in the statistics if the
7151 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007152
7153 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
7154
7155
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007156redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
7157redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
7158redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007159 Return an HTTP redirection if/unless a condition is matched
7160 May be used in sections : defaults | frontend | listen | backend
7161 no | yes | yes | yes
7162
7163 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01007164 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007165
Willy Tarreau0140f252008-11-19 21:07:09 +01007166 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007167 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007168 the HTTP "Location" header. When used in an "http-request" rule,
7169 <loc> value follows the log-format rules and can include some
7170 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007171
7172 <pfx> With "redirect prefix", the "Location" header is built from the
7173 concatenation of <pfx> and the complete URI path, including the
7174 query string, unless the "drop-query" option is specified (see
7175 below). As a special case, if <pfx> equals exactly "/", then
7176 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007177 redirect to the same URL (for instance, to insert a cookie). When
7178 used in an "http-request" rule, <pfx> value follows the log-format
7179 rules and can include some dynamic values (see Custom Log Format
7180 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007181
7182 <sch> With "redirect scheme", then the "Location" header is built by
7183 concatenating <sch> with "://" then the first occurrence of the
7184 "Host" header, and then the URI path, including the query string
7185 unless the "drop-query" option is specified (see below). If no
7186 path is found or if the path is "*", then "/" is used instead. If
7187 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007188 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007189 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007190 HTTPS. When used in an "http-request" rule, <sch> value follows
7191 the log-format rules and can include some dynamic values (see
7192 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01007193
7194 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007195 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
7196 with 302 used by default if no code is specified. 301 means
7197 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02007198 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007199 cache the redirection. 303 is equivalent to 302 except that the
7200 browser will fetch the location with a GET method. 307 is just
7201 like 302 but makes it clear that the same method must be reused.
7202 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01007203
7204 <option> There are several options which can be specified to adjust the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007205 expected behavior of a redirection :
Willy Tarreau0140f252008-11-19 21:07:09 +01007206
7207 - "drop-query"
7208 When this keyword is used in a prefix-based redirection, then the
7209 location will be set without any possible query-string, which is useful
7210 for directing users to a non-secure page for instance. It has no effect
7211 with a location-type redirect.
7212
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007213 - "append-slash"
7214 This keyword may be used in conjunction with "drop-query" to redirect
7215 users who use a URL not ending with a '/' to the same one with the '/'.
7216 It can be useful to ensure that search engines will only see one URL.
7217 For this, a return code 301 is preferred.
7218
Willy Tarreau0140f252008-11-19 21:07:09 +01007219 - "set-cookie NAME[=value]"
7220 A "Set-Cookie" header will be added with NAME (and optionally "=value")
7221 to the response. This is sometimes used to indicate that a user has
7222 been seen, for instance to protect against some types of DoS. No other
7223 cookie option is added, so the cookie will be a session cookie. Note
7224 that for a browser, a sole cookie name without an equal sign is
7225 different from a cookie with an equal sign.
7226
7227 - "clear-cookie NAME[=]"
7228 A "Set-Cookie" header will be added with NAME (and optionally "="), but
7229 with the "Max-Age" attribute set to zero. This will tell the browser to
7230 delete this cookie. It is useful for instance on logout pages. It is
7231 important to note that clearing the cookie "NAME" will not remove a
7232 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
7233 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007234
7235 Example: move the login URL only to HTTPS.
7236 acl clear dst_port 80
7237 acl secure dst_port 8080
7238 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01007239 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01007240 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01007241 acl cookie_set hdr_sub(cookie) SEEN=1
7242
7243 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01007244 redirect prefix https://mysite.com if login_page !secure
7245 redirect prefix http://mysite.com drop-query if login_page !uid_given
7246 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01007247 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007248
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007249 Example: send redirects for request for articles without a '/'.
7250 acl missing_slash path_reg ^/article/[^/]*$
7251 redirect code 301 prefix / drop-query append-slash if missing_slash
7252
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007253 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01007254 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007255
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007256 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02007257 http-request redirect code 301 location \
7258 http://www.%[hdr(host)]%[capture.req.uri] \
7259 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007260
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007261 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007262
7263
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007264redisp (deprecated)
7265redispatch (deprecated)
7266 Enable or disable session redistribution in case of connection failure
7267 May be used in sections: defaults | frontend | listen | backend
7268 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007269 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007270
7271 In HTTP mode, if a server designated by a cookie is down, clients may
7272 definitely stick to it because they cannot flush the cookie, so they will not
7273 be able to access the service anymore.
7274
7275 Specifying "redispatch" will allow the proxy to break their persistence and
7276 redistribute them to a working server.
7277
7278 It also allows to retry last connection to another server in case of multiple
7279 connection failures. Of course, it requires having "retries" set to a nonzero
7280 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007281
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007282 This form is deprecated, do not use it in any new configuration, use the new
7283 "option redispatch" instead.
7284
7285 See also : "option redispatch"
7286
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007287
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01007288reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01007289 Add a header at the end of the HTTP request
7290 May be used in sections : defaults | frontend | listen | backend
7291 no | yes | yes | yes
7292 Arguments :
7293 <string> is the complete line to be added. Any space or known delimiter
7294 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007295 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007296
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01007297 <cond> is an optional matching condition built from ACLs. It makes it
7298 possible to ignore this rule when other conditions are not met.
7299
Willy Tarreau303c0352008-01-17 19:01:39 +01007300 A new line consisting in <string> followed by a line feed will be added after
7301 the last header of an HTTP request.
7302
7303 Header transformations only apply to traffic which passes through HAProxy,
7304 and not to traffic generated by HAProxy, such as health-checks or error
7305 responses.
7306
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01007307 Example : add "X-Proto: SSL" to requests coming via port 81
7308 acl is-ssl dst_port 81
7309 reqadd X-Proto:\ SSL if is-ssl
7310
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007311 See also: "rspadd", "http-request", section 6 about HTTP header manipulation,
7312 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007313
7314
Willy Tarreau5321c422010-01-28 20:35:13 +01007315reqallow <search> [{if | unless} <cond>]
7316reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007317 Definitely allow an HTTP request if a line matches a regular expression
7318 May be used in sections : defaults | frontend | listen | backend
7319 no | yes | yes | yes
7320 Arguments :
7321 <search> is the regular expression applied to HTTP headers and to the
7322 request line. This is an extended regular expression. Parenthesis
7323 grouping is supported and no preliminary backslash is required.
7324 Any space or known delimiter must be escaped using a backslash
7325 ('\'). The pattern applies to a full line at a time. The
7326 "reqallow" keyword strictly matches case while "reqiallow"
7327 ignores case.
7328
Willy Tarreau5321c422010-01-28 20:35:13 +01007329 <cond> is an optional matching condition built from ACLs. It makes it
7330 possible to ignore this rule when other conditions are not met.
7331
Willy Tarreau303c0352008-01-17 19:01:39 +01007332 A request containing any line which matches extended regular expression
7333 <search> will mark the request as allowed, even if any later test would
7334 result in a deny. The test applies both to the request line and to request
7335 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007336 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007337
7338 It is easier, faster and more powerful to use ACLs to write access policies.
7339 Reqdeny, reqallow and reqpass should be avoided in new designs.
7340
7341 Example :
7342 # allow www.* but refuse *.local
7343 reqiallow ^Host:\ www\.
7344 reqideny ^Host:\ .*\.local
7345
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007346 See also: "reqdeny", "block", "http-request", section 6 about HTTP header
7347 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007348
7349
Willy Tarreau5321c422010-01-28 20:35:13 +01007350reqdel <search> [{if | unless} <cond>]
7351reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007352 Delete all headers matching a regular expression in an HTTP request
7353 May be used in sections : defaults | frontend | listen | backend
7354 no | yes | yes | yes
7355 Arguments :
7356 <search> is the regular expression applied to HTTP headers and to the
7357 request line. This is an extended regular expression. Parenthesis
7358 grouping is supported and no preliminary backslash is required.
7359 Any space or known delimiter must be escaped using a backslash
7360 ('\'). The pattern applies to a full line at a time. The "reqdel"
7361 keyword strictly matches case while "reqidel" ignores case.
7362
Willy Tarreau5321c422010-01-28 20:35:13 +01007363 <cond> is an optional matching condition built from ACLs. It makes it
7364 possible to ignore this rule when other conditions are not met.
7365
Willy Tarreau303c0352008-01-17 19:01:39 +01007366 Any header line matching extended regular expression <search> in the request
7367 will be completely deleted. Most common use of this is to remove unwanted
7368 and/or dangerous headers or cookies from a request before passing it to the
7369 next servers.
7370
7371 Header transformations only apply to traffic which passes through HAProxy,
7372 and not to traffic generated by HAProxy, such as health-checks or error
7373 responses. Keep in mind that header names are not case-sensitive.
7374
7375 Example :
7376 # remove X-Forwarded-For header and SERVER cookie
7377 reqidel ^X-Forwarded-For:.*
7378 reqidel ^Cookie:.*SERVER=
7379
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007380 See also: "reqadd", "reqrep", "rspdel", "http-request", section 6 about
7381 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007382
7383
Willy Tarreau5321c422010-01-28 20:35:13 +01007384reqdeny <search> [{if | unless} <cond>]
7385reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007386 Deny an HTTP request if a line matches a regular expression
7387 May be used in sections : defaults | frontend | listen | backend
7388 no | yes | yes | yes
7389 Arguments :
7390 <search> is the regular expression applied to HTTP headers and to the
7391 request line. This is an extended regular expression. Parenthesis
7392 grouping is supported and no preliminary backslash is required.
7393 Any space or known delimiter must be escaped using a backslash
7394 ('\'). The pattern applies to a full line at a time. The
7395 "reqdeny" keyword strictly matches case while "reqideny" ignores
7396 case.
7397
Willy Tarreau5321c422010-01-28 20:35:13 +01007398 <cond> is an optional matching condition built from ACLs. It makes it
7399 possible to ignore this rule when other conditions are not met.
7400
Willy Tarreau303c0352008-01-17 19:01:39 +01007401 A request containing any line which matches extended regular expression
7402 <search> will mark the request as denied, even if any later test would
7403 result in an allow. The test applies both to the request line and to request
7404 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007405 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007406
Willy Tarreauced27012008-01-17 20:35:34 +01007407 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007408 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01007409 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01007410
Willy Tarreau303c0352008-01-17 19:01:39 +01007411 It is easier, faster and more powerful to use ACLs to write access policies.
7412 Reqdeny, reqallow and reqpass should be avoided in new designs.
7413
7414 Example :
7415 # refuse *.local, then allow www.*
7416 reqideny ^Host:\ .*\.local
7417 reqiallow ^Host:\ www\.
7418
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007419 See also: "reqallow", "rspdeny", "block", "http-request", section 6 about
7420 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007421
7422
Willy Tarreau5321c422010-01-28 20:35:13 +01007423reqpass <search> [{if | unless} <cond>]
7424reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007425 Ignore any HTTP request line matching a regular expression in next rules
7426 May be used in sections : defaults | frontend | listen | backend
7427 no | yes | yes | yes
7428 Arguments :
7429 <search> is the regular expression applied to HTTP headers and to the
7430 request line. This is an extended regular expression. Parenthesis
7431 grouping is supported and no preliminary backslash is required.
7432 Any space or known delimiter must be escaped using a backslash
7433 ('\'). The pattern applies to a full line at a time. The
7434 "reqpass" keyword strictly matches case while "reqipass" ignores
7435 case.
7436
Willy Tarreau5321c422010-01-28 20:35:13 +01007437 <cond> is an optional matching condition built from ACLs. It makes it
7438 possible to ignore this rule when other conditions are not met.
7439
Willy Tarreau303c0352008-01-17 19:01:39 +01007440 A request containing any line which matches extended regular expression
7441 <search> will skip next rules, without assigning any deny or allow verdict.
7442 The test applies both to the request line and to request headers. Keep in
7443 mind that URLs in request line are case-sensitive while header names are not.
7444
7445 It is easier, faster and more powerful to use ACLs to write access policies.
7446 Reqdeny, reqallow and reqpass should be avoided in new designs.
7447
7448 Example :
7449 # refuse *.local, then allow www.*, but ignore "www.private.local"
7450 reqipass ^Host:\ www.private\.local
7451 reqideny ^Host:\ .*\.local
7452 reqiallow ^Host:\ www\.
7453
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007454 See also: "reqallow", "reqdeny", "block", "http-request", section 6 about
7455 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007456
7457
Willy Tarreau5321c422010-01-28 20:35:13 +01007458reqrep <search> <string> [{if | unless} <cond>]
7459reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007460 Replace a regular expression with a string in an HTTP request line
7461 May be used in sections : defaults | frontend | listen | backend
7462 no | yes | yes | yes
7463 Arguments :
7464 <search> is the regular expression applied to HTTP headers and to the
7465 request line. This is an extended regular expression. Parenthesis
7466 grouping is supported and no preliminary backslash is required.
7467 Any space or known delimiter must be escaped using a backslash
7468 ('\'). The pattern applies to a full line at a time. The "reqrep"
7469 keyword strictly matches case while "reqirep" ignores case.
7470
7471 <string> is the complete line to be added. Any space or known delimiter
7472 must be escaped using a backslash ('\'). References to matched
7473 pattern groups are possible using the common \N form, with N
7474 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007475 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007476
Willy Tarreau5321c422010-01-28 20:35:13 +01007477 <cond> is an optional matching condition built from ACLs. It makes it
7478 possible to ignore this rule when other conditions are not met.
7479
Willy Tarreau303c0352008-01-17 19:01:39 +01007480 Any line matching extended regular expression <search> in the request (both
7481 the request line and header lines) will be completely replaced with <string>.
7482 Most common use of this is to rewrite URLs or domain names in "Host" headers.
7483
7484 Header transformations only apply to traffic which passes through HAProxy,
7485 and not to traffic generated by HAProxy, such as health-checks or error
7486 responses. Note that for increased readability, it is suggested to add enough
7487 spaces between the request and the response. Keep in mind that URLs in
7488 request line are case-sensitive while header names are not.
7489
7490 Example :
7491 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007492 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01007493 # replace "www.mydomain.com" with "www" in the host name.
7494 reqirep ^Host:\ www.mydomain.com Host:\ www
7495
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007496 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", "http-request",
7497 section 6 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007498
7499
Willy Tarreau5321c422010-01-28 20:35:13 +01007500reqtarpit <search> [{if | unless} <cond>]
7501reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007502 Tarpit an HTTP request containing a line matching a regular expression
7503 May be used in sections : defaults | frontend | listen | backend
7504 no | yes | yes | yes
7505 Arguments :
7506 <search> is the regular expression applied to HTTP headers and to the
7507 request line. This is an extended regular expression. Parenthesis
7508 grouping is supported and no preliminary backslash is required.
7509 Any space or known delimiter must be escaped using a backslash
7510 ('\'). The pattern applies to a full line at a time. The
7511 "reqtarpit" keyword strictly matches case while "reqitarpit"
7512 ignores case.
7513
Willy Tarreau5321c422010-01-28 20:35:13 +01007514 <cond> is an optional matching condition built from ACLs. It makes it
7515 possible to ignore this rule when other conditions are not met.
7516
Willy Tarreau303c0352008-01-17 19:01:39 +01007517 A request containing any line which matches extended regular expression
7518 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01007519 be kept open for a pre-defined time, then will return an HTTP error 500 so
7520 that the attacker does not suspect it has been tarpitted. The status 500 will
7521 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01007522 delay is defined by "timeout tarpit", or "timeout connect" if the former is
7523 not set.
7524
7525 The goal of the tarpit is to slow down robots attacking servers with
7526 identifiable requests. Many robots limit their outgoing number of connections
7527 and stay connected waiting for a reply which can take several minutes to
7528 come. Depending on the environment and attack, it may be particularly
7529 efficient at reducing the load on the network and firewalls.
7530
Willy Tarreau5321c422010-01-28 20:35:13 +01007531 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01007532 # ignore user-agents reporting any flavor of "Mozilla" or "MSIE", but
Willy Tarreau303c0352008-01-17 19:01:39 +01007533 # block all others.
7534 reqipass ^User-Agent:\.*(Mozilla|MSIE)
7535 reqitarpit ^User-Agent:
7536
Willy Tarreau5321c422010-01-28 20:35:13 +01007537 # block bad guys
7538 acl badguys src 10.1.0.3 172.16.13.20/28
7539 reqitarpit . if badguys
7540
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007541 See also: "reqallow", "reqdeny", "reqpass", "http-request", section 6
7542 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007543
7544
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007545retries <value>
7546 Set the number of retries to perform on a server after a connection failure
7547 May be used in sections: defaults | frontend | listen | backend
7548 yes | no | yes | yes
7549 Arguments :
7550 <value> is the number of times a connection attempt should be retried on
7551 a server when a connection either is refused or times out. The
7552 default value is 3.
7553
7554 It is important to understand that this value applies to the number of
7555 connection attempts, not full requests. When a connection has effectively
7556 been established to a server, there will be no more retry.
7557
7558 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07007559 a turn-around timer of min("timeout connect", one second) is applied before
7560 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007561
7562 When "option redispatch" is set, the last retry may be performed on another
7563 server even if a cookie references a different server.
7564
7565 See also : "option redispatch"
7566
7567
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007568rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01007569 Add a header at the end of the HTTP response
7570 May be used in sections : defaults | frontend | listen | backend
7571 no | yes | yes | yes
7572 Arguments :
7573 <string> is the complete line to be added. Any space or known delimiter
7574 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007575 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007576
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007577 <cond> is an optional matching condition built from ACLs. It makes it
7578 possible to ignore this rule when other conditions are not met.
7579
Willy Tarreau303c0352008-01-17 19:01:39 +01007580 A new line consisting in <string> followed by a line feed will be added after
7581 the last header of an HTTP response.
7582
7583 Header transformations only apply to traffic which passes through HAProxy,
7584 and not to traffic generated by HAProxy, such as health-checks or error
7585 responses.
7586
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007587 See also: "rspdel" "reqadd", "http-response", section 6 about HTTP header
7588 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007589
7590
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007591rspdel <search> [{if | unless} <cond>]
7592rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007593 Delete all headers matching a regular expression in an HTTP response
7594 May be used in sections : defaults | frontend | listen | backend
7595 no | yes | yes | yes
7596 Arguments :
7597 <search> is the regular expression applied to HTTP headers and to the
7598 response line. This is an extended regular expression, so
7599 parenthesis grouping is supported and no preliminary backslash
7600 is required. Any space or known delimiter must be escaped using
7601 a backslash ('\'). The pattern applies to a full line at a time.
7602 The "rspdel" keyword strictly matches case while "rspidel"
7603 ignores case.
7604
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007605 <cond> is an optional matching condition built from ACLs. It makes it
7606 possible to ignore this rule when other conditions are not met.
7607
Willy Tarreau303c0352008-01-17 19:01:39 +01007608 Any header line matching extended regular expression <search> in the response
7609 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02007610 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01007611 client.
7612
7613 Header transformations only apply to traffic which passes through HAProxy,
7614 and not to traffic generated by HAProxy, such as health-checks or error
7615 responses. Keep in mind that header names are not case-sensitive.
7616
7617 Example :
7618 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02007619 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01007620
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007621 See also: "rspadd", "rsprep", "reqdel", "http-response", section 6 about
7622 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007623
7624
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007625rspdeny <search> [{if | unless} <cond>]
7626rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007627 Block an HTTP response if a line matches a regular expression
7628 May be used in sections : defaults | frontend | listen | backend
7629 no | yes | yes | yes
7630 Arguments :
7631 <search> is the regular expression applied to HTTP headers and to the
7632 response line. This is an extended regular expression, so
7633 parenthesis grouping is supported and no preliminary backslash
7634 is required. Any space or known delimiter must be escaped using
7635 a backslash ('\'). The pattern applies to a full line at a time.
7636 The "rspdeny" keyword strictly matches case while "rspideny"
7637 ignores case.
7638
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007639 <cond> is an optional matching condition built from ACLs. It makes it
7640 possible to ignore this rule when other conditions are not met.
7641
Willy Tarreau303c0352008-01-17 19:01:39 +01007642 A response containing any line which matches extended regular expression
7643 <search> will mark the request as denied. The test applies both to the
7644 response line and to response headers. Keep in mind that header names are not
7645 case-sensitive.
7646
7647 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01007648 block the response before it reaches the client. If a response is denied, it
7649 will be replaced with an HTTP 502 error so that the client never retrieves
7650 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01007651
7652 It is easier, faster and more powerful to use ACLs to write access policies.
7653 Rspdeny should be avoided in new designs.
7654
7655 Example :
7656 # Ensure that no content type matching ms-word will leak
7657 rspideny ^Content-type:\.*/ms-word
7658
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007659 See also: "reqdeny", "acl", "block", "http-response", section 6 about
7660 HTTP header manipulation and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007661
7662
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007663rsprep <search> <string> [{if | unless} <cond>]
7664rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007665 Replace a regular expression with a string in an HTTP response line
7666 May be used in sections : defaults | frontend | listen | backend
7667 no | yes | yes | yes
7668 Arguments :
7669 <search> is the regular expression applied to HTTP headers and to the
7670 response line. This is an extended regular expression, so
7671 parenthesis grouping is supported and no preliminary backslash
7672 is required. Any space or known delimiter must be escaped using
7673 a backslash ('\'). The pattern applies to a full line at a time.
7674 The "rsprep" keyword strictly matches case while "rspirep"
7675 ignores case.
7676
7677 <string> is the complete line to be added. Any space or known delimiter
7678 must be escaped using a backslash ('\'). References to matched
7679 pattern groups are possible using the common \N form, with N
7680 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007681 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007682
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007683 <cond> is an optional matching condition built from ACLs. It makes it
7684 possible to ignore this rule when other conditions are not met.
7685
Willy Tarreau303c0352008-01-17 19:01:39 +01007686 Any line matching extended regular expression <search> in the response (both
7687 the response line and header lines) will be completely replaced with
7688 <string>. Most common use of this is to rewrite Location headers.
7689
7690 Header transformations only apply to traffic which passes through HAProxy,
7691 and not to traffic generated by HAProxy, such as health-checks or error
7692 responses. Note that for increased readability, it is suggested to add enough
7693 spaces between the request and the response. Keep in mind that header names
7694 are not case-sensitive.
7695
7696 Example :
7697 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
7698 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
7699
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007700 See also: "rspadd", "rspdel", "reqrep", "http-response", section 6 about
7701 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007702
7703
David du Colombier486df472011-03-17 10:40:26 +01007704server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007705 Declare a server in a backend
7706 May be used in sections : defaults | frontend | listen | backend
7707 no | no | yes | yes
7708 Arguments :
7709 <name> is the internal name assigned to this server. This name will
Davor Ocelice9ed2812017-12-25 17:49:28 +01007710 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05007711 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007712
David du Colombier486df472011-03-17 10:40:26 +01007713 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
7714 resolvable hostname is supported, but this name will be resolved
7715 during start-up. Address "0.0.0.0" or "*" has a special meaning.
7716 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02007717 address as the one from the client connection. This is useful in
7718 transparent proxy architectures where the client's connection is
7719 intercepted and haproxy must forward to the original destination
7720 address. This is more or less what the "transparent" keyword does
7721 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01007722 to report statistics. Optionally, an address family prefix may be
7723 used before the address to force the family regardless of the
7724 address format, which can be useful to specify a path to a unix
7725 socket with no slash ('/'). Currently supported prefixes are :
7726 - 'ipv4@' -> address is always IPv4
7727 - 'ipv6@' -> address is always IPv6
7728 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007729 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemandb2f07452015-05-12 14:27:13 +02007730 You may want to reference some environment variables in the
7731 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +01007732 variables. The "init-addr" setting can be used to modify the way
7733 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007734
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007735 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007736 be sent to this port. If unset, the same port the client
7737 connected to will be used. The port may also be prefixed by a "+"
7738 or a "-". In this case, the server's port will be determined by
7739 adding this value to the client's port.
7740
7741 <param*> is a list of parameters for this server. The "server" keywords
7742 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007743 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007744
7745 Examples :
7746 server first 10.1.1.1:1080 cookie first check inter 1000
7747 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01007748 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02007749 server backup "${SRV_BACKUP}:1080" backup
7750 server www1_dc1 "${LAN_DC1}.101:80"
7751 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007752
Willy Tarreau55dcaf62015-09-27 15:03:15 +02007753 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
7754 sun_path length is used for the address length. Some other programs
7755 such as socat use the string length only by default. Pass the option
7756 ",unix-tightsocklen=0" to any abstract socket definition in socat to
7757 make it compatible with HAProxy's.
7758
Mark Lamourinec2247f02012-01-04 13:02:01 -05007759 See also: "default-server", "http-send-name-header" and section 5 about
7760 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007761
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007762server-state-file-name [<file>]
7763 Set the server state file to read, load and apply to servers available in
7764 this backend. It only applies when the directive "load-server-state-from-file"
7765 is set to "local". When <file> is not provided or if this directive is not
7766 set, then backend name is used. If <file> starts with a slash '/', then it is
7767 considered as an absolute path. Otherwise, <file> is concatenated to the
7768 global directive "server-state-file-base".
7769
7770 Example: the minimal configuration below would make HAProxy look for the
7771 state server file '/etc/haproxy/states/bk':
7772
7773 global
7774 server-state-file-base /etc/haproxy/states
7775
7776 backend bk
7777 load-server-state-from-file
7778
7779 See also: "server-state-file-base", "load-server-state-from-file", and
7780 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007781
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02007782server-template <prefix> <num | range> <fqdn>[:<port>] [params*]
7783 Set a template to initialize servers with shared parameters.
7784 The names of these servers are built from <prefix> and <num | range> parameters.
7785 May be used in sections : defaults | frontend | listen | backend
7786 no | no | yes | yes
7787
7788 Arguments:
7789 <prefix> A prefix for the server names to be built.
7790
7791 <num | range>
7792 If <num> is provided, this template initializes <num> servers
7793 with 1 up to <num> as server name suffixes. A range of numbers
7794 <num_low>-<num_high> may also be used to use <num_low> up to
7795 <num_high> as server name suffixes.
7796
7797 <fqdn> A FQDN for all the servers this template initializes.
7798
7799 <port> Same meaning as "server" <port> argument (see "server" keyword).
7800
7801 <params*>
7802 Remaining server parameters among all those supported by "server"
7803 keyword.
7804
7805 Examples:
7806 # Initializes 3 servers with srv1, srv2 and srv3 as names,
7807 # google.com as FQDN, and health-check enabled.
7808 server-template srv 1-3 google.com:80 check
7809
7810 # or
7811 server-template srv 3 google.com:80 check
7812
7813 # would be equivalent to:
7814 server srv1 google.com:80 check
7815 server srv2 google.com:80 check
7816 server srv3 google.com:80 check
7817
7818
7819
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007820source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02007821source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007822source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007823 Set the source address for outgoing connections
7824 May be used in sections : defaults | frontend | listen | backend
7825 yes | no | yes | yes
7826 Arguments :
7827 <addr> is the IPv4 address HAProxy will bind to before connecting to a
7828 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01007829
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007830 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01007831 the most appropriate address to reach its destination. Optionally
7832 an address family prefix may be used before the address to force
7833 the family regardless of the address format, which can be useful
7834 to specify a path to a unix socket with no slash ('/'). Currently
7835 supported prefixes are :
7836 - 'ipv4@' -> address is always IPv4
7837 - 'ipv6@' -> address is always IPv6
7838 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007839 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02007840 You may want to reference some environment variables in the
7841 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007842
7843 <port> is an optional port. It is normally not needed but may be useful
7844 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02007845 the system will select a free port. Note that port ranges are not
7846 supported in the backend. If you want to force port ranges, you
7847 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007848
7849 <addr2> is the IP address to present to the server when connections are
7850 forwarded in full transparent proxy mode. This is currently only
7851 supported on some patched Linux kernels. When this address is
7852 specified, clients connecting to the server will be presented
7853 with this address, while health checks will still use the address
7854 <addr>.
7855
7856 <port2> is the optional port to present to the server when connections
7857 are forwarded in full transparent proxy mode (see <addr2> above).
7858 The default value of zero means the system will select a free
7859 port.
7860
Willy Tarreaubce70882009-09-07 11:51:47 +02007861 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
7862 This is the name of a comma-separated header list which can
7863 contain multiple IP addresses. By default, the last occurrence is
7864 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01007865 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02007866 by previous proxy, typically Stunnel. In order to use another
7867 occurrence from the last one, please see the <occ> parameter
7868 below. When the header (or occurrence) is not found, no binding
7869 is performed so that the proxy's default IP address is used. Also
7870 keep in mind that the header name is case insensitive, as for any
7871 HTTP header.
7872
7873 <occ> is the occurrence number of a value to be used in a multi-value
7874 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007875 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02007876 address. Positive values indicate a position from the first
7877 occurrence, 1 being the first one. Negative values indicate
7878 positions relative to the last one, -1 being the last one. This
7879 is helpful for situations where an X-Forwarded-For header is set
7880 at the entry point of an infrastructure and must be used several
7881 proxy layers away. When this value is not specified, -1 is
7882 assumed. Passing a zero here disables the feature.
7883
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007884 <name> is an optional interface name to which to bind to for outgoing
7885 traffic. On systems supporting this features (currently, only
7886 Linux), this allows one to bind all traffic to the server to
7887 this interface even if it is not the one the system would select
7888 based on routing tables. This should be used with extreme care.
7889 Note that using this option requires root privileges.
7890
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007891 The "source" keyword is useful in complex environments where a specific
7892 address only is allowed to connect to the servers. It may be needed when a
7893 private address must be used through a public gateway for instance, and it is
7894 known that the system cannot determine the adequate source address by itself.
7895
7896 An extension which is available on certain patched Linux kernels may be used
7897 through the "usesrc" optional keyword. It makes it possible to connect to the
7898 servers with an IP address which does not belong to the system itself. This
7899 is called "full transparent proxy mode". For this to work, the destination
7900 servers have to route their traffic back to this address through the machine
7901 running HAProxy, and IP forwarding must generally be enabled on this machine.
7902
7903 In this "full transparent proxy" mode, it is possible to force a specific IP
7904 address to be presented to the servers. This is not much used in fact. A more
7905 common use is to tell HAProxy to present the client's IP address. For this,
7906 there are two methods :
7907
7908 - present the client's IP and port addresses. This is the most transparent
7909 mode, but it can cause problems when IP connection tracking is enabled on
7910 the machine, because a same connection may be seen twice with different
7911 states. However, this solution presents the huge advantage of not
7912 limiting the system to the 64k outgoing address+port couples, because all
7913 of the client ranges may be used.
7914
7915 - present only the client's IP address and select a spare port. This
7916 solution is still quite elegant but slightly less transparent (downstream
7917 firewalls logs will not match upstream's). It also presents the downside
7918 of limiting the number of concurrent connections to the usual 64k ports.
7919 However, since the upstream and downstream ports are different, local IP
7920 connection tracking on the machine will not be upset by the reuse of the
7921 same session.
7922
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007923 This option sets the default source for all servers in the backend. It may
7924 also be specified in a "defaults" section. Finer source address specification
7925 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007926 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007927
Baptiste Assmann91bd3372015-07-17 21:59:42 +02007928 In order to work, "usesrc" requires root privileges.
7929
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007930 Examples :
7931 backend private
7932 # Connect to the servers using our 192.168.1.200 source address
7933 source 192.168.1.200
7934
7935 backend transparent_ssl1
7936 # Connect to the SSL farm from the client's source address
7937 source 192.168.1.200 usesrc clientip
7938
7939 backend transparent_ssl2
7940 # Connect to the SSL farm from the client's source address and port
7941 # not recommended if IP conntrack is present on the local machine.
7942 source 192.168.1.200 usesrc client
7943
7944 backend transparent_ssl3
7945 # Connect to the SSL farm from the client's source address. It
7946 # is more conntrack-friendly.
7947 source 192.168.1.200 usesrc clientip
7948
7949 backend transparent_smtp
7950 # Connect to the SMTP farm from the client's source address/port
7951 # with Tproxy version 4.
7952 source 0.0.0.0 usesrc clientip
7953
Willy Tarreaubce70882009-09-07 11:51:47 +02007954 backend transparent_http
7955 # Connect to the servers using the client's IP as seen by previous
7956 # proxy.
7957 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
7958
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007959 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007960 the Linux kernel on www.balabit.com, the "bind" keyword.
7961
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007962
Willy Tarreau844e3c52008-01-11 16:28:18 +01007963srvtimeout <timeout> (deprecated)
7964 Set the maximum inactivity time on the server side.
7965 May be used in sections : defaults | frontend | listen | backend
7966 yes | no | yes | yes
7967 Arguments :
7968 <timeout> is the timeout value specified in milliseconds by default, but
7969 can be in any other unit if the number is suffixed by the unit,
7970 as explained at the top of this document.
7971
7972 The inactivity timeout applies when the server is expected to acknowledge or
7973 send data. In HTTP mode, this timeout is particularly important to consider
7974 during the first phase of the server's response, when it has to send the
7975 headers, as it directly represents the server's processing time for the
7976 request. To find out what value to put there, it's often good to start with
7977 what would be considered as unacceptable response times, then check the logs
7978 to observe the response time distribution, and adjust the value accordingly.
7979
7980 The value is specified in milliseconds by default, but can be in any other
7981 unit if the number is suffixed by the unit, as specified at the top of this
7982 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
7983 recommended that the client timeout remains equal to the server timeout in
7984 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007985 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01007986 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +01007987 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01007988
7989 This parameter is specific to backends, but can be specified once for all in
7990 "defaults" sections. This is in fact one of the easiest solutions not to
7991 forget about it. An unspecified timeout results in an infinite timeout, which
7992 is not recommended. Such a usage is accepted and works but reports a warning
7993 during startup because it may results in accumulation of expired sessions in
7994 the system if the system's timeouts are not configured either.
7995
7996 This parameter is provided for compatibility but is currently deprecated.
7997 Please use "timeout server" instead.
7998
Willy Tarreauce887fd2012-05-12 12:50:00 +02007999 See also : "timeout server", "timeout tunnel", "timeout client" and
8000 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01008001
8002
Cyril Bonté66c327d2010-10-12 00:14:37 +02008003stats admin { if | unless } <cond>
8004 Enable statistics admin level if/unless a condition is matched
8005 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008006 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02008007
8008 This statement enables the statistics admin level if/unless a condition is
8009 matched.
8010
8011 The admin level allows to enable/disable servers from the web interface. By
8012 default, statistics page is read-only for security reasons.
8013
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008014 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8015 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008016 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008017
Cyril Bonté23b39d92011-02-10 22:54:44 +01008018 Currently, the POST request is limited to the buffer size minus the reserved
8019 buffer space, which means that if the list of servers is too long, the
8020 request won't be processed. It is recommended to alter few servers at a
8021 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008022
8023 Example :
8024 # statistics admin level only for localhost
8025 backend stats_localhost
8026 stats enable
8027 stats admin if LOCALHOST
8028
8029 Example :
8030 # statistics admin level always enabled because of the authentication
8031 backend stats_auth
8032 stats enable
8033 stats auth admin:AdMiN123
8034 stats admin if TRUE
8035
8036 Example :
8037 # statistics admin level depends on the authenticated user
8038 userlist stats-auth
8039 group admin users admin
8040 user admin insecure-password AdMiN123
8041 group readonly users haproxy
8042 user haproxy insecure-password haproxy
8043
8044 backend stats_auth
8045 stats enable
8046 acl AUTH http_auth(stats-auth)
8047 acl AUTH_ADMIN http_auth_group(stats-auth) admin
8048 stats http-request auth unless AUTH
8049 stats admin if AUTH_ADMIN
8050
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008051 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
8052 "bind-process", section 3.4 about userlists and section 7 about
8053 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008054
8055
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008056stats auth <user>:<passwd>
8057 Enable statistics with authentication and grant access to an account
8058 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008059 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008060 Arguments :
8061 <user> is a user name to grant access to
8062
8063 <passwd> is the cleartext password associated to this user
8064
8065 This statement enables statistics with default settings, and restricts access
8066 to declared users only. It may be repeated as many times as necessary to
8067 allow as many users as desired. When a user tries to access the statistics
8068 without a valid account, a "401 Forbidden" response will be returned so that
8069 the browser asks the user to provide a valid user and password. The real
8070 which will be returned to the browser is configurable using "stats realm".
8071
8072 Since the authentication method is HTTP Basic Authentication, the passwords
8073 circulate in cleartext on the network. Thus, it was decided that the
8074 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008075 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008076
8077 It is also possible to reduce the scope of the proxies which appear in the
8078 report using "stats scope".
8079
8080 Though this statement alone is enough to enable statistics reporting, it is
8081 recommended to set all other settings in order to avoid relying on default
8082 unobvious parameters.
8083
8084 Example :
8085 # public access (limited to this backend only)
8086 backend public_www
8087 server srv1 192.168.0.1:80
8088 stats enable
8089 stats hide-version
8090 stats scope .
8091 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008092 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008093 stats auth admin1:AdMiN123
8094 stats auth admin2:AdMiN321
8095
8096 # internal monitoring access (unlimited)
8097 backend private_monitoring
8098 stats enable
8099 stats uri /admin?stats
8100 stats refresh 5s
8101
8102 See also : "stats enable", "stats realm", "stats scope", "stats uri"
8103
8104
8105stats enable
8106 Enable statistics reporting with default settings
8107 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008108 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008109 Arguments : none
8110
8111 This statement enables statistics reporting with default settings defined
8112 at build time. Unless stated otherwise, these settings are used :
8113 - stats uri : /haproxy?stats
8114 - stats realm : "HAProxy Statistics"
8115 - stats auth : no authentication
8116 - stats scope : no restriction
8117
8118 Though this statement alone is enough to enable statistics reporting, it is
8119 recommended to set all other settings in order to avoid relying on default
8120 unobvious parameters.
8121
8122 Example :
8123 # public access (limited to this backend only)
8124 backend public_www
8125 server srv1 192.168.0.1:80
8126 stats enable
8127 stats hide-version
8128 stats scope .
8129 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008130 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008131 stats auth admin1:AdMiN123
8132 stats auth admin2:AdMiN321
8133
8134 # internal monitoring access (unlimited)
8135 backend private_monitoring
8136 stats enable
8137 stats uri /admin?stats
8138 stats refresh 5s
8139
8140 See also : "stats auth", "stats realm", "stats uri"
8141
8142
Willy Tarreaud63335a2010-02-26 12:56:52 +01008143stats hide-version
8144 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008145 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008146 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008147 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008148
Willy Tarreaud63335a2010-02-26 12:56:52 +01008149 By default, the stats page reports some useful status information along with
8150 the statistics. Among them is HAProxy's version. However, it is generally
8151 considered dangerous to report precise version to anyone, as it can help them
8152 target known weaknesses with specific attacks. The "stats hide-version"
8153 statement removes the version from the statistics report. This is recommended
8154 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008155
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008156 Though this statement alone is enough to enable statistics reporting, it is
8157 recommended to set all other settings in order to avoid relying on default
8158 unobvious parameters.
8159
Willy Tarreaud63335a2010-02-26 12:56:52 +01008160 Example :
8161 # public access (limited to this backend only)
8162 backend public_www
8163 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008164 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008165 stats hide-version
8166 stats scope .
8167 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008168 stats realm HAProxy\ Statistics
Willy Tarreaud63335a2010-02-26 12:56:52 +01008169 stats auth admin1:AdMiN123
8170 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008171
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008172 # internal monitoring access (unlimited)
8173 backend private_monitoring
8174 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008175 stats uri /admin?stats
8176 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01008177
Willy Tarreaud63335a2010-02-26 12:56:52 +01008178 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008179
Willy Tarreau983e01e2010-01-11 18:42:06 +01008180
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02008181stats http-request { allow | deny | auth [realm <realm>] }
8182 [ { if | unless } <condition> ]
8183 Access control for statistics
8184
8185 May be used in sections: defaults | frontend | listen | backend
8186 no | no | yes | yes
8187
8188 As "http-request", these set of options allow to fine control access to
8189 statistics. Each option may be followed by if/unless and acl.
8190 First option with matched condition (or option without condition) is final.
8191 For "deny" a 403 error will be returned, for "allow" normal processing is
8192 performed, for "auth" a 401/407 error code is returned so the client
8193 should be asked to enter a username and password.
8194
8195 There is no fixed limit to the number of http-request statements per
8196 instance.
8197
8198 See also : "http-request", section 3.4 about userlists and section 7
8199 about ACL usage.
8200
8201
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008202stats realm <realm>
8203 Enable statistics and set authentication realm
8204 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008205 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008206 Arguments :
8207 <realm> is the name of the HTTP Basic Authentication realm reported to
8208 the browser. The browser uses it to display it in the pop-up
8209 inviting the user to enter a valid username and password.
8210
8211 The realm is read as a single word, so any spaces in it should be escaped
8212 using a backslash ('\').
8213
8214 This statement is useful only in conjunction with "stats auth" since it is
8215 only related to authentication.
8216
8217 Though this statement alone is enough to enable statistics reporting, it is
8218 recommended to set all other settings in order to avoid relying on default
8219 unobvious parameters.
8220
8221 Example :
8222 # public access (limited to this backend only)
8223 backend public_www
8224 server srv1 192.168.0.1:80
8225 stats enable
8226 stats hide-version
8227 stats scope .
8228 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008229 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008230 stats auth admin1:AdMiN123
8231 stats auth admin2:AdMiN321
8232
8233 # internal monitoring access (unlimited)
8234 backend private_monitoring
8235 stats enable
8236 stats uri /admin?stats
8237 stats refresh 5s
8238
8239 See also : "stats auth", "stats enable", "stats uri"
8240
8241
8242stats refresh <delay>
8243 Enable statistics with automatic refresh
8244 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008245 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008246 Arguments :
8247 <delay> is the suggested refresh delay, specified in seconds, which will
8248 be returned to the browser consulting the report page. While the
8249 browser is free to apply any delay, it will generally respect it
8250 and refresh the page this every seconds. The refresh interval may
8251 be specified in any other non-default time unit, by suffixing the
8252 unit after the value, as explained at the top of this document.
8253
8254 This statement is useful on monitoring displays with a permanent page
8255 reporting the load balancer's activity. When set, the HTML report page will
8256 include a link "refresh"/"stop refresh" so that the user can select whether
8257 he wants automatic refresh of the page or not.
8258
8259 Though this statement alone is enough to enable statistics reporting, it is
8260 recommended to set all other settings in order to avoid relying on default
8261 unobvious parameters.
8262
8263 Example :
8264 # public access (limited to this backend only)
8265 backend public_www
8266 server srv1 192.168.0.1:80
8267 stats enable
8268 stats hide-version
8269 stats scope .
8270 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008271 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008272 stats auth admin1:AdMiN123
8273 stats auth admin2:AdMiN321
8274
8275 # internal monitoring access (unlimited)
8276 backend private_monitoring
8277 stats enable
8278 stats uri /admin?stats
8279 stats refresh 5s
8280
8281 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8282
8283
8284stats scope { <name> | "." }
8285 Enable statistics and limit access scope
8286 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008287 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008288 Arguments :
8289 <name> is the name of a listen, frontend or backend section to be
8290 reported. The special name "." (a single dot) designates the
8291 section in which the statement appears.
8292
8293 When this statement is specified, only the sections enumerated with this
8294 statement will appear in the report. All other ones will be hidden. This
8295 statement may appear as many times as needed if multiple sections need to be
8296 reported. Please note that the name checking is performed as simple string
8297 comparisons, and that it is never checked that a give section name really
8298 exists.
8299
8300 Though this statement alone is enough to enable statistics reporting, it is
8301 recommended to set all other settings in order to avoid relying on default
8302 unobvious parameters.
8303
8304 Example :
8305 # public access (limited to this backend only)
8306 backend public_www
8307 server srv1 192.168.0.1:80
8308 stats enable
8309 stats hide-version
8310 stats scope .
8311 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008312 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008313 stats auth admin1:AdMiN123
8314 stats auth admin2:AdMiN321
8315
8316 # internal monitoring access (unlimited)
8317 backend private_monitoring
8318 stats enable
8319 stats uri /admin?stats
8320 stats refresh 5s
8321
8322 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8323
Willy Tarreaud63335a2010-02-26 12:56:52 +01008324
Willy Tarreauc9705a12010-07-27 20:05:50 +02008325stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01008326 Enable reporting of a description on the statistics page.
8327 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008328 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008329
Willy Tarreauc9705a12010-07-27 20:05:50 +02008330 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01008331 description from global section is automatically used instead.
8332
8333 This statement is useful for users that offer shared services to their
8334 customers, where node or description should be different for each customer.
8335
8336 Though this statement alone is enough to enable statistics reporting, it is
8337 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008338 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008339
8340 Example :
8341 # internal monitoring access (unlimited)
8342 backend private_monitoring
8343 stats enable
8344 stats show-desc Master node for Europe, Asia, Africa
8345 stats uri /admin?stats
8346 stats refresh 5s
8347
8348 See also: "show-node", "stats enable", "stats uri" and "description" in
8349 global section.
8350
8351
8352stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02008353 Enable reporting additional information on the statistics page
8354 May be used in sections : defaults | frontend | listen | backend
8355 yes | yes | yes | yes
8356 Arguments : none
8357
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008358 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01008359 - cap: capabilities (proxy)
8360 - mode: one of tcp, http or health (proxy)
8361 - id: SNMP ID (proxy, socket, server)
8362 - IP (socket, server)
8363 - cookie (backend, server)
8364
8365 Though this statement alone is enough to enable statistics reporting, it is
8366 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008367 unobvious parameters. Default behavior is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008368
8369 See also: "stats enable", "stats uri".
8370
8371
8372stats show-node [ <name> ]
8373 Enable reporting of a host name on the statistics page.
8374 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008375 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008376 Arguments:
8377 <name> is an optional name to be reported. If unspecified, the
8378 node name from global section is automatically used instead.
8379
8380 This statement is useful for users that offer shared services to their
8381 customers, where node or description might be different on a stats page
Davor Ocelice9ed2812017-12-25 17:49:28 +01008382 provided for each customer. Default behavior is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008383
8384 Though this statement alone is enough to enable statistics reporting, it is
8385 recommended to set all other settings in order to avoid relying on default
8386 unobvious parameters.
8387
8388 Example:
8389 # internal monitoring access (unlimited)
8390 backend private_monitoring
8391 stats enable
8392 stats show-node Europe-1
8393 stats uri /admin?stats
8394 stats refresh 5s
8395
8396 See also: "show-desc", "stats enable", "stats uri", and "node" in global
8397 section.
8398
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008399
8400stats uri <prefix>
8401 Enable statistics and define the URI prefix to access them
8402 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008403 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008404 Arguments :
8405 <prefix> is the prefix of any URI which will be redirected to stats. This
8406 prefix may contain a question mark ('?') to indicate part of a
8407 query string.
8408
8409 The statistics URI is intercepted on the relayed traffic, so it appears as a
8410 page within the normal application. It is strongly advised to ensure that the
8411 selected URI will never appear in the application, otherwise it will never be
8412 possible to reach it in the application.
8413
8414 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008415 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008416 It is generally a good idea to include a question mark in the URI so that
8417 intermediate proxies refrain from caching the results. Also, since any string
8418 beginning with the prefix will be accepted as a stats request, the question
8419 mark helps ensuring that no valid URI will begin with the same words.
8420
8421 It is sometimes very convenient to use "/" as the URI prefix, and put that
8422 statement in a "listen" instance of its own. That makes it easy to dedicate
8423 an address or a port to statistics only.
8424
8425 Though this statement alone is enough to enable statistics reporting, it is
8426 recommended to set all other settings in order to avoid relying on default
8427 unobvious parameters.
8428
8429 Example :
8430 # public access (limited to this backend only)
8431 backend public_www
8432 server srv1 192.168.0.1:80
8433 stats enable
8434 stats hide-version
8435 stats scope .
8436 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008437 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008438 stats auth admin1:AdMiN123
8439 stats auth admin2:AdMiN321
8440
8441 # internal monitoring access (unlimited)
8442 backend private_monitoring
8443 stats enable
8444 stats uri /admin?stats
8445 stats refresh 5s
8446
8447 See also : "stats auth", "stats enable", "stats realm"
8448
8449
Willy Tarreaud63335a2010-02-26 12:56:52 +01008450stick match <pattern> [table <table>] [{if | unless} <cond>]
8451 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008452 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01008453 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008454
8455 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008456 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008457 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01008458 will be analyzed in the hope to find a matching entry in a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008459 stickiness table. This rule is mandatory.
8460
8461 <table> is an optional stickiness table name. If unspecified, the same
8462 backend's table is used. A stickiness table is declared using
8463 the "stick-table" statement.
8464
8465 <cond> is an optional matching condition. It makes it possible to match
8466 on a certain criterion only when other conditions are met (or
8467 not met). For instance, it could be used to match on a source IP
8468 address except when a request passes through a known proxy, in
8469 which case we'd match on a header containing that IP address.
8470
8471 Some protocols or applications require complex stickiness rules and cannot
8472 always simply rely on cookies nor hashing. The "stick match" statement
8473 describes a rule to extract the stickiness criterion from an incoming request
8474 or connection. See section 7 for a complete list of possible patterns and
8475 transformation rules.
8476
8477 The table has to be declared using the "stick-table" statement. It must be of
8478 a type compatible with the pattern. By default it is the one which is present
8479 in the same backend. It is possible to share a table with other backends by
8480 referencing it using the "table" keyword. If another table is referenced,
8481 the server's ID inside the backends are used. By default, all server IDs
8482 start at 1 in each backend, so the server ordering is enough. But in case of
8483 doubt, it is highly recommended to force server IDs using their "id" setting.
8484
8485 It is possible to restrict the conditions where a "stick match" statement
8486 will apply, using "if" or "unless" followed by a condition. See section 7 for
8487 ACL based conditions.
8488
8489 There is no limit on the number of "stick match" statements. The first that
8490 applies and matches will cause the request to be directed to the same server
8491 as was used for the request which created the entry. That way, multiple
8492 matches can be used as fallbacks.
8493
8494 The stick rules are checked after the persistence cookies, so they will not
8495 affect stickiness if a cookie has already been used to select a server. That
8496 way, it becomes very easy to insert cookies and match on IP addresses in
8497 order to maintain stickiness between HTTP and HTTPS.
8498
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008499 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8500 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008501 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008502
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008503 Example :
8504 # forward SMTP users to the same server they just used for POP in the
8505 # last 30 minutes
8506 backend pop
8507 mode tcp
8508 balance roundrobin
8509 stick store-request src
8510 stick-table type ip size 200k expire 30m
8511 server s1 192.168.1.1:110
8512 server s2 192.168.1.1:110
8513
8514 backend smtp
8515 mode tcp
8516 balance roundrobin
8517 stick match src table pop
8518 server s1 192.168.1.1:25
8519 server s2 192.168.1.1:25
8520
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008521 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02008522 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008523
8524
8525stick on <pattern> [table <table>] [{if | unless} <condition>]
8526 Define a request pattern to associate a user to a server
8527 May be used in sections : defaults | frontend | listen | backend
8528 no | no | yes | yes
8529
8530 Note : This form is exactly equivalent to "stick match" followed by
8531 "stick store-request", all with the same arguments. Please refer
8532 to both keywords for details. It is only provided as a convenience
8533 for writing more maintainable configurations.
8534
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008535 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8536 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008537 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008538
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008539 Examples :
8540 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01008541 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008542
8543 # ...is strictly equivalent to this one :
8544 stick match src table pop if !localhost
8545 stick store-request src table pop if !localhost
8546
8547
8548 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
8549 # well as HTTP without cookie. Share the same table between both accesses.
8550 backend http
8551 mode http
8552 balance roundrobin
8553 stick on src table https
8554 cookie SRV insert indirect nocache
8555 server s1 192.168.1.1:80 cookie s1
8556 server s2 192.168.1.1:80 cookie s2
8557
8558 backend https
8559 mode tcp
8560 balance roundrobin
8561 stick-table type ip size 200k expire 30m
8562 stick on src
8563 server s1 192.168.1.1:443
8564 server s2 192.168.1.1:443
8565
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008566 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008567
8568
8569stick store-request <pattern> [table <table>] [{if | unless} <condition>]
8570 Define a request pattern used to create an entry in a stickiness table
8571 May be used in sections : defaults | frontend | listen | backend
8572 no | no | yes | yes
8573
8574 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008575 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008576 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01008577 will be analyzed, extracted and stored in the table once a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008578 server is selected.
8579
8580 <table> is an optional stickiness table name. If unspecified, the same
8581 backend's table is used. A stickiness table is declared using
8582 the "stick-table" statement.
8583
8584 <cond> is an optional storage condition. It makes it possible to store
8585 certain criteria only when some conditions are met (or not met).
8586 For instance, it could be used to store the source IP address
8587 except when the request passes through a known proxy, in which
8588 case we'd store a converted form of a header containing that IP
8589 address.
8590
8591 Some protocols or applications require complex stickiness rules and cannot
8592 always simply rely on cookies nor hashing. The "stick store-request" statement
8593 describes a rule to decide what to extract from the request and when to do
8594 it, in order to store it into a stickiness table for further requests to
8595 match it using the "stick match" statement. Obviously the extracted part must
8596 make sense and have a chance to be matched in a further request. Storing a
8597 client's IP address for instance often makes sense. Storing an ID found in a
8598 URL parameter also makes sense. Storing a source port will almost never make
8599 any sense because it will be randomly matched. See section 7 for a complete
8600 list of possible patterns and transformation rules.
8601
8602 The table has to be declared using the "stick-table" statement. It must be of
8603 a type compatible with the pattern. By default it is the one which is present
8604 in the same backend. It is possible to share a table with other backends by
8605 referencing it using the "table" keyword. If another table is referenced,
8606 the server's ID inside the backends are used. By default, all server IDs
8607 start at 1 in each backend, so the server ordering is enough. But in case of
8608 doubt, it is highly recommended to force server IDs using their "id" setting.
8609
8610 It is possible to restrict the conditions where a "stick store-request"
8611 statement will apply, using "if" or "unless" followed by a condition. This
8612 condition will be evaluated while parsing the request, so any criteria can be
8613 used. See section 7 for ACL based conditions.
8614
8615 There is no limit on the number of "stick store-request" statements, but
8616 there is a limit of 8 simultaneous stores per request or response. This
8617 makes it possible to store up to 8 criteria, all extracted from either the
8618 request or the response, regardless of the number of rules. Only the 8 first
8619 ones which match will be kept. Using this, it is possible to feed multiple
8620 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008621 another protocol or access method. Using multiple store-request rules with
8622 the same table is possible and may be used to find the best criterion to rely
8623 on, by arranging the rules by decreasing preference order. Only the first
8624 extracted criterion for a given table will be stored. All subsequent store-
8625 request rules referencing the same table will be skipped and their ACLs will
8626 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008627
8628 The "store-request" rules are evaluated once the server connection has been
8629 established, so that the table will contain the real server that processed
8630 the request.
8631
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008632 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8633 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008634 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008635
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008636 Example :
8637 # forward SMTP users to the same server they just used for POP in the
8638 # last 30 minutes
8639 backend pop
8640 mode tcp
8641 balance roundrobin
8642 stick store-request src
8643 stick-table type ip size 200k expire 30m
8644 server s1 192.168.1.1:110
8645 server s2 192.168.1.1:110
8646
8647 backend smtp
8648 mode tcp
8649 balance roundrobin
8650 stick match src table pop
8651 server s1 192.168.1.1:25
8652 server s2 192.168.1.1:25
8653
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008654 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02008655 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008656
8657
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008658stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02008659 size <size> [expire <expire>] [nopurge] [peers <peersect>]
8660 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08008661 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008662 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008663 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008664
8665 Arguments :
8666 ip a table declared with "type ip" will only store IPv4 addresses.
8667 This form is very compact (about 50 bytes per entry) and allows
8668 very fast entry lookup and stores with almost no overhead. This
8669 is mainly used to store client source IP addresses.
8670
David du Colombier9a6d3c92011-03-17 10:40:24 +01008671 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
8672 This form is very compact (about 60 bytes per entry) and allows
8673 very fast entry lookup and stores with almost no overhead. This
8674 is mainly used to store client source IP addresses.
8675
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008676 integer a table declared with "type integer" will store 32bit integers
8677 which can represent a client identifier found in a request for
8678 instance.
8679
8680 string a table declared with "type string" will store substrings of up
8681 to <len> characters. If the string provided by the pattern
8682 extractor is larger than <len>, it will be truncated before
8683 being stored. During matching, at most <len> characters will be
8684 compared between the string in the table and the extracted
8685 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008686 to 32 characters.
8687
8688 binary a table declared with "type binary" will store binary blocks
8689 of <len> bytes. If the block provided by the pattern
8690 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02008691 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008692 is shorter than <len>, it will be padded by 0. When not
8693 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008694
8695 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008696 "string" type table (See type "string" above). Or the number
8697 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008698 changing this parameter as memory usage will proportionally
8699 increase.
8700
8701 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01008702 value directly impacts memory usage. Count approximately
8703 50 bytes per entry, plus the size of a string if any. The size
8704 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008705
8706 [nopurge] indicates that we refuse to purge older entries when the table
8707 is full. When not specified and the table is full when haproxy
8708 wants to store an entry in it, it will flush a few of the oldest
8709 entries in order to release some space for the new ones. This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01008710 most often the desired behavior. In some specific cases, it
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008711 be desirable to refuse new entries instead of purging the older
8712 ones. That may be the case when the amount of data to store is
8713 far above the hardware limits and we prefer not to offer access
8714 to new clients than to reject the ones already connected. When
8715 using this parameter, be sure to properly set the "expire"
8716 parameter (see below).
8717
Emeric Brunf099e792010-09-27 12:05:28 +02008718 <peersect> is the name of the peers section to use for replication. Entries
8719 which associate keys to server IDs are kept synchronized with
8720 the remote peers declared in this section. All entries are also
8721 automatically learned from the local peer (old process) during a
8722 soft restart.
8723
Willy Tarreau1abc6732015-05-01 19:21:02 +02008724 NOTE : each peers section may be referenced only by tables
8725 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008726
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008727 <expire> defines the maximum duration of an entry in the table since it
8728 was last created, refreshed or matched. The expiration delay is
8729 defined using the standard time format, similarly as the various
8730 timeouts. The maximum duration is slightly above 24 days. See
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03008731 section 2.4 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02008732 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008733 be removed once full. Be sure not to use the "nopurge" parameter
8734 if not expiration delay is specified.
8735
Willy Tarreau08d5f982010-06-06 13:34:54 +02008736 <data_type> is used to store additional information in the stick-table. This
8737 may be used by ACLs in order to control various criteria related
8738 to the activity of the client matching the stick-table. For each
8739 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02008740 that the additional data can fit. Several data types may be
8741 stored with an entry. Multiple data types may be specified after
8742 the "store" keyword, as a comma-separated list. Alternatively,
8743 it is possible to repeat the "store" keyword followed by one or
8744 several data types. Except for the "server_id" type which is
8745 automatically detected and enabled, all data types must be
8746 explicitly declared to be stored. If an ACL references a data
8747 type which is not stored, the ACL will simply not match. Some
8748 data types require an argument which must be passed just after
8749 the type between parenthesis. See below for the supported data
8750 types and their arguments.
8751
8752 The data types that can be stored with an entry are the following :
8753 - server_id : this is an integer which holds the numeric ID of the server a
8754 request was assigned to. It is used by the "stick match", "stick store",
8755 and "stick on" rules. It is automatically enabled when referenced.
8756
8757 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
8758 integer which may be used for anything. Most of the time it will be used
8759 to put a special tag on some entries, for instance to note that a
Davor Ocelice9ed2812017-12-25 17:49:28 +01008760 specific behavior was detected and must be known for future matches.
Willy Tarreauc9705a12010-07-27 20:05:50 +02008761
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008762 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
8763 over a period. It is a positive 32-bit integer integer which may be used
8764 for anything. Just like <gpc0>, it counts events, but instead of keeping
Davor Ocelice9ed2812017-12-25 17:49:28 +01008765 a cumulative number, it maintains the rate at which the counter is
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008766 incremented. Most of the time it will be used to measure the frequency of
Davor Ocelice9ed2812017-12-25 17:49:28 +01008767 occurrence of certain events (e.g. requests to a specific URL).
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008768
Frédéric Lécaille6778b272018-01-29 15:22:53 +01008769 - gpc1 : second General Purpose Counter. It is a positive 32-bit integer
8770 integer which may be used for anything. Most of the time it will be used
8771 to put a special tag on some entries, for instance to note that a
8772 specific behavior was detected and must be known for future matches.
8773
8774 - gpc1_rate(<period>) : increment rate of the second General Purpose Counter
8775 over a period. It is a positive 32-bit integer integer which may be used
8776 for anything. Just like <gpc1>, it counts events, but instead of keeping
8777 a cumulative number, it maintains the rate at which the counter is
8778 incremented. Most of the time it will be used to measure the frequency of
8779 occurrence of certain events (e.g. requests to a specific URL).
8780
Willy Tarreauc9705a12010-07-27 20:05:50 +02008781 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
8782 the absolute number of connections received from clients which matched
8783 this entry. It does not mean the connections were accepted, just that
8784 they were received.
8785
8786 - conn_cur : Current Connections. It is a positive 32-bit integer which
8787 stores the concurrent connection counts for the entry. It is incremented
8788 once an incoming connection matches the entry, and decremented once the
8789 connection leaves. That way it is possible to know at any time the exact
8790 number of concurrent connections for an entry.
8791
8792 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8793 integer parameter <period> which indicates in milliseconds the length
8794 of the period over which the average is measured. It reports the average
8795 incoming connection rate over that period, in connections per period. The
8796 result is an integer which can be matched using ACLs.
8797
8798 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
8799 the absolute number of sessions received from clients which matched this
8800 entry. A session is a connection that was accepted by the layer 4 rules.
8801
8802 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8803 integer parameter <period> which indicates in milliseconds the length
8804 of the period over which the average is measured. It reports the average
8805 incoming session rate over that period, in sessions per period. The
8806 result is an integer which can be matched using ACLs.
8807
8808 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
8809 counts the absolute number of HTTP requests received from clients which
8810 matched this entry. It does not matter whether they are valid requests or
8811 not. Note that this is different from sessions when keep-alive is used on
8812 the client side.
8813
8814 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8815 integer parameter <period> which indicates in milliseconds the length
8816 of the period over which the average is measured. It reports the average
8817 HTTP request rate over that period, in requests per period. The result is
8818 an integer which can be matched using ACLs. It does not matter whether
8819 they are valid requests or not. Note that this is different from sessions
8820 when keep-alive is used on the client side.
8821
8822 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
8823 counts the absolute number of HTTP requests errors induced by clients
8824 which matched this entry. Errors are counted on invalid and truncated
8825 requests, as well as on denied or tarpitted requests, and on failed
8826 authentications. If the server responds with 4xx, then the request is
8827 also counted as an error since it's an error triggered by the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01008828 (e.g. vulnerability scan).
Willy Tarreauc9705a12010-07-27 20:05:50 +02008829
8830 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8831 integer parameter <period> which indicates in milliseconds the length
8832 of the period over which the average is measured. It reports the average
8833 HTTP request error rate over that period, in requests per period (see
8834 http_err_cnt above for what is accounted as an error). The result is an
8835 integer which can be matched using ACLs.
8836
8837 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01008838 integer which counts the cumulative number of bytes received from clients
Willy Tarreauc9705a12010-07-27 20:05:50 +02008839 which matched this entry. Headers are included in the count. This may be
8840 used to limit abuse of upload features on photo or video servers.
8841
8842 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8843 integer parameter <period> which indicates in milliseconds the length
8844 of the period over which the average is measured. It reports the average
8845 incoming bytes rate over that period, in bytes per period. It may be used
8846 to detect users which upload too much and too fast. Warning: with large
8847 uploads, it is possible that the amount of uploaded data will be counted
8848 once upon termination, thus causing spikes in the average transfer speed
8849 instead of having a smooth one. This may partially be smoothed with
8850 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
8851 recommended for better fairness.
8852
8853 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01008854 integer which counts the cumulative number of bytes sent to clients which
Willy Tarreauc9705a12010-07-27 20:05:50 +02008855 matched this entry. Headers are included in the count. This may be used
8856 to limit abuse of bots sucking the whole site.
8857
8858 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
8859 an integer parameter <period> which indicates in milliseconds the length
8860 of the period over which the average is measured. It reports the average
8861 outgoing bytes rate over that period, in bytes per period. It may be used
8862 to detect users which download too much and too fast. Warning: with large
8863 transfers, it is possible that the amount of transferred data will be
8864 counted once upon termination, thus causing spikes in the average
8865 transfer speed instead of having a smooth one. This may partially be
8866 smoothed with "option contstats" though this is not perfect yet. Use of
8867 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02008868
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008869 There is only one stick-table per proxy. At the moment of writing this doc,
8870 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008871 to be required, simply create a dummy backend with a stick-table in it and
8872 reference it.
8873
8874 It is important to understand that stickiness based on learning information
8875 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01008876 lost upon restart unless peers are properly configured to transfer such
8877 information upon restart (recommended). In general it can be good as a
8878 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008879
Willy Tarreauc9705a12010-07-27 20:05:50 +02008880 Last, memory requirements may be important when storing many data types.
8881 Indeed, storing all indicators above at once in each entry requires 116 bytes
8882 per entry, or 116 MB for a 1-million entries table. This is definitely not
8883 something that can be ignored.
8884
8885 Example:
8886 # Keep track of counters of up to 1 million IP addresses over 5 minutes
8887 # and store a general purpose counter and the average connection rate
8888 # computed over a sliding window of 30 seconds.
8889 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
8890
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03008891 See also : "stick match", "stick on", "stick store-request", section 2.4
David du Colombiera13d1b92011-03-17 10:40:22 +01008892 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008893
8894
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008895stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01008896 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008897 May be used in sections : defaults | frontend | listen | backend
8898 no | no | yes | yes
8899
8900 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008901 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008902 describes what elements of the response or connection will
Davor Ocelice9ed2812017-12-25 17:49:28 +01008903 be analyzed, extracted and stored in the table once a
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008904 server is selected.
8905
8906 <table> is an optional stickiness table name. If unspecified, the same
8907 backend's table is used. A stickiness table is declared using
8908 the "stick-table" statement.
8909
8910 <cond> is an optional storage condition. It makes it possible to store
8911 certain criteria only when some conditions are met (or not met).
8912 For instance, it could be used to store the SSL session ID only
8913 when the response is a SSL server hello.
8914
8915 Some protocols or applications require complex stickiness rules and cannot
8916 always simply rely on cookies nor hashing. The "stick store-response"
8917 statement describes a rule to decide what to extract from the response and
8918 when to do it, in order to store it into a stickiness table for further
8919 requests to match it using the "stick match" statement. Obviously the
8920 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008921 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008922 See section 7 for a complete list of possible patterns and transformation
8923 rules.
8924
8925 The table has to be declared using the "stick-table" statement. It must be of
8926 a type compatible with the pattern. By default it is the one which is present
8927 in the same backend. It is possible to share a table with other backends by
8928 referencing it using the "table" keyword. If another table is referenced,
8929 the server's ID inside the backends are used. By default, all server IDs
8930 start at 1 in each backend, so the server ordering is enough. But in case of
8931 doubt, it is highly recommended to force server IDs using their "id" setting.
8932
8933 It is possible to restrict the conditions where a "stick store-response"
8934 statement will apply, using "if" or "unless" followed by a condition. This
8935 condition will be evaluated while parsing the response, so any criteria can
8936 be used. See section 7 for ACL based conditions.
8937
8938 There is no limit on the number of "stick store-response" statements, but
8939 there is a limit of 8 simultaneous stores per request or response. This
8940 makes it possible to store up to 8 criteria, all extracted from either the
8941 request or the response, regardless of the number of rules. Only the 8 first
8942 ones which match will be kept. Using this, it is possible to feed multiple
8943 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008944 another protocol or access method. Using multiple store-response rules with
8945 the same table is possible and may be used to find the best criterion to rely
8946 on, by arranging the rules by decreasing preference order. Only the first
8947 extracted criterion for a given table will be stored. All subsequent store-
8948 response rules referencing the same table will be skipped and their ACLs will
8949 not be evaluated. However, even if a store-request rule references a table, a
8950 store-response rule may also use the same table. This means that each table
8951 may learn exactly one element from the request and one element from the
8952 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008953
8954 The table will contain the real server that processed the request.
8955
8956 Example :
8957 # Learn SSL session ID from both request and response and create affinity.
8958 backend https
8959 mode tcp
8960 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02008961 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008962 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008963
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008964 acl clienthello req_ssl_hello_type 1
8965 acl serverhello rep_ssl_hello_type 2
8966
8967 # use tcp content accepts to detects ssl client and server hello.
8968 tcp-request inspect-delay 5s
8969 tcp-request content accept if clienthello
8970
8971 # no timeout on response inspect delay by default.
8972 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008973
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008974 # SSL session ID (SSLID) may be present on a client or server hello.
8975 # Its length is coded on 1 byte at offset 43 and its value starts
8976 # at offset 44.
8977
8978 # Match and learn on request if client hello.
8979 stick on payload_lv(43,1) if clienthello
8980
8981 # Learn on response if server hello.
8982 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02008983
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008984 server s1 192.168.1.1:443
8985 server s2 192.168.1.1:443
8986
8987 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
8988 extraction.
8989
8990
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008991tcp-check connect [params*]
8992 Opens a new connection
8993 May be used in sections: defaults | frontend | listen | backend
8994 no | no | yes | yes
8995
8996 When an application lies on more than a single TCP port or when HAProxy
8997 load-balance many services in a single backend, it makes sense to probe all
8998 the services individually before considering a server as operational.
8999
9000 When there are no TCP port configured on the server line neither server port
9001 directive, then the 'tcp-check connect port <port>' must be the first step
9002 of the sequence.
9003
9004 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
9005 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
9006 do.
9007
9008 Parameters :
9009 They are optional and can be used to describe how HAProxy should open and
9010 use the TCP connection.
9011
9012 port if not set, check port or server port is used.
9013 It tells HAProxy where to open the connection to.
9014 <port> must be a valid TCP port source integer, from 1 to 65535.
9015
9016 send-proxy send a PROXY protocol string
9017
9018 ssl opens a ciphered connection
9019
9020 Examples:
9021 # check HTTP and HTTPs services on a server.
9022 # first open port 80 thanks to server line port directive, then
9023 # tcp-check opens port 443, ciphered and run a request on it:
9024 option tcp-check
9025 tcp-check connect
9026 tcp-check send GET\ /\ HTTP/1.0\r\n
9027 tcp-check send Host:\ haproxy.1wt.eu\r\n
9028 tcp-check send \r\n
9029 tcp-check expect rstring (2..|3..)
9030 tcp-check connect port 443 ssl
9031 tcp-check send GET\ /\ HTTP/1.0\r\n
9032 tcp-check send Host:\ haproxy.1wt.eu\r\n
9033 tcp-check send \r\n
9034 tcp-check expect rstring (2..|3..)
9035 server www 10.0.0.1 check port 80
9036
9037 # check both POP and IMAP from a single server:
9038 option tcp-check
9039 tcp-check connect port 110
9040 tcp-check expect string +OK\ POP3\ ready
9041 tcp-check connect port 143
9042 tcp-check expect string *\ OK\ IMAP4\ ready
9043 server mail 10.0.0.1 check
9044
9045 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
9046
9047
9048tcp-check expect [!] <match> <pattern>
Davor Ocelice9ed2812017-12-25 17:49:28 +01009049 Specify data to be collected and analyzed during a generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009050 May be used in sections: defaults | frontend | listen | backend
9051 no | no | yes | yes
9052
9053 Arguments :
9054 <match> is a keyword indicating how to look for a specific pattern in the
9055 response. The keyword may be one of "string", "rstring" or
9056 binary.
9057 The keyword may be preceded by an exclamation mark ("!") to negate
9058 the match. Spaces are allowed between the exclamation mark and the
9059 keyword. See below for more details on the supported keywords.
9060
9061 <pattern> is the pattern to look for. It may be a string or a regular
9062 expression. If the pattern contains spaces, they must be escaped
9063 with the usual backslash ('\').
9064 If the match is set to binary, then the pattern must be passed as
Davor Ocelice9ed2812017-12-25 17:49:28 +01009065 a series of hexadecimal digits in an even number. Each sequence of
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009066 two digits will represent a byte. The hexadecimal digits may be
9067 used upper or lower case.
9068
9069
9070 The available matches are intentionally similar to their http-check cousins :
9071
9072 string <string> : test the exact string matches in the response buffer.
9073 A health check response will be considered valid if the
9074 response's buffer contains this exact string. If the
9075 "string" keyword is prefixed with "!", then the response
9076 will be considered invalid if the body contains this
9077 string. This can be used to look for a mandatory pattern
9078 in a protocol response, or to detect a failure when a
9079 specific error appears in a protocol banner.
9080
9081 rstring <regex> : test a regular expression on the response buffer.
9082 A health check response will be considered valid if the
9083 response's buffer matches this expression. If the
9084 "rstring" keyword is prefixed with "!", then the response
9085 will be considered invalid if the body matches the
9086 expression.
9087
9088 binary <hexstring> : test the exact string in its hexadecimal form matches
9089 in the response buffer. A health check response will
9090 be considered valid if the response's buffer contains
9091 this exact hexadecimal string.
9092 Purpose is to match data on binary protocols.
9093
9094 It is important to note that the responses will be limited to a certain size
9095 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
9096 Thus, too large responses may not contain the mandatory pattern when using
9097 "string", "rstring" or binary. If a large response is absolutely required, it
9098 is possible to change the default max size by setting the global variable.
9099 However, it is worth keeping in mind that parsing very large responses can
9100 waste some CPU cycles, especially when regular expressions are used, and that
9101 it is always better to focus the checks on smaller resources. Also, in its
9102 current state, the check will not find any string nor regex past a null
9103 character in the response. Similarly it is not possible to request matching
9104 the null character.
9105
9106 Examples :
9107 # perform a POP check
9108 option tcp-check
9109 tcp-check expect string +OK\ POP3\ ready
9110
9111 # perform an IMAP check
9112 option tcp-check
9113 tcp-check expect string *\ OK\ IMAP4\ ready
9114
9115 # look for the redis master server
9116 option tcp-check
9117 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02009118 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009119 tcp-check send info\ replication\r\n
9120 tcp-check expect string role:master
9121 tcp-check send QUIT\r\n
9122 tcp-check expect string +OK
9123
9124
9125 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
9126 "tcp-check send-binary", "http-check expect", tune.chksize
9127
9128
9129tcp-check send <data>
9130 Specify a string to be sent as a question during a generic health check
9131 May be used in sections: defaults | frontend | listen | backend
9132 no | no | yes | yes
9133
9134 <data> : the data to be sent as a question during a generic health check
9135 session. For now, <data> must be a string.
9136
9137 Examples :
9138 # look for the redis master server
9139 option tcp-check
9140 tcp-check send info\ replication\r\n
9141 tcp-check expect string role:master
9142
9143 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9144 "tcp-check send-binary", tune.chksize
9145
9146
Davor Ocelice9ed2812017-12-25 17:49:28 +01009147tcp-check send-binary <hexstring>
9148 Specify a hex digits string to be sent as a binary question during a raw
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009149 tcp health check
9150 May be used in sections: defaults | frontend | listen | backend
9151 no | no | yes | yes
9152
9153 <data> : the data to be sent as a question during a generic health check
9154 session. For now, <data> must be a string.
Davor Ocelice9ed2812017-12-25 17:49:28 +01009155 <hexstring> : test the exact string in its hexadecimal form matches in the
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009156 response buffer. A health check response will be considered
9157 valid if the response's buffer contains this exact
9158 hexadecimal string.
9159 Purpose is to send binary data to ask on binary protocols.
9160
9161 Examples :
9162 # redis check in binary
9163 option tcp-check
9164 tcp-check send-binary 50494e470d0a # PING\r\n
9165 tcp-check expect binary 2b504F4e47 # +PONG
9166
9167
9168 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9169 "tcp-check send", tune.chksize
9170
9171
Willy Tarreaue9656522010-08-17 15:40:09 +02009172tcp-request connection <action> [{if | unless} <condition>]
9173 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02009174 May be used in sections : defaults | frontend | listen | backend
9175 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02009176 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009177 <action> defines the action to perform if the condition applies. See
9178 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02009179
Willy Tarreaue9656522010-08-17 15:40:09 +02009180 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009181
9182 Immediately after acceptance of a new incoming connection, it is possible to
9183 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02009184 or dropped or have its counters tracked. Those conditions cannot make use of
9185 any data contents because the connection has not been read from yet, and the
9186 buffers are not yet allocated. This is used to selectively and very quickly
9187 accept or drop connections from various sources with a very low overhead. If
9188 some contents need to be inspected in order to take the decision, the
9189 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009190
Willy Tarreaue9656522010-08-17 15:40:09 +02009191 The "tcp-request connection" rules are evaluated in their exact declaration
9192 order. If no rule matches or if there is no rule, the default action is to
9193 accept the incoming connection. There is no specific limit to the number of
9194 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009195
Willy Tarreaua9083d02015-05-08 15:27:59 +02009196 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02009197 - accept :
9198 accepts the connection if the condition is true (when used with "if")
9199 or false (when used with "unless"). The first such rule executed ends
9200 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009201
Willy Tarreaue9656522010-08-17 15:40:09 +02009202 - reject :
9203 rejects the connection if the condition is true (when used with "if")
9204 or false (when used with "unless"). The first such rule executed ends
9205 the rules evaluation. Rejected connections do not even become a
9206 session, which is why they are accounted separately for in the stats,
9207 as "denied connections". They are not considered for the session
9208 rate-limit and are not logged either. The reason is that these rules
9209 should only be used to filter extremely high connection rates such as
9210 the ones encountered during a massive DDoS attack. Under these extreme
9211 conditions, the simple action of logging each event would make the
9212 system collapse and would considerably lower the filtering capacity. If
9213 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +02009214 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009215
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009216 - expect-proxy layer4 :
9217 configures the client-facing connection to receive a PROXY protocol
9218 header before any byte is read from the socket. This is equivalent to
9219 having the "accept-proxy" keyword on the "bind" line, except that using
9220 the TCP rule allows the PROXY protocol to be accepted only for certain
9221 IP address ranges using an ACL. This is convenient when multiple layers
9222 of load balancers are passed through by traffic coming from public
9223 hosts.
9224
Bertrand Jacquin90759682016-06-06 15:35:39 +01009225 - expect-netscaler-cip layer4 :
9226 configures the client-facing connection to receive a NetScaler Client
9227 IP insertion protocol header before any byte is read from the socket.
9228 This is equivalent to having the "accept-netscaler-cip" keyword on the
9229 "bind" line, except that using the TCP rule allows the PROXY protocol
9230 to be accepted only for certain IP address ranges using an ACL. This
9231 is convenient when multiple layers of load balancers are passed
9232 through by traffic coming from public hosts.
9233
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009234 - capture <sample> len <length> :
9235 This only applies to "tcp-request content" rules. It captures sample
9236 expression <sample> from the request buffer, and converts it to a
9237 string of at most <len> characters. The resulting string is stored into
9238 the next request "capture" slot, so it will possibly appear next to
9239 some captured HTTP headers. It will then automatically appear in the
9240 logs, and it will be possible to extract it using sample fetch rules to
9241 feed it into headers or anything. The length should be limited given
9242 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02009243 session life. Please check section 7.3 (Fetching samples) and "capture
9244 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009245
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009246 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02009247 enables tracking of sticky counters from current connection. These
Willy Tarreau09448f72014-06-25 18:12:15 +02009248 rules do not stop evaluation and do not change default action. 3 sets
Willy Tarreaue9656522010-08-17 15:40:09 +02009249 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009250 first "track-sc0" rule executed enables tracking of the counters of the
9251 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02009252 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009253 set. The first "track-sc2" rule executed enables tracking of the
9254 counters of the specified table as the third set. It is a recommended
9255 practice to use the first set of counters for the per-frontend counters
9256 and the second set for the per-backend ones. But this is just a
9257 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009258
Willy Tarreaue9656522010-08-17 15:40:09 +02009259 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009260 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02009261 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +01009262 request or connection will be analyzed, extracted, combined,
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009263 and used to select which table entry to update the counters.
9264 Note that "tcp-request connection" cannot use content-based
9265 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009266
Willy Tarreaue9656522010-08-17 15:40:09 +02009267 <table> is an optional table to be used instead of the default one,
9268 which is the stick-table declared in the current proxy. All
9269 the counters for the matches and updates for the key will
9270 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009271
Willy Tarreaue9656522010-08-17 15:40:09 +02009272 Once a "track-sc*" rule is executed, the key is looked up in the table
9273 and if it is not found, an entry is allocated for it. Then a pointer to
9274 that entry is kept during all the session's life, and this entry's
9275 counters are updated as often as possible, every time the session's
9276 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009277 Counters are only updated for events that happen after the tracking has
9278 been started. For example, connection counters will not be updated when
9279 tracking layer 7 information, since the connection event happens before
9280 layer7 information is extracted.
9281
Willy Tarreaue9656522010-08-17 15:40:09 +02009282 If the entry tracks concurrent connection counters, one connection is
9283 counted for as long as the entry is tracked, and the entry will not
9284 expire during that time. Tracking counters also provides a performance
9285 advantage over just checking the keys, because only one table lookup is
9286 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009287
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009288 - sc-inc-gpc0(<sc-id>):
9289 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
9290 counter designated by <sc-id>. If an error occurs, this action silently
9291 fails and the actions evaluation continues.
9292
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009293 - sc-inc-gpc1(<sc-id>):
9294 The "sc-inc-gpc1" increments the GPC1 counter according to the sticky
9295 counter designated by <sc-id>. If an error occurs, this action silently
9296 fails and the actions evaluation continues.
9297
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009298 - sc-set-gpt0(<sc-id>) <int>:
9299 This action sets the GPT0 tag according to the sticky counter designated
9300 by <sc-id> and the value of <int>. The expected result is a boolean. If
9301 an error occurs, this action silently fails and the actions evaluation
9302 continues.
9303
William Lallemand2e785f22016-05-25 01:48:42 +02009304 - set-src <expr> :
9305 Is used to set the source IP address to the value of specified
9306 expression. Useful if you want to mask source IP for privacy.
9307 If you want to provide an IP from a HTTP header use "http-request
9308 set-src"
9309
9310 <expr> Is a standard HAProxy expression formed by a sample-fetch
9311 followed by some converters.
9312
9313 Example:
9314
9315 tcp-request connection set-src src,ipmask(24)
9316
Willy Tarreau0c630532016-10-21 17:52:58 +02009317 When possible, set-src preserves the original source port as long as the
9318 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +02009319
William Lallemand44be6402016-05-25 01:51:35 +02009320 - set-src-port <expr> :
9321 Is used to set the source port address to the value of specified
9322 expression.
9323
9324 <expr> Is a standard HAProxy expression formed by a sample-fetch
9325 followed by some converters.
9326
9327 Example:
9328
9329 tcp-request connection set-src-port int(4000)
9330
Willy Tarreau0c630532016-10-21 17:52:58 +02009331 When possible, set-src-port preserves the original source address as long
9332 as the address family supports a port, otherwise it forces the source
9333 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02009334
William Lallemand13e9b0c2016-05-25 02:34:07 +02009335 - set-dst <expr> :
9336 Is used to set the destination IP address to the value of specified
9337 expression. Useful if you want to mask IP for privacy in log.
9338 If you want to provide an IP from a HTTP header use "http-request
9339 set-dst". If you want to connect to the new address/port, use
9340 '0.0.0.0:0' as a server address in the backend.
9341
9342 <expr> Is a standard HAProxy expression formed by a sample-fetch
9343 followed by some converters.
9344
9345 Example:
9346
9347 tcp-request connection set-dst dst,ipmask(24)
9348 tcp-request connection set-dst ipv4(10.0.0.1)
9349
Willy Tarreau0c630532016-10-21 17:52:58 +02009350 When possible, set-dst preserves the original destination port as long as
9351 the address family allows it, otherwise the destination port is set to 0.
9352
William Lallemand13e9b0c2016-05-25 02:34:07 +02009353 - set-dst-port <expr> :
9354 Is used to set the destination port address to the value of specified
9355 expression. If you want to connect to the new address/port, use
9356 '0.0.0.0:0' as a server address in the backend.
9357
9358
9359 <expr> Is a standard HAProxy expression formed by a sample-fetch
9360 followed by some converters.
9361
9362 Example:
9363
9364 tcp-request connection set-dst-port int(4000)
9365
Willy Tarreau0c630532016-10-21 17:52:58 +02009366 When possible, set-dst-port preserves the original destination address as
9367 long as the address family supports a port, otherwise it forces the
9368 destination address to IPv4 "0.0.0.0" before rewriting the port.
9369
Willy Tarreau2d392c22015-08-24 01:43:45 +02009370 - "silent-drop" :
9371 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +01009372 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +02009373 to prevent the client from being notified. The effect it then that the
9374 client still sees an established connection while there's none on
9375 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9376 except that it doesn't use any local resource at all on the machine
9377 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +01009378 slow down stronger attackers. It is important to understand the impact
9379 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009380 client and HAProxy (firewalls, proxies, load balancers) will also keep
9381 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +01009382 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009383 TCP_REPAIR socket option is used to block the emission of a TCP
9384 reset. On other systems, the socket's TTL is reduced to 1 so that the
9385 TCP reset doesn't pass the first router, though it's still delivered to
9386 local networks. Do not use it unless you fully understand how it works.
9387
Willy Tarreaue9656522010-08-17 15:40:09 +02009388 Note that the "if/unless" condition is optional. If no condition is set on
9389 the action, it is simply performed unconditionally. That can be useful for
9390 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009391
Willy Tarreaue9656522010-08-17 15:40:09 +02009392 Example: accept all connections from white-listed hosts, reject too fast
9393 connection without counting them, and track accepted connections.
9394 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009395
Willy Tarreaue9656522010-08-17 15:40:09 +02009396 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009397 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009398 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009399
Willy Tarreaue9656522010-08-17 15:40:09 +02009400 Example: accept all connections from white-listed hosts, count all other
9401 connections and reject too fast ones. This results in abusive ones
9402 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009403
Willy Tarreaue9656522010-08-17 15:40:09 +02009404 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009405 tcp-request connection track-sc0 src
9406 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009407
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009408 Example: enable the PROXY protocol for traffic coming from all known proxies.
9409
9410 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
9411
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009412 See section 7 about ACL usage.
9413
Willy Tarreau4f614292016-10-21 17:49:36 +02009414 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009415
9416
Willy Tarreaue9656522010-08-17 15:40:09 +02009417tcp-request content <action> [{if | unless} <condition>]
9418 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02009419 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02009420 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02009421 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009422 <action> defines the action to perform if the condition applies. See
9423 below.
Willy Tarreau62644772008-07-16 18:36:06 +02009424
Willy Tarreaue9656522010-08-17 15:40:09 +02009425 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02009426
Davor Ocelice9ed2812017-12-25 17:49:28 +01009427 A request's contents can be analyzed at an early stage of request processing
Willy Tarreaue9656522010-08-17 15:40:09 +02009428 called "TCP content inspection". During this stage, ACL-based rules are
9429 evaluated every time the request contents are updated, until either an
9430 "accept" or a "reject" rule matches, or the TCP request inspection delay
9431 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02009432
Willy Tarreaue9656522010-08-17 15:40:09 +02009433 The first difference between these rules and "tcp-request connection" rules
9434 is that "tcp-request content" rules can make use of contents to take a
9435 decision. Most often, these decisions will consider a protocol recognition or
9436 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01009437 both frontends and backends. In case of HTTP keep-alive with the client, all
9438 tcp-request content rules are evaluated again, so haproxy keeps a record of
9439 what sticky counters were assigned by a "tcp-request connection" versus a
9440 "tcp-request content" rule, and flushes all the content-related ones after
9441 processing an HTTP request, so that they may be evaluated again by the rules
9442 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009443 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01009444 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02009445
Willy Tarreaue9656522010-08-17 15:40:09 +02009446 Content-based rules are evaluated in their exact declaration order. If no
9447 rule matches or if there is no rule, the default action is to accept the
9448 contents. There is no specific limit to the number of rules which may be
9449 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02009450
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009451 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009452 - accept : the request is accepted
9453 - reject : the request is rejected and the connection is closed
9454 - capture : the specified sample expression is captured
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009455 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009456 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009457 - sc-inc-gpc1(<sc-id>)
Thierry Fournierb9125672016-03-29 19:34:37 +02009458 - sc-set-gpt0(<sc-id>) <int>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009459 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +01009460 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +02009461 - silent-drop
Davor Ocelice9ed2812017-12-25 17:49:28 +01009462 - send-spoe-group <engine-name> <group-name>
Willy Tarreau62644772008-07-16 18:36:06 +02009463
Willy Tarreaue9656522010-08-17 15:40:09 +02009464 They have the same meaning as their counter-parts in "tcp-request connection"
9465 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02009466
Willy Tarreauf3338342014-01-28 21:40:28 +01009467 While there is nothing mandatory about it, it is recommended to use the
9468 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
9469 content" rules in the frontend, and track-sc2 for "tcp-request content"
9470 rules in the backend, because that makes the configuration more readable
9471 and easier to troubleshoot, but this is just a guideline and all counters
9472 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02009473
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009474 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02009475 the action, it is simply performed unconditionally. That can be useful for
9476 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02009477
Willy Tarreaue9656522010-08-17 15:40:09 +02009478 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02009479 rules, since HTTP-specific ACL matches are able to preliminarily parse the
9480 contents of a buffer before extracting the required data. If the buffered
9481 contents do not parse as a valid HTTP message, then the ACL does not match.
9482 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01009483 processing, so there is no risk of parsing something differently. In an HTTP
9484 backend connected to from an HTTP frontend, it is guaranteed that HTTP
9485 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02009486
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009487 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009488 are present when the rule is processed. The rule processing engine is able to
9489 wait until the inspect delay expires when the data to be tracked is not yet
9490 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009491
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009492 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +02009493 declared inline. For "tcp-request session" rules, only session-level
9494 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009495
Daniel Schneller0b547052016-03-21 20:46:57 +01009496 <var-name> The name of the variable starts with an indication about
9497 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01009498 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01009499 "sess" : the variable is shared with the whole session
9500 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009501 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01009502 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009503 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01009504 "res" : the variable is shared only during response
9505 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009506 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01009507 The name may only contain characters 'a-z', 'A-Z', '0-9',
9508 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009509
9510 <expr> Is a standard HAProxy expression formed by a sample-fetch
9511 followed by some converters.
9512
Christopher Faulet85d79c92016-11-09 16:54:56 +01009513 The "unset-var" is used to unset a variable. See above for details about
9514 <var-name>.
9515
Christopher Faulet76c09ef2017-09-21 11:03:52 +02009516 The "send-spoe-group" is used to trigger sending of a group of SPOE
9517 messages. To do so, the SPOE engine used to send messages must be defined, as
9518 well as the SPOE group to send. Of course, the SPOE engine must refer to an
9519 existing SPOE filter. If not engine name is provided on the SPOE filter line,
9520 the SPOE agent name must be used.
9521
9522 <engine-name> The SPOE engine name.
9523
9524 <group-name> The SPOE group name as specified in the engine configuration.
9525
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009526 Example:
9527
9528 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +01009529 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009530
Willy Tarreau62644772008-07-16 18:36:06 +02009531 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02009532 # Accept HTTP requests containing a Host header saying "example.com"
9533 # and reject everything else.
9534 acl is_host_com hdr(Host) -i example.com
9535 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02009536 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02009537 tcp-request content reject
9538
9539 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02009540 # reject SMTP connection if client speaks first
9541 tcp-request inspect-delay 30s
9542 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009543 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02009544
9545 # Forward HTTPS connection only if client speaks
9546 tcp-request inspect-delay 30s
9547 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009548 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02009549 tcp-request content reject
9550
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009551 Example:
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009552 # Track the last IP(stick-table type string) from X-Forwarded-For
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009553 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009554 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009555 # Or track the last IP(stick-table type ip|ipv6) from X-Forwarded-For
9556 tcp-request content track-sc0 req.hdr_ip(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009557
9558 Example:
9559 # track request counts per "base" (concatenation of Host+URL)
9560 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009561 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009562
Willy Tarreaue9656522010-08-17 15:40:09 +02009563 Example: track per-frontend and per-backend counters, block abusers at the
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009564 frontend when the backend detects abuse(and marks gpc0).
Willy Tarreaue9656522010-08-17 15:40:09 +02009565
9566 frontend http
Davor Ocelice9ed2812017-12-25 17:49:28 +01009567 # Use General Purpose Counter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02009568 # protecting all our sites
9569 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009570 tcp-request connection track-sc0 src
9571 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02009572 ...
9573 use_backend http_dynamic if { path_end .php }
9574
9575 backend http_dynamic
9576 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009577 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02009578 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009579 acl click_too_fast sc1_http_req_rate gt 10
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009580 acl mark_as_abuser sc0_inc_gpc0(http) gt 0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009581 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02009582 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02009583
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009584 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02009585
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03009586 See also : "tcp-request connection", "tcp-request session",
9587 "tcp-request inspect-delay", and "http-request".
Willy Tarreau62644772008-07-16 18:36:06 +02009588
9589
9590tcp-request inspect-delay <timeout>
9591 Set the maximum allowed time to wait for data during content inspection
9592 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02009593 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02009594 Arguments :
9595 <timeout> is the timeout value specified in milliseconds by default, but
9596 can be in any other unit if the number is suffixed by the unit,
9597 as explained at the top of this document.
9598
9599 People using haproxy primarily as a TCP relay are often worried about the
9600 risk of passing any type of protocol to a server without any analysis. In
9601 order to be able to analyze the request contents, we must first withhold
9602 the data then analyze them. This statement simply enables withholding of
9603 data for at most the specified amount of time.
9604
Willy Tarreaufb356202010-08-03 14:02:05 +02009605 TCP content inspection applies very early when a connection reaches a
9606 frontend, then very early when the connection is forwarded to a backend. This
9607 means that a connection may experience a first delay in the frontend and a
9608 second delay in the backend if both have tcp-request rules.
9609
Willy Tarreau62644772008-07-16 18:36:06 +02009610 Note that when performing content inspection, haproxy will evaluate the whole
9611 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009612 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02009613 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01009614 contents are definitive. If no delay is set, haproxy will not wait at all
9615 and will immediately apply a verdict based on the available information.
9616 Obviously this is unlikely to be very useful and might even be racy, so such
9617 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02009618
9619 As soon as a rule matches, the request is released and continues as usual. If
9620 the timeout is reached and no rule matches, the default policy will be to let
9621 it pass through unaffected.
9622
9623 For most protocols, it is enough to set it to a few seconds, as most clients
9624 send the full request immediately upon connection. Add 3 or more seconds to
9625 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01009626 to use large values, for instance to ensure that the client never talks
Davor Ocelice9ed2812017-12-25 17:49:28 +01009627 before the server (e.g. SMTP), or to wait for a client to talk before passing
9628 data to the server (e.g. SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02009629 least the inspection delay, otherwise it will expire first. If the client
9630 closes the connection or if the buffer is full, the delay immediately expires
9631 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02009632
Willy Tarreau55165fe2009-05-10 12:02:55 +02009633 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02009634 "timeout client".
9635
9636
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009637tcp-response content <action> [{if | unless} <condition>]
9638 Perform an action on a session response depending on a layer 4-7 condition
9639 May be used in sections : defaults | frontend | listen | backend
9640 no | no | yes | yes
9641 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009642 <action> defines the action to perform if the condition applies. See
9643 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009644
9645 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
9646
Davor Ocelice9ed2812017-12-25 17:49:28 +01009647 Response contents can be analyzed at an early stage of response processing
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009648 called "TCP content inspection". During this stage, ACL-based rules are
9649 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009650 "accept", "close" or a "reject" rule matches, or a TCP response inspection
9651 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009652
9653 Most often, these decisions will consider a protocol recognition or validity.
9654
9655 Content-based rules are evaluated in their exact declaration order. If no
9656 rule matches or if there is no rule, the default action is to accept the
9657 contents. There is no specific limit to the number of rules which may be
9658 inserted.
9659
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009660 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009661 - accept :
9662 accepts the response if the condition is true (when used with "if")
9663 or false (when used with "unless"). The first such rule executed ends
9664 the rules evaluation.
9665
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009666 - close :
9667 immediately closes the connection with the server if the condition is
9668 true (when used with "if"), or false (when used with "unless"). The
9669 first such rule executed ends the rules evaluation. The main purpose of
9670 this action is to force a connection to be finished between a client
9671 and a server after an exchange when the application protocol expects
9672 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009673 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009674 protocols.
9675
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009676 - reject :
9677 rejects the response if the condition is true (when used with "if")
9678 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009679 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009680
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009681 - set-var(<var-name>) <expr>
9682 Sets a variable.
9683
Christopher Faulet85d79c92016-11-09 16:54:56 +01009684 - unset-var(<var-name>)
9685 Unsets a variable.
9686
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009687 - sc-inc-gpc0(<sc-id>):
9688 This action increments the GPC0 counter according to the sticky
9689 counter designated by <sc-id>. If an error occurs, this action fails
9690 silently and the actions evaluation continues.
9691
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009692 - sc-inc-gpc1(<sc-id>):
9693 This action increments the GPC1 counter according to the sticky
9694 counter designated by <sc-id>. If an error occurs, this action fails
9695 silently and the actions evaluation continues.
9696
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009697 - sc-set-gpt0(<sc-id>) <int> :
9698 This action sets the GPT0 tag according to the sticky counter designated
9699 by <sc-id> and the value of <int>. The expected result is a boolean. If
9700 an error occurs, this action silently fails and the actions evaluation
9701 continues.
9702
Willy Tarreau2d392c22015-08-24 01:43:45 +02009703 - "silent-drop" :
9704 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +01009705 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +02009706 to prevent the client from being notified. The effect it then that the
9707 client still sees an established connection while there's none on
9708 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9709 except that it doesn't use any local resource at all on the machine
9710 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +01009711 slow down stronger attackers. It is important to understand the impact
9712 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009713 client and HAProxy (firewalls, proxies, load balancers) will also keep
9714 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +01009715 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009716 TCP_REPAIR socket option is used to block the emission of a TCP
9717 reset. On other systems, the socket's TTL is reduced to 1 so that the
9718 TCP reset doesn't pass the first router, though it's still delivered to
9719 local networks. Do not use it unless you fully understand how it works.
9720
Christopher Faulet76c09ef2017-09-21 11:03:52 +02009721 - send-spoe-group <engine-name> <group-name>
9722 Send a group of SPOE messages.
9723
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009724 Note that the "if/unless" condition is optional. If no condition is set on
9725 the action, it is simply performed unconditionally. That can be useful for
9726 for changing the default action to a reject.
9727
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009728 It is perfectly possible to match layer 7 contents with "tcp-response
9729 content" rules, but then it is important to ensure that a full response has
9730 been buffered, otherwise no contents will match. In order to achieve this,
9731 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009732 period.
9733
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009734 The "set-var" is used to set the content of a variable. The variable is
9735 declared inline.
9736
Daniel Schneller0b547052016-03-21 20:46:57 +01009737 <var-name> The name of the variable starts with an indication about
9738 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01009739 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01009740 "sess" : the variable is shared with the whole session
9741 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009742 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01009743 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009744 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01009745 "res" : the variable is shared only during response
9746 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009747 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01009748 The name may only contain characters 'a-z', 'A-Z', '0-9',
9749 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009750
9751 <expr> Is a standard HAProxy expression formed by a sample-fetch
9752 followed by some converters.
9753
9754 Example:
9755
9756 tcp-request content set-var(sess.my_var) src
9757
Christopher Faulet85d79c92016-11-09 16:54:56 +01009758 The "unset-var" is used to unset a variable. See above for details about
9759 <var-name>.
9760
9761 Example:
9762
9763 tcp-request content unset-var(sess.my_var)
9764
Christopher Faulet76c09ef2017-09-21 11:03:52 +02009765 The "send-spoe-group" is used to trigger sending of a group of SPOE
9766 messages. To do so, the SPOE engine used to send messages must be defined, as
9767 well as the SPOE group to send. Of course, the SPOE engine must refer to an
9768 existing SPOE filter. If not engine name is provided on the SPOE filter line,
9769 the SPOE agent name must be used.
9770
9771 <engine-name> The SPOE engine name.
9772
9773 <group-name> The SPOE group name as specified in the engine configuration.
9774
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009775 See section 7 about ACL usage.
9776
9777 See also : "tcp-request content", "tcp-response inspect-delay"
9778
9779
Willy Tarreau4f614292016-10-21 17:49:36 +02009780tcp-request session <action> [{if | unless} <condition>]
9781 Perform an action on a validated session depending on a layer 5 condition
9782 May be used in sections : defaults | frontend | listen | backend
9783 no | yes | yes | no
9784 Arguments :
9785 <action> defines the action to perform if the condition applies. See
9786 below.
9787
9788 <condition> is a standard layer5-only ACL-based condition (see section 7).
9789
Davor Ocelice9ed2812017-12-25 17:49:28 +01009790 Once a session is validated, (i.e. after all handshakes have been completed),
Willy Tarreau4f614292016-10-21 17:49:36 +02009791 it is possible to evaluate some conditions to decide whether this session
9792 must be accepted or dropped or have its counters tracked. Those conditions
9793 cannot make use of any data contents because no buffers are allocated yet and
9794 the processing cannot wait at this stage. The main use case it to copy some
9795 early information into variables (since variables are accessible in the
9796 session), or to keep track of some information collected after the handshake,
9797 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
Davor Ocelice9ed2812017-12-25 17:49:28 +01009798 from the PROXY protocol header (e.g. track a source forwarded this way). The
Willy Tarreau4f614292016-10-21 17:49:36 +02009799 extracted information can thus be copied to a variable or tracked using
9800 "track-sc" rules. Of course it is also possible to decide to accept/reject as
9801 with other rulesets. Most operations performed here could also be performed
9802 in "tcp-request content" rules, except that in HTTP these rules are evaluated
9803 for each new request, and that might not always be acceptable. For example a
9804 rule might increment a counter on each evaluation. It would also be possible
9805 that a country is resolved by geolocation from the source IP address,
9806 assigned to a session-wide variable, then the source address rewritten from
9807 an HTTP header for all requests. If some contents need to be inspected in
9808 order to take the decision, the "tcp-request content" statements must be used
9809 instead.
9810
9811 The "tcp-request session" rules are evaluated in their exact declaration
9812 order. If no rule matches or if there is no rule, the default action is to
9813 accept the incoming session. There is no specific limit to the number of
9814 rules which may be inserted.
9815
9816 Several types of actions are supported :
9817 - accept : the request is accepted
9818 - reject : the request is rejected and the connection is closed
9819 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
9820 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009821 - sc-inc-gpc1(<sc-id>)
Willy Tarreau4f614292016-10-21 17:49:36 +02009822 - sc-set-gpt0(<sc-id>) <int>
9823 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +01009824 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +02009825 - silent-drop
9826
9827 These actions have the same meaning as their respective counter-parts in
9828 "tcp-request connection" and "tcp-request content", so please refer to these
9829 sections for a complete description.
9830
9831 Note that the "if/unless" condition is optional. If no condition is set on
9832 the action, it is simply performed unconditionally. That can be useful for
9833 "track-sc*" actions as well as for changing the default action to a reject.
9834
9835 Example: track the original source address by default, or the one advertised
9836 in the PROXY protocol header for connection coming from the local
9837 proxies. The first connection-level rule enables receipt of the
9838 PROXY protocol for these ones, the second rule tracks whatever
9839 address we decide to keep after optional decoding.
9840
9841 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
9842 tcp-request session track-sc0 src
9843
9844 Example: accept all sessions from white-listed hosts, reject too fast
9845 sessions without counting them, and track accepted sessions.
9846 This results in session rate being capped from abusive sources.
9847
9848 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
9849 tcp-request session reject if { src_sess_rate gt 10 }
9850 tcp-request session track-sc0 src
9851
9852 Example: accept all sessions from white-listed hosts, count all other
9853 sessions and reject too fast ones. This results in abusive ones
9854 being blocked as long as they don't slow down.
9855
9856 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
9857 tcp-request session track-sc0 src
9858 tcp-request session reject if { sc0_sess_rate gt 10 }
9859
9860 See section 7 about ACL usage.
9861
9862 See also : "tcp-request connection", "tcp-request content", "stick-table"
9863
9864
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009865tcp-response inspect-delay <timeout>
9866 Set the maximum allowed time to wait for a response during content inspection
9867 May be used in sections : defaults | frontend | listen | backend
9868 no | no | yes | yes
9869 Arguments :
9870 <timeout> is the timeout value specified in milliseconds by default, but
9871 can be in any other unit if the number is suffixed by the unit,
9872 as explained at the top of this document.
9873
9874 See also : "tcp-response content", "tcp-request inspect-delay".
9875
9876
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009877timeout check <timeout>
9878 Set additional check timeout, but only after a connection has been already
9879 established.
9880
9881 May be used in sections: defaults | frontend | listen | backend
9882 yes | no | yes | yes
9883 Arguments:
9884 <timeout> is the timeout value specified in milliseconds by default, but
9885 can be in any other unit if the number is suffixed by the unit,
9886 as explained at the top of this document.
9887
9888 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
9889 for check and "timeout check" as an additional read timeout. The "min" is
Davor Ocelice9ed2812017-12-25 17:49:28 +01009890 used so that people running with *very* long "timeout connect" (e.g. those
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009891 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01009892 (Please also note that there is no valid reason to have such long connect
9893 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
9894 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009895
9896 If "timeout check" is not set haproxy uses "inter" for complete check
9897 timeout (connect + read) exactly like all <1.3.15 version.
9898
9899 In most cases check request is much simpler and faster to handle than normal
9900 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01009901 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009902
9903 This parameter is specific to backends, but can be specified once for all in
9904 "defaults" sections. This is in fact one of the easiest solutions not to
9905 forget about it.
9906
Willy Tarreau41a340d2008-01-22 12:25:31 +01009907 See also: "timeout connect", "timeout queue", "timeout server",
9908 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009909
9910
Willy Tarreau0ba27502007-12-24 16:55:16 +01009911timeout client <timeout>
9912timeout clitimeout <timeout> (deprecated)
9913 Set the maximum inactivity time on the client side.
9914 May be used in sections : defaults | frontend | listen | backend
9915 yes | yes | yes | no
9916 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009917 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009918 can be in any other unit if the number is suffixed by the unit,
9919 as explained at the top of this document.
9920
9921 The inactivity timeout applies when the client is expected to acknowledge or
9922 send data. In HTTP mode, this timeout is particularly important to consider
9923 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +01009924 response while it is reading data sent by the server. That said, for the
9925 first phase, it is preferable to set the "timeout http-request" to better
9926 protect HAProxy from Slowloris like attacks. The value is specified in
9927 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +01009928 suffixed by the unit, as specified at the top of this document. In TCP mode
9929 (and to a lesser extent, in HTTP mode), it is highly recommended that the
9930 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009931 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01009932 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +01009933 (e.g. 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
9934 sessions (e.g. WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +02009935 which overrides "timeout client" and "timeout server" for tunnels, as well as
9936 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009937
9938 This parameter is specific to frontends, but can be specified once for all in
9939 "defaults" sections. This is in fact one of the easiest solutions not to
9940 forget about it. An unspecified timeout results in an infinite timeout, which
9941 is not recommended. Such a usage is accepted and works but reports a warning
9942 during startup because it may results in accumulation of expired sessions in
9943 the system if the system's timeouts are not configured either.
9944
Willy Tarreau95c4e142017-11-26 12:18:55 +01009945 This also applies to HTTP/2 connections, which will be closed with GOAWAY.
Lukas Tribus75df9d72017-11-24 19:05:12 +01009946
Willy Tarreau0ba27502007-12-24 16:55:16 +01009947 This parameter replaces the old, deprecated "clitimeout". It is recommended
9948 to use it to write new configurations. The form "timeout clitimeout" is
9949 provided only by backwards compatibility but its use is strongly discouraged.
9950
Baptiste Assmann2e1941e2016-03-06 23:24:12 +01009951 See also : "clitimeout", "timeout server", "timeout tunnel",
9952 "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009953
9954
Willy Tarreau05cdd962014-05-10 14:30:07 +02009955timeout client-fin <timeout>
9956 Set the inactivity timeout on the client side for half-closed connections.
9957 May be used in sections : defaults | frontend | listen | backend
9958 yes | yes | yes | no
9959 Arguments :
9960 <timeout> is the timeout value specified in milliseconds by default, but
9961 can be in any other unit if the number is suffixed by the unit,
9962 as explained at the top of this document.
9963
9964 The inactivity timeout applies when the client is expected to acknowledge or
9965 send data while one direction is already shut down. This timeout is different
9966 from "timeout client" in that it only applies to connections which are closed
9967 in one direction. This is particularly useful to avoid keeping connections in
9968 FIN_WAIT state for too long when clients do not disconnect cleanly. This
9969 problem is particularly common long connections such as RDP or WebSocket.
9970 Note that this timeout can override "timeout tunnel" when a connection shuts
Willy Tarreau599391a2017-11-24 10:16:00 +01009971 down in one direction. It is applied to idle HTTP/2 connections once a GOAWAY
9972 frame was sent, often indicating an expectation that the connection quickly
9973 ends.
Willy Tarreau05cdd962014-05-10 14:30:07 +02009974
9975 This parameter is specific to frontends, but can be specified once for all in
9976 "defaults" sections. By default it is not set, so half-closed connections
9977 will use the other timeouts (timeout.client or timeout.tunnel).
9978
9979 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
9980
9981
Willy Tarreau0ba27502007-12-24 16:55:16 +01009982timeout connect <timeout>
9983timeout contimeout <timeout> (deprecated)
9984 Set the maximum time to wait for a connection attempt to a server to succeed.
9985 May be used in sections : defaults | frontend | listen | backend
9986 yes | no | yes | yes
9987 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009988 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009989 can be in any other unit if the number is suffixed by the unit,
9990 as explained at the top of this document.
9991
9992 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009993 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01009994 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +01009995 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009996 connect timeout also presets both queue and tarpit timeouts to the same value
9997 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009998
9999 This parameter is specific to backends, but can be specified once for all in
10000 "defaults" sections. This is in fact one of the easiest solutions not to
10001 forget about it. An unspecified timeout results in an infinite timeout, which
10002 is not recommended. Such a usage is accepted and works but reports a warning
10003 during startup because it may results in accumulation of failed sessions in
10004 the system if the system's timeouts are not configured either.
10005
10006 This parameter replaces the old, deprecated "contimeout". It is recommended
10007 to use it to write new configurations. The form "timeout contimeout" is
10008 provided only by backwards compatibility but its use is strongly discouraged.
10009
Willy Tarreau41a340d2008-01-22 12:25:31 +010010010 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
10011 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010012
10013
Willy Tarreaub16a5742010-01-10 14:46:16 +010010014timeout http-keep-alive <timeout>
10015 Set the maximum allowed time to wait for a new HTTP request to appear
10016 May be used in sections : defaults | frontend | listen | backend
10017 yes | yes | yes | yes
10018 Arguments :
10019 <timeout> is the timeout value specified in milliseconds by default, but
10020 can be in any other unit if the number is suffixed by the unit,
10021 as explained at the top of this document.
10022
10023 By default, the time to wait for a new request in case of keep-alive is set
10024 by "timeout http-request". However this is not always convenient because some
10025 people want very short keep-alive timeouts in order to release connections
10026 faster, and others prefer to have larger ones but still have short timeouts
10027 once the request has started to present itself.
10028
10029 The "http-keep-alive" timeout covers these needs. It will define how long to
10030 wait for a new HTTP request to start coming after a response was sent. Once
10031 the first byte of request has been seen, the "http-request" timeout is used
10032 to wait for the complete request to come. Note that empty lines prior to a
10033 new request do not refresh the timeout and are not counted as a new request.
10034
10035 There is also another difference between the two timeouts : when a connection
10036 expires during timeout http-keep-alive, no error is returned, the connection
10037 just closes. If the connection expires in "http-request" while waiting for a
10038 connection to complete, a HTTP 408 error is returned.
10039
10040 In general it is optimal to set this value to a few tens to hundreds of
10041 milliseconds, to allow users to fetch all objects of a page at once but
Davor Ocelice9ed2812017-12-25 17:49:28 +010010042 without waiting for further clicks. Also, if set to a very small value (e.g.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010043 1 millisecond) it will probably only accept pipelined requests but not the
10044 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +020010045 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010046
10047 If this parameter is not set, the "http-request" timeout applies, and if both
10048 are not set, "timeout client" still applies at the lower level. It should be
10049 set in the frontend to take effect, unless the frontend is in TCP mode, in
10050 which case the HTTP backend's timeout will be used.
10051
Willy Tarreau95c4e142017-11-26 12:18:55 +010010052 When using HTTP/2 "timeout client" is applied instead. This is so we can keep
10053 using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2
Lukas Tribus75df9d72017-11-24 19:05:12 +010010054 (where we only have one connection per client and a connection setup).
10055
Willy Tarreaub16a5742010-01-10 14:46:16 +010010056 See also : "timeout http-request", "timeout client".
10057
10058
Willy Tarreau036fae02008-01-06 13:24:40 +010010059timeout http-request <timeout>
10060 Set the maximum allowed time to wait for a complete HTTP request
10061 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010062 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +010010063 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010064 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +010010065 can be in any other unit if the number is suffixed by the unit,
10066 as explained at the top of this document.
10067
10068 In order to offer DoS protection, it may be required to lower the maximum
10069 accepted time to receive a complete HTTP request without affecting the client
10070 timeout. This helps protecting against established connections on which
10071 nothing is sent. The client timeout cannot offer a good protection against
10072 this abuse because it is an inactivity timeout, which means that if the
10073 attacker sends one character every now and then, the timeout will not
10074 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +020010075 types, the request will be aborted if it does not complete in time. When the
10076 timeout expires, an HTTP 408 response is sent to the client to inform it
10077 about the problem, and the connection is closed. The logs will report
10078 termination codes "cR". Some recent browsers are having problems with this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010079 standard, well-documented behavior, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +020010080 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
10081 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +010010082
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010083 By default, this timeout only applies to the header part of the request,
10084 and not to any data. As soon as the empty line is received, this timeout is
10085 not used anymore. When combined with "option http-buffer-request", this
10086 timeout also applies to the body of the request..
10087 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +010010088 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +010010089
10090 Generally it is enough to set it to a few seconds, as most clients send the
10091 full request immediately upon connection. Add 3 or more seconds to cover TCP
Davor Ocelice9ed2812017-12-25 17:49:28 +010010092 retransmits but that's all. Setting it to very low values (e.g. 50 ms) will
Willy Tarreau036fae02008-01-06 13:24:40 +010010093 generally work on local networks as long as there are no packet losses. This
10094 will prevent people from sending bare HTTP requests using telnet.
10095
10096 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010097 chunk of the incoming request. It should be set in the frontend to take
10098 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
10099 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +010010100
Willy Tarreau0f228a02015-05-01 15:37:53 +020010101 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010102 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +010010103
Willy Tarreau844e3c52008-01-11 16:28:18 +010010104
10105timeout queue <timeout>
10106 Set the maximum time to wait in the queue for a connection slot to be free
10107 May be used in sections : defaults | frontend | listen | backend
10108 yes | no | yes | yes
10109 Arguments :
10110 <timeout> is the timeout value specified in milliseconds by default, but
10111 can be in any other unit if the number is suffixed by the unit,
10112 as explained at the top of this document.
10113
10114 When a server's maxconn is reached, connections are left pending in a queue
10115 which may be server-specific or global to the backend. In order not to wait
10116 indefinitely, a timeout is applied to requests pending in the queue. If the
10117 timeout is reached, it is considered that the request will almost never be
10118 served, so it is dropped and a 503 error is returned to the client.
10119
10120 The "timeout queue" statement allows to fix the maximum time for a request to
10121 be left pending in a queue. If unspecified, the same value as the backend's
10122 connection timeout ("timeout connect") is used, for backwards compatibility
10123 with older versions with no "timeout queue" parameter.
10124
10125 See also : "timeout connect", "contimeout".
10126
10127
10128timeout server <timeout>
10129timeout srvtimeout <timeout> (deprecated)
10130 Set the maximum inactivity time on the server side.
10131 May be used in sections : defaults | frontend | listen | backend
10132 yes | no | yes | yes
10133 Arguments :
10134 <timeout> is the timeout value specified in milliseconds by default, but
10135 can be in any other unit if the number is suffixed by the unit,
10136 as explained at the top of this document.
10137
10138 The inactivity timeout applies when the server is expected to acknowledge or
10139 send data. In HTTP mode, this timeout is particularly important to consider
10140 during the first phase of the server's response, when it has to send the
10141 headers, as it directly represents the server's processing time for the
10142 request. To find out what value to put there, it's often good to start with
10143 what would be considered as unacceptable response times, then check the logs
10144 to observe the response time distribution, and adjust the value accordingly.
10145
10146 The value is specified in milliseconds by default, but can be in any other
10147 unit if the number is suffixed by the unit, as specified at the top of this
10148 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
10149 recommended that the client timeout remains equal to the server timeout in
10150 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010151 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +010010152 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +010010153 seconds (e.g. 4 or 5 seconds minimum). If some long-lived sessions are mixed
10154 with short-lived sessions (e.g. WebSocket and HTTP), it's worth considering
Willy Tarreauce887fd2012-05-12 12:50:00 +020010155 "timeout tunnel", which overrides "timeout client" and "timeout server" for
10156 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010157
10158 This parameter is specific to backends, but can be specified once for all in
10159 "defaults" sections. This is in fact one of the easiest solutions not to
10160 forget about it. An unspecified timeout results in an infinite timeout, which
10161 is not recommended. Such a usage is accepted and works but reports a warning
10162 during startup because it may results in accumulation of expired sessions in
10163 the system if the system's timeouts are not configured either.
10164
10165 This parameter replaces the old, deprecated "srvtimeout". It is recommended
10166 to use it to write new configurations. The form "timeout srvtimeout" is
10167 provided only by backwards compatibility but its use is strongly discouraged.
10168
Willy Tarreauce887fd2012-05-12 12:50:00 +020010169 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010170
Willy Tarreau05cdd962014-05-10 14:30:07 +020010171
10172timeout server-fin <timeout>
10173 Set the inactivity timeout on the server side for half-closed connections.
10174 May be used in sections : defaults | frontend | listen | backend
10175 yes | no | yes | yes
10176 Arguments :
10177 <timeout> is the timeout value specified in milliseconds by default, but
10178 can be in any other unit if the number is suffixed by the unit,
10179 as explained at the top of this document.
10180
10181 The inactivity timeout applies when the server is expected to acknowledge or
10182 send data while one direction is already shut down. This timeout is different
10183 from "timeout server" in that it only applies to connections which are closed
10184 in one direction. This is particularly useful to avoid keeping connections in
10185 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
10186 This problem is particularly common long connections such as RDP or WebSocket.
10187 Note that this timeout can override "timeout tunnel" when a connection shuts
10188 down in one direction. This setting was provided for completeness, but in most
10189 situations, it should not be needed.
10190
10191 This parameter is specific to backends, but can be specified once for all in
10192 "defaults" sections. By default it is not set, so half-closed connections
10193 will use the other timeouts (timeout.server or timeout.tunnel).
10194
10195 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
10196
Willy Tarreau844e3c52008-01-11 16:28:18 +010010197
10198timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +010010199 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +010010200 May be used in sections : defaults | frontend | listen | backend
10201 yes | yes | yes | yes
10202 Arguments :
10203 <timeout> is the tarpit duration specified in milliseconds by default, but
10204 can be in any other unit if the number is suffixed by the unit,
10205 as explained at the top of this document.
10206
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010207 When a connection is tarpitted using "http-request tarpit" or
10208 "reqtarpit", it is maintained open with no activity for a certain
10209 amount of time, then closed. "timeout tarpit" defines how long it will
10210 be maintained open.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010211
10212 The value is specified in milliseconds by default, but can be in any other
10213 unit if the number is suffixed by the unit, as specified at the top of this
10214 document. If unspecified, the same value as the backend's connection timeout
10215 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +010010216 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010217
10218 See also : "timeout connect", "contimeout".
10219
10220
Willy Tarreauce887fd2012-05-12 12:50:00 +020010221timeout tunnel <timeout>
10222 Set the maximum inactivity time on the client and server side for tunnels.
10223 May be used in sections : defaults | frontend | listen | backend
10224 yes | no | yes | yes
10225 Arguments :
10226 <timeout> is the timeout value specified in milliseconds by default, but
10227 can be in any other unit if the number is suffixed by the unit,
10228 as explained at the top of this document.
10229
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010230 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +020010231 between a client and a server, and the connection remains inactive in both
10232 directions. This timeout supersedes both the client and server timeouts once
10233 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
Davor Ocelice9ed2812017-12-25 17:49:28 +010010234 analyzer remains attached to either connection (e.g. tcp content rules are
10235 accepted). In HTTP, this timeout is used when a connection is upgraded (e.g.
Willy Tarreauce887fd2012-05-12 12:50:00 +020010236 when switching to the WebSocket protocol, or forwarding a CONNECT request
10237 to a proxy), or after the first response when no keepalive/close option is
10238 specified.
10239
Willy Tarreau05cdd962014-05-10 14:30:07 +020010240 Since this timeout is usually used in conjunction with long-lived connections,
10241 it usually is a good idea to also set "timeout client-fin" to handle the
10242 situation where a client suddenly disappears from the net and does not
10243 acknowledge a close, or sends a shutdown and does not acknowledge pending
10244 data anymore. This can happen in lossy networks where firewalls are present,
10245 and is detected by the presence of large amounts of sessions in a FIN_WAIT
10246 state.
10247
Willy Tarreauce887fd2012-05-12 12:50:00 +020010248 The value is specified in milliseconds by default, but can be in any other
10249 unit if the number is suffixed by the unit, as specified at the top of this
10250 document. Whatever the expected normal idle time, it is a good practice to
10251 cover at least one or several TCP packet losses by specifying timeouts that
Davor Ocelice9ed2812017-12-25 17:49:28 +010010252 are slightly above multiples of 3 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreauce887fd2012-05-12 12:50:00 +020010253
10254 This parameter is specific to backends, but can be specified once for all in
10255 "defaults" sections. This is in fact one of the easiest solutions not to
10256 forget about it.
10257
10258 Example :
10259 defaults http
10260 option http-server-close
10261 timeout connect 5s
10262 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +020010263 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +020010264 timeout server 30s
10265 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
10266
Willy Tarreau05cdd962014-05-10 14:30:07 +020010267 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +020010268
10269
Willy Tarreau844e3c52008-01-11 16:28:18 +010010270transparent (deprecated)
10271 Enable client-side transparent proxying
10272 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +010010273 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +010010274 Arguments : none
10275
10276 This keyword was introduced in order to provide layer 7 persistence to layer
10277 3 load balancers. The idea is to use the OS's ability to redirect an incoming
10278 connection for a remote address to a local process (here HAProxy), and let
10279 this process know what address was initially requested. When this option is
10280 used, sessions without cookies will be forwarded to the original destination
10281 IP address of the incoming request (which should match that of another
10282 equipment), while requests with cookies will still be forwarded to the
10283 appropriate server.
10284
10285 The "transparent" keyword is deprecated, use "option transparent" instead.
10286
10287 Note that contrary to a common belief, this option does NOT make HAProxy
10288 present the client's IP to the server when establishing the connection.
10289
Willy Tarreau844e3c52008-01-11 16:28:18 +010010290 See also: "option transparent"
10291
William Lallemanda73203e2012-03-12 12:48:57 +010010292unique-id-format <string>
10293 Generate a unique ID for each request.
10294 May be used in sections : defaults | frontend | listen | backend
10295 yes | yes | yes | no
10296 Arguments :
10297 <string> is a log-format string.
10298
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010299 This keyword creates a ID for each request using the custom log format. A
10300 unique ID is useful to trace a request passing through many components of
10301 a complex infrastructure. The newly created ID may also be logged using the
10302 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +010010303
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010304 The format should be composed from elements that are guaranteed to be
10305 unique when combined together. For instance, if multiple haproxy instances
10306 are involved, it might be important to include the node name. It is often
10307 needed to log the incoming connection's source and destination addresses
10308 and ports. Note that since multiple requests may be performed over the same
10309 connection, including a request counter may help differentiate them.
10310 Similarly, a timestamp may protect against a rollover of the counter.
10311 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +010010312
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010313 It is recommended to use hexadecimal notation for many fields since it
10314 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +010010315
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010316 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010317
Julien Vehentf21be322014-03-07 08:27:34 -050010318 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010319
10320 will generate:
10321
10322 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10323
10324 See also: "unique-id-header"
10325
10326unique-id-header <name>
10327 Add a unique ID header in the HTTP request.
10328 May be used in sections : defaults | frontend | listen | backend
10329 yes | yes | yes | no
10330 Arguments :
10331 <name> is the name of the header.
10332
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010333 Add a unique-id header in the HTTP request sent to the server, using the
10334 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +010010335
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010336 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010337
Julien Vehentf21be322014-03-07 08:27:34 -050010338 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010339 unique-id-header X-Unique-ID
10340
10341 will generate:
10342
10343 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10344
10345 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +010010346
Willy Tarreauf51658d2014-04-23 01:21:56 +020010347use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010348 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010349 May be used in sections : defaults | frontend | listen | backend
10350 no | yes | yes | no
10351 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010352 <backend> is the name of a valid backend or "listen" section, or a
10353 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010354
Willy Tarreauf51658d2014-04-23 01:21:56 +020010355 <condition> is a condition composed of ACLs, as described in section 7. If
10356 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010357
10358 When doing content-switching, connections arrive on a frontend and are then
10359 dispatched to various backends depending on a number of conditions. The
10360 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010361 "use_backend" keyword. While it is normally used with HTTP processing, it can
Davor Ocelice9ed2812017-12-25 17:49:28 +010010362 also be used in pure TCP, either without content using stateless ACLs (e.g.
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010363 source address validation) or combined with a "tcp-request" rule to wait for
10364 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010365
10366 There may be as many "use_backend" rules as desired. All of these rules are
10367 evaluated in their declaration order, and the first one which matches will
10368 assign the backend.
10369
10370 In the first form, the backend will be used if the condition is met. In the
10371 second form, the backend will be used if the condition is not met. If no
10372 condition is valid, the backend defined with "default_backend" will be used.
10373 If no default backend is defined, either the servers in the same section are
10374 used (in case of a "listen" section) or, in case of a frontend, no server is
10375 used and a 503 service unavailable response is returned.
10376
Willy Tarreau51aecc72009-07-12 09:47:04 +020010377 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010378 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +020010379 and backend processing will immediately follow, or the backend will wait for
10380 a complete HTTP request to get in. This feature is useful when a frontend
10381 must decode several protocols on a unique port, one of them being HTTP.
10382
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010383 When <backend> is a simple name, it is resolved at configuration time, and an
10384 error is reported if the specified backend does not exist. If <backend> is
10385 a log-format string instead, no check may be done at configuration time, so
10386 the backend name is resolved dynamically at run time. If the resulting
10387 backend name does not correspond to any valid backend, no other rule is
10388 evaluated, and the default_backend directive is applied instead. Note that
10389 when using dynamic backend names, it is highly recommended to use a prefix
10390 that no other backend uses in order to ensure that an unauthorized backend
10391 cannot be forced from the request.
10392
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010393 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010394 used to detect the association between frontends and backends to compute the
10395 backend's "fullconn" setting. This cannot be done for dynamic names.
10396
10397 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
10398 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +010010399
Willy Tarreau036fae02008-01-06 13:24:40 +010010400
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010401use-server <server> if <condition>
10402use-server <server> unless <condition>
10403 Only use a specific server if/unless an ACL-based condition is matched.
10404 May be used in sections : defaults | frontend | listen | backend
10405 no | no | yes | yes
10406 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010407 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010408
10409 <condition> is a condition composed of ACLs, as described in section 7.
10410
10411 By default, connections which arrive to a backend are load-balanced across
10412 the available servers according to the configured algorithm, unless a
10413 persistence mechanism such as a cookie is used and found in the request.
10414
10415 Sometimes it is desirable to forward a particular request to a specific
10416 server without having to declare a dedicated backend for this server. This
10417 can be achieved using the "use-server" rules. These rules are evaluated after
10418 the "redirect" rules and before evaluating cookies, and they have precedence
10419 on them. There may be as many "use-server" rules as desired. All of these
10420 rules are evaluated in their declaration order, and the first one which
10421 matches will assign the server.
10422
10423 If a rule designates a server which is down, and "option persist" is not used
10424 and no force-persist rule was validated, it is ignored and evaluation goes on
10425 with the next rules until one matches.
10426
10427 In the first form, the server will be used if the condition is met. In the
10428 second form, the server will be used if the condition is not met. If no
10429 condition is valid, the processing continues and the server will be assigned
10430 according to other persistence mechanisms.
10431
10432 Note that even if a rule is matched, cookie processing is still performed but
10433 does not assign the server. This allows prefixed cookies to have their prefix
10434 stripped.
10435
10436 The "use-server" statement works both in HTTP and TCP mode. This makes it
10437 suitable for use with content-based inspection. For instance, a server could
10438 be selected in a farm according to the TLS SNI field. And if these servers
10439 have their weight set to zero, they will not be used for other traffic.
10440
10441 Example :
10442 # intercept incoming TLS requests based on the SNI field
10443 use-server www if { req_ssl_sni -i www.example.com }
10444 server www 192.168.0.1:443 weight 0
10445 use-server mail if { req_ssl_sni -i mail.example.com }
10446 server mail 192.168.0.1:587 weight 0
10447 use-server imap if { req_ssl_sni -i imap.example.com }
Lukas Tribus98a3e3f2017-03-26 12:55:35 +000010448 server imap 192.168.0.1:993 weight 0
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010449 # all the rest is forwarded to this server
10450 server default 192.168.0.2:443 check
10451
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010452 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010453
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010454
Davor Ocelice9ed2812017-12-25 17:49:28 +0100104555. Bind and server options
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010456--------------------------
10457
10458The "bind", "server" and "default-server" keywords support a number of settings
10459depending on some build options and on the system HAProxy was built on. These
10460settings generally each consist in one word sometimes followed by a value,
10461written on the same line as the "bind" or "server" line. All these options are
10462described in this section.
10463
10464
104655.1. Bind options
10466-----------------
10467
10468The "bind" keyword supports a certain number of settings which are all passed
10469as arguments on the same line. The order in which those arguments appear makes
10470no importance, provided that they appear after the bind address. All of these
10471parameters are optional. Some of them consist in a single words (booleans),
10472while other ones expect a value after them. In this case, the value must be
10473provided immediately after the setting name.
10474
10475The currently supported settings are the following ones.
10476
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010010477accept-netscaler-cip <magic number>
10478 Enforces the use of the NetScaler Client IP insertion protocol over any
10479 connection accepted by any of the TCP sockets declared on the same line. The
10480 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
10481 the incoming connection to be used everywhere an address is used, with the
10482 only exception of "tcp-request connection" rules which will only see the
10483 real connection address. Logs will reflect the addresses indicated in the
10484 protocol, unless it is violated, in which case the real address will still
10485 be used. This keyword combined with support from external components can be
10486 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010010487 mechanism which is not always reliable and not even always usable. See also
10488 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
10489 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010010490
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010491accept-proxy
10492 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020010493 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
10494 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010495 3/4 addresses of the incoming connection to be used everywhere an address is
10496 used, with the only exception of "tcp-request connection" rules which will
10497 only see the real connection address. Logs will reflect the addresses
10498 indicated in the protocol, unless it is violated, in which case the real
Davor Ocelice9ed2812017-12-25 17:49:28 +010010499 address will still be used. This keyword combined with support from external
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010500 components can be used as an efficient and reliable alternative to the
10501 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020010502 usable. See also "tcp-request connection expect-proxy" for a finer-grained
10503 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010504
Olivier Houchardc2aae742017-09-22 18:26:28 +020010505allow-0rtt
10506 Allow receiving early data when using TLS 1.3. This is disabled by default,
10507 due to security considerations.
10508
Willy Tarreauab861d32013-04-02 02:30:41 +020010509alpn <protocols>
10510 This enables the TLS ALPN extension and advertises the specified protocol
10511 list as supported on top of ALPN. The protocol list consists in a comma-
10512 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
10513 quotes). This requires that the SSL library is build with support for TLS
10514 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
Willy Tarreau95c4e142017-11-26 12:18:55 +010010515 initial NPN extension. ALPN is required to enable HTTP/2 on an HTTP frontend.
10516 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
10517 now obsolete NPN extension. At the time of writing this, most browsers still
10518 support both ALPN and NPN for HTTP/2 so a fallback to NPN may still work for
10519 a while. But ALPN must be used whenever possible. If both HTTP/2 and HTTP/1.1
10520 are expected to be supported, both versions can be advertised, in order of
10521 preference, like below :
10522
10523 bind :443 ssl crt pub.pem alpn h2,http/1.1
Willy Tarreauab861d32013-04-02 02:30:41 +020010524
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010525backlog <backlog>
10526 Sets the socket's backlog to this value. If unspecified, the frontend's
10527 backlog is used instead, which generally defaults to the maxconn value.
10528
Emmanuel Hocdete7f2b732017-01-09 16:15:54 +010010529curves <curves>
10530 This setting is only available when support for OpenSSL was built in. It sets
10531 the string describing the list of elliptic curves algorithms ("curve suite")
10532 that are negotiated during the SSL/TLS handshake with ECDHE. The format of the
10533 string is a colon-delimited list of curve name.
10534 Example: "X25519:P-256" (without quote)
10535 When "curves" is set, "ecdhe" parameter is ignored.
10536
Emeric Brun7fb34422012-09-28 15:26:15 +020010537ecdhe <named curve>
10538 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010010539 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
10540 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020010541
Emeric Brunfd33a262012-10-11 16:28:27 +020010542ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020010543 This setting is only available when support for OpenSSL was built in. It
10544 designates a PEM file from which to load CA certificates used to verify
10545 client's certificate.
10546
Emeric Brunb6dc9342012-09-28 17:55:37 +020010547ca-ignore-err [all|<errorID>,...]
10548 This setting is only available when support for OpenSSL was built in.
10549 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
10550 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
10551 error is ignored.
10552
Christopher Faulet31af49d2015-06-09 17:29:50 +020010553ca-sign-file <cafile>
10554 This setting is only available when support for OpenSSL was built in. It
10555 designates a PEM file containing both the CA certificate and the CA private
10556 key used to create and sign server's certificates. This is a mandatory
10557 setting when the dynamic generation of certificates is enabled. See
10558 'generate-certificates' for details.
10559
Bertrand Jacquind4d0a232016-11-13 16:37:12 +000010560ca-sign-pass <passphrase>
Christopher Faulet31af49d2015-06-09 17:29:50 +020010561 This setting is only available when support for OpenSSL was built in. It is
10562 the CA private key passphrase. This setting is optional and used only when
10563 the dynamic generation of certificates is enabled. See
10564 'generate-certificates' for details.
10565
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010566ciphers <ciphers>
10567 This setting is only available when support for OpenSSL was built in. It sets
10568 the string describing the list of cipher algorithms ("cipher suite") that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010569 negotiated during the SSL/TLS handshake. The format of the string is defined
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010570 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
10571 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
Davor Ocelice9ed2812017-12-25 17:49:28 +010010572 Depending on the compatibility and security requirements, the list of suitable
Daniel Schneller87e43022017-09-01 19:29:57 +020010573 ciphers depends on a variety of variables. For background information and
10574 recommendations see e. g. (https://wiki.mozilla.org/Security/Server_Side_TLS)
10575 and (https://mozilla.github.io/server-side-tls/ssl-config-generator/).
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010576
Emeric Brunfd33a262012-10-11 16:28:27 +020010577crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020010578 This setting is only available when support for OpenSSL was built in. It
10579 designates a PEM file from which to load certificate revocation list used
10580 to verify client's certificate.
10581
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010582crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000010583 This setting is only available when support for OpenSSL was built in. It
10584 designates a PEM file containing both the required certificates and any
10585 associated private keys. This file can be built by concatenating multiple
10586 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
10587 requires an intermediate certificate, this can also be concatenated into this
10588 file.
10589
10590 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
10591 are loaded.
10592
10593 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +010010594 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010595 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
10596 specified multiple times in order to load certificates from multiple files or
10597 directories. The certificates will be presented to clients who provide a
10598 valid TLS Server Name Indication field matching one of their CN or alt
Davor Ocelice9ed2812017-12-25 17:49:28 +010010599 subjects. Wildcards are supported, where a wildcard character '*' is used
10600 instead of the first hostname component (e.g. *.example.org matches
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010601 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000010602
10603 If no SNI is provided by the client or if the SSL library does not support
10604 TLS extensions, or if the client provides an SNI hostname which does not
10605 match any certificate, then the first loaded certificate will be presented.
10606 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010010607 recommended to load the default one first as a file or to ensure that it will
10608 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000010609
Emeric Brune032bfa2012-09-28 13:01:45 +020010610 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010611
Davor Ocelice9ed2812017-12-25 17:49:28 +010010612 Some CAs (such as GoDaddy) offer a drop down list of server types that do not
Alex Davies0fbf0162013-03-02 16:04:50 +000010613 include HAProxy when obtaining a certificate. If this happens be sure to
Davor Ocelice9ed2812017-12-25 17:49:28 +010010614 choose a web server that the CA believes requires an intermediate CA (for
10615 GoDaddy, selection Apache Tomcat will get the correct bundle, but many
Alex Davies0fbf0162013-03-02 16:04:50 +000010616 others, e.g. nginx, result in a wrong bundle that will not work for some
10617 clients).
10618
Emeric Brun4147b2e2014-06-16 18:36:30 +020010619 For each PEM file, haproxy checks for the presence of file at the same path
10620 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
10621 Status Request extension (also known as "OCSP stapling") is automatically
10622 enabled. The content of this file is optional. If not empty, it must contain
10623 a valid OCSP Response in DER format. In order to be valid an OCSP Response
10624 must comply with the following rules: it has to indicate a good status,
10625 it has to be a single response for the certificate of the PEM file, and it
10626 has to be valid at the moment of addition. If these rules are not respected
10627 the OCSP Response is ignored and a warning is emitted. In order to identify
10628 which certificate an OCSP Response applies to, the issuer's certificate is
10629 necessary. If the issuer's certificate is not found in the PEM file, it will
10630 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
10631 if it exists otherwise it will fail with an error.
10632
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010633 For each PEM file, haproxy also checks for the presence of file at the same
10634 path suffixed by ".sctl". If such file is found, support for Certificate
10635 Transparency (RFC6962) TLS extension is enabled. The file must contain a
10636 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
10637 to check basic syntax, but no signatures are verified.
10638
yanbzhu6c25e9e2016-01-05 12:52:02 -050010639 There are cases where it is desirable to support multiple key types, e.g. RSA
10640 and ECDSA in the cipher suites offered to the clients. This allows clients
10641 that support EC certificates to be able to use EC ciphers, while
10642 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050010643
10644 In order to provide this functionality, multiple PEM files, each with a
10645 different key type, are required. To associate these PEM files into a
10646 "cert bundle" that is recognized by haproxy, they must be named in the
10647 following way: All PEM files that are to be bundled must have the same base
10648 name, with a suffix indicating the key type. Currently, three suffixes are
10649 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
10650 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
10651 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
10652 suffix matters. To load this bundle into haproxy, specify the base name only:
10653
10654 Example : bind :8443 ssl crt example.pem
10655
yanbzhu6c25e9e2016-01-05 12:52:02 -050010656 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -050010657 a cert bundle.
10658
Davor Ocelice9ed2812017-12-25 17:49:28 +010010659 HAProxy will load all PEM files in the bundle at the same time to try to
yanbzhud19630c2015-12-14 15:10:25 -050010660 support multiple key types. PEM files are combined based on Common Name
10661 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
10662 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
10663 entries in the certificates in that bundle, haproxy will not be able to
10664 provide multi-cert support.
10665
10666 Assuming bundle in the example above contained the following:
10667
10668 Filename | CN | SAN
10669 -------------------+-----------------+-------------------
10670 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -050010671 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -050010672 example.pem.ecdsa | www.example.com | ecdsa.example.com
10673 -------------------+-----------------+-------------------
10674
10675 Users connecting with an SNI of "www.example.com" will be able
10676 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
10677 "rsa.example.com" will only be able to use RSA cipher suites, and users
10678 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020010679 suites. With BoringSSL and Openssl >= 1.1.1 multi-cert is natively supported,
10680 no need to bundle certificates. ECDSA certificate will be preferred if client
10681 support it.
yanbzhud19630c2015-12-14 15:10:25 -050010682
10683 If a directory name is given as the <cert> argument, haproxy will
10684 automatically search and load bundled files in that directory.
10685
10686 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
10687 bundling. Each certificate can have its own .ocsp and .issuer file. At this
10688 time, sctl is not supported in multi-certificate bundling.
10689
Emeric Brunb6dc9342012-09-28 17:55:37 +020010690crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000010691 This setting is only available when support for OpenSSL was built in. Sets a
Davor Ocelice9ed2812017-12-25 17:49:28 +010010692 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010693 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000010694 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020010695
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010696crt-list <file>
10697 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010698 designates a list of PEM file with an optional ssl configuration and a SNI
10699 filter per certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010700
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010701 <crtfile> [\[<sslbindconf> ...\]] [[!]<snifilter> ...]
10702
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020010703 sslbindconf support "npn", "alpn", "verify", "ca-file", "no-ca-names",
10704 crl-file", "ecdhe", "curves", "ciphers" configuration. With BoringSSL
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020010705 and Openssl >= 1.1.1 "ssl-min-ver" and "ssl-max-ver" are also supported.
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010706 It override the configuration set in bind line for the certificate.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010707
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020010708 Wildcards are supported in the SNI filter. Negative filter are also supported,
10709 only useful in combination with a wildcard filter to exclude a particular SNI.
10710 The certificates will be presented to clients who provide a valid TLS Server
10711 Name Indication field matching one of the SNI filters. If no SNI filter is
10712 specified, the CN and alt subjects are used. This directive may be specified
10713 multiple times. See the "crt" option for more information. The default
10714 certificate is still needed to meet OpenSSL expectations. If it is not used,
10715 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010716
yanbzhu6c25e9e2016-01-05 12:52:02 -050010717 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +020010718 the base name is given in the crt-list. SNI filter will do the same work on
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020010719 all bundled certificates. With BoringSSL and Openssl >= 1.1.1 multi-cert is
10720 natively supported, avoid multi-cert bundling. RSA and ECDSA certificates can
10721 be declared in a row, and set different ssl and filter parameter.
yanbzhud19630c2015-12-14 15:10:25 -050010722
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010723 crt-list file example:
10724 cert1.pem
Emmanuel Hocdet05942112017-02-20 16:11:50 +010010725 cert2.pem [alpn h2,http/1.1]
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010726 certW.pem *.domain.tld !secure.domain.tld
Emmanuel Hocdet05942112017-02-20 16:11:50 +010010727 certS.pem [curves X25519:P-256 ciphers ECDHE-ECDSA-AES256-GCM-SHA384] secure.domain.tld
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010728
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010729defer-accept
10730 Is an optional keyword which is supported only on certain Linux kernels. It
10731 states that a connection will only be accepted once some data arrive on it,
10732 or at worst after the first retransmit. This should be used only on protocols
Davor Ocelice9ed2812017-12-25 17:49:28 +010010733 for which the client talks first (e.g. HTTP). It can slightly improve
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010734 performance by ensuring that most of the request is already available when
10735 the connection is accepted. On the other hand, it will not be able to detect
10736 connections which don't talk. It is important to note that this option is
10737 broken in all kernels up to 2.6.31, as the connection is never accepted until
10738 the client talks. This can cause issues with front firewalls which would see
10739 an established connection while the proxy will only see it in SYN_RECV. This
10740 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
10741
William Lallemandf6975e92017-05-26 17:42:10 +020010742expose-fd listeners
10743 This option is only usable with the stats socket. It gives your stats socket
10744 the capability to pass listeners FD to another HAProxy process.
William Lallemande202b1e2017-06-01 17:38:56 +020010745 During a reload with the master-worker mode, the process is automatically
10746 reexecuted adding -x and one of the stats socket with this option.
Davor Ocelice9ed2812017-12-25 17:49:28 +010010747 See also "-x" in the management guide.
William Lallemandf6975e92017-05-26 17:42:10 +020010748
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010749force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010750 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010751 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010752 for high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010753 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010754
10755force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010756 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010757 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010758 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010759
10760force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010761 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010762 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010763 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010764
10765force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010766 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010767 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010768 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010769
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020010770force-tlsv13
10771 This option enforces use of TLSv1.3 only on SSL connections instantiated from
10772 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010773 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020010774
Christopher Faulet31af49d2015-06-09 17:29:50 +020010775generate-certificates
10776 This setting is only available when support for OpenSSL was built in. It
10777 enables the dynamic SSL certificates generation. A CA certificate and its
10778 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
10779 a transparent forward proxy, SSL requests generate errors because of a common
10780 name mismatch on the certificate presented to the client. With this option
10781 enabled, HAProxy will try to forge a certificate using the SNI hostname
10782 indicated by the client. This is done only if no certificate matches the SNI
10783 hostname (see 'crt-list'). If an error occurs, the default certificate is
10784 used, else the 'strict-sni' option is set.
10785 It can also be used when HAProxy is configured as a reverse proxy to ease the
10786 deployment of an architecture with many backends.
10787
10788 Creating a SSL certificate is an expensive operation, so a LRU cache is used
10789 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
Davor Ocelice9ed2812017-12-25 17:49:28 +010010790 increases the HAProxy's memory footprint to reduce latency when the same
Christopher Faulet31af49d2015-06-09 17:29:50 +020010791 certificate is used many times.
10792
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010793gid <gid>
10794 Sets the group of the UNIX sockets to the designated system gid. It can also
10795 be set by default in the global section's "unix-bind" statement. Note that
10796 some platforms simply ignore this. This setting is equivalent to the "group"
10797 setting except that the group ID is used instead of its name. This setting is
10798 ignored by non UNIX sockets.
10799
10800group <group>
10801 Sets the group of the UNIX sockets to the designated system group. It can
10802 also be set by default in the global section's "unix-bind" statement. Note
10803 that some platforms simply ignore this. This setting is equivalent to the
10804 "gid" setting except that the group name is used instead of its gid. This
10805 setting is ignored by non UNIX sockets.
10806
10807id <id>
10808 Fixes the socket ID. By default, socket IDs are automatically assigned, but
10809 sometimes it is more convenient to fix them to ease monitoring. This value
10810 must be strictly positive and unique within the listener/frontend. This
10811 option can only be used when defining only a single socket.
10812
10813interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010010814 Restricts the socket to a specific interface. When specified, only packets
10815 received from that particular interface are processed by the socket. This is
10816 currently only supported on Linux. The interface must be a primary system
10817 interface, not an aliased interface. It is also possible to bind multiple
10818 frontends to the same address if they are bound to different interfaces. Note
10819 that binding to a network interface requires root privileges. This parameter
Jérôme Magnin61275192018-02-07 11:39:58 +010010820 is only compatible with TCPv4/TCPv6 sockets. When specified, return traffic
10821 uses the same interface as inbound traffic, and its associated routing table,
10822 even if there are explicit routes through different interfaces configured.
10823 This can prove useful to address asymmetric routing issues when the same
10824 client IP addresses need to be able to reach frontends hosted on different
10825 interfaces.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010826
Willy Tarreauabb175f2012-09-24 12:43:26 +020010827level <level>
10828 This setting is used with the stats sockets only to restrict the nature of
10829 the commands that can be issued on the socket. It is ignored by other
10830 sockets. <level> can be one of :
Davor Ocelice9ed2812017-12-25 17:49:28 +010010831 - "user" is the least privileged level; only non-sensitive stats can be
Willy Tarreauabb175f2012-09-24 12:43:26 +020010832 read, and no change is allowed. It would make sense on systems where it
10833 is not easy to restrict access to the socket.
10834 - "operator" is the default level and fits most common uses. All data can
Davor Ocelice9ed2812017-12-25 17:49:28 +010010835 be read, and only non-sensitive changes are permitted (e.g. clear max
Willy Tarreauabb175f2012-09-24 12:43:26 +020010836 counters).
Davor Ocelice9ed2812017-12-25 17:49:28 +010010837 - "admin" should be used with care, as everything is permitted (e.g. clear
Willy Tarreauabb175f2012-09-24 12:43:26 +020010838 all counters).
10839
Andjelko Iharosc4df59e2017-07-20 11:59:48 +020010840severity-output <format>
10841 This setting is used with the stats sockets only to configure severity
10842 level output prepended to informational feedback messages. Severity
10843 level of messages can range between 0 and 7, conforming to syslog
10844 rfc5424. Valid and successful socket commands requesting data
10845 (i.e. "show map", "get acl foo" etc.) will never have a severity level
10846 prepended. It is ignored by other sockets. <format> can be one of :
10847 - "none" (default) no severity level is prepended to feedback messages.
10848 - "number" severity level is prepended as a number.
10849 - "string" severity level is prepended as a string following the
10850 rfc5424 convention.
10851
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010852maxconn <maxconn>
10853 Limits the sockets to this number of concurrent connections. Extraneous
10854 connections will remain in the system's backlog until a connection is
10855 released. If unspecified, the limit will be the same as the frontend's
10856 maxconn. Note that in case of port ranges or multiple addresses, the same
10857 value will be applied to each socket. This setting enables different
10858 limitations on expensive sockets, for instance SSL entries which may easily
10859 eat all memory.
10860
10861mode <mode>
10862 Sets the octal mode used to define access permissions on the UNIX socket. It
10863 can also be set by default in the global section's "unix-bind" statement.
10864 Note that some platforms simply ignore this. This setting is ignored by non
10865 UNIX sockets.
10866
10867mss <maxseg>
10868 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
10869 connections. This can be used to force a lower MSS for certain specific
10870 ports, for instance for connections passing through a VPN. Note that this
10871 relies on a kernel feature which is theoretically supported under Linux but
10872 was buggy in all versions prior to 2.6.28. It may or may not work on other
10873 operating systems. It may also not change the advertised value but change the
10874 effective size of outgoing segments. The commonly advertised value for TCPv4
10875 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
10876 positive, it will be used as the advertised MSS. If it is negative, it will
10877 indicate by how much to reduce the incoming connection's advertised MSS for
10878 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
10879
10880name <name>
10881 Sets an optional name for these sockets, which will be reported on the stats
10882 page.
10883
Willy Tarreaud72f0f32015-10-13 14:50:22 +020010884namespace <name>
10885 On Linux, it is possible to specify which network namespace a socket will
10886 belong to. This directive makes it possible to explicitly bind a listener to
10887 a namespace different from the default one. Please refer to your operating
10888 system's documentation to find more details about network namespaces.
10889
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010890nice <nice>
10891 Sets the 'niceness' of connections initiated from the socket. Value must be
10892 in the range -1024..1024 inclusive, and defaults to zero. Positive values
10893 means that such connections are more friendly to others and easily offer
10894 their place in the scheduler. On the opposite, negative values mean that
10895 connections want to run with a higher priority than others. The difference
10896 only happens under high loads when the system is close to saturation.
10897 Negative values are appropriate for low-latency or administration services,
10898 and high values are generally recommended for CPU intensive tasks such as SSL
10899 processing or bulk transfers which are less sensible to latency. For example,
10900 it may make sense to use a positive value for an SMTP socket and a negative
10901 one for an RDP socket.
10902
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020010903no-ca-names
10904 This setting is only available when support for OpenSSL was built in. It
10905 prevents from send CA names in server hello message when ca-file is used.
10906
Emeric Brun9b3009b2012-10-05 11:55:06 +020010907no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010908 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010909 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010910 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010911 be enabled using any configuration option. This option is also available on
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010912 global statement "ssl-default-bind-options". Use "ssl-min-ver" and
10913 "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010914
Emeric Brun90ad8722012-10-02 14:00:59 +020010915no-tls-tickets
10916 This setting is only available when support for OpenSSL was built in. It
10917 disables the stateless session resumption (RFC 5077 TLS Ticket
10918 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010919 session resumption is more expensive in CPU usage. This option is also
10920 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020010921
Emeric Brun9b3009b2012-10-05 11:55:06 +020010922no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010923 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010924 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010925 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010926 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010927 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
10928 and "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010929
Emeric Brun9b3009b2012-10-05 11:55:06 +020010930no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020010931 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010932 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010933 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010934 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010935 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
10936 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020010937
Emeric Brun9b3009b2012-10-05 11:55:06 +020010938no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020010939 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010940 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010941 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010942 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010943 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
10944 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020010945
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020010946no-tlsv13
10947 This setting is only available when support for OpenSSL was built in. It
10948 disables support for TLSv1.3 on any sockets instantiated from the listener
10949 when SSL is supported. Note that SSLv2 is forced disabled in the code and
10950 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010951 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
10952 and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020010953
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020010954npn <protocols>
10955 This enables the NPN TLS extension and advertises the specified protocol list
10956 as supported on top of NPN. The protocol list consists in a comma-delimited
10957 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
10958 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020010959 enabled (check with haproxy -vv). Note that the NPN extension has been
Willy Tarreau95c4e142017-11-26 12:18:55 +010010960 replaced with the ALPN extension (see the "alpn" keyword), though this one is
10961 only available starting with OpenSSL 1.0.2. If HTTP/2 is desired on an older
10962 version of OpenSSL, NPN might still be used as most clients still support it
10963 at the time of writing this. It is possible to enable both NPN and ALPN
10964 though it probably doesn't make any sense out of testing.
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020010965
Lukas Tribus53ae85c2017-05-04 15:45:40 +000010966prefer-client-ciphers
10967 Use the client's preference when selecting the cipher suite, by default
10968 the server's preference is enforced. This option is also available on
10969 global statement "ssl-default-bind-options".
Lukas Tribus926594f2018-05-18 17:55:57 +020010970 Note that with OpenSSL >= 1.1.1 ChaCha20-Poly1305 is reprioritized anyway
10971 (without setting this option), if a ChaCha20-Poly1305 cipher is at the top of
10972 the client cipher list.
Lukas Tribus53ae85c2017-05-04 15:45:40 +000010973
Christopher Fauletc644fa92017-11-23 22:44:11 +010010974process <process-set>[/<thread-set>]
10975 This restricts the list of processes and/or threads on which this listener is
10976 allowed to run. It does not enforce any process but eliminates those which do
Davor Ocelice9ed2812017-12-25 17:49:28 +010010977 not match. If the frontend uses a "bind-process" setting, the intersection
Christopher Fauletc644fa92017-11-23 22:44:11 +010010978 between the two is applied. If in the end the listener is not allowed to run
10979 on any remaining process, a warning is emitted, and the listener will either
10980 run on the first process of the listener if a single process was specified,
10981 or on all of its processes if multiple processes were specified. If a thread
Davor Ocelice9ed2812017-12-25 17:49:28 +010010982 set is specified, it limits the threads allowed to process incoming
Christopher Fauletc644fa92017-11-23 22:44:11 +010010983 connections for this listener, for the corresponding process set. For the
10984 unlikely case where several ranges are needed, this directive may be
10985 repeated. <process-set> and <thread-set> must use the format
10986
10987 all | odd | even | number[-[number]]
10988
10989 Ranges can be partially defined. The higher bound can be omitted. In such
10990 case, it is replaced by the corresponding maximum value. The main purpose of
10991 this directive is to be used with the stats sockets and have one different
10992 socket per process. The second purpose is to have multiple bind lines sharing
10993 the same IP:port but not the same process in a listener, so that the system
10994 can distribute the incoming connections into multiple queues and allow a
10995 smoother inter-process load balancing. Currently Linux 3.9 and above is known
10996 for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020010997
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010998ssl
10999 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011000 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011001 certificate is necessary (see "crt" above). All contents in the buffers will
11002 appear in clear text, so that ACLs and HTTP processing will only have access
Emmanuel Hocdetbd695fe2017-05-15 15:53:41 +020011003 to deciphered contents. SSLv3 is disabled per default, use "ssl-min-ver SSLv3"
11004 to enable it.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011005
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011006ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11007 This option enforces use of <version> or lower on SSL connections instantiated
11008 from this listener. This option is also available on global statement
11009 "ssl-default-bind-options". See also "ssl-min-ver".
11010
11011ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11012 This option enforces use of <version> or upper on SSL connections instantiated
11013 from this listener. This option is also available on global statement
11014 "ssl-default-bind-options". See also "ssl-max-ver".
11015
Emmanuel Hocdet65623372013-01-24 17:17:15 +010011016strict-sni
11017 This setting is only available when support for OpenSSL was built in. The
11018 SSL/TLS negotiation is allow only if the client provided an SNI which match
11019 a certificate. The default certificate is not used.
11020 See the "crt" option for more information.
11021
Willy Tarreau2af207a2015-02-04 00:45:58 +010011022tcp-ut <delay>
Tim Düsterhus4896c442016-11-29 02:15:19 +010011023 Sets the TCP User Timeout for all incoming connections instantiated from this
Willy Tarreau2af207a2015-02-04 00:45:58 +010011024 listening socket. This option is available on Linux since version 2.6.37. It
11025 allows haproxy to configure a timeout for sockets which contain data not
Davor Ocelice9ed2812017-12-25 17:49:28 +010011026 receiving an acknowledgment for the configured delay. This is especially
Willy Tarreau2af207a2015-02-04 00:45:58 +010011027 useful on long-lived connections experiencing long idle periods such as
11028 remote terminals or database connection pools, where the client and server
11029 timeouts must remain high to allow a long period of idle, but where it is
11030 important to detect that the client has disappeared in order to release all
11031 resources associated with its connection (and the server's session). The
11032 argument is a delay expressed in milliseconds by default. This only works
11033 for regular TCP connections, and is ignored for other protocols.
11034
Willy Tarreau1c862c52012-10-05 16:21:00 +020011035tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010011036 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020011037 enables TCP Fast Open on the listening socket, which means that clients which
11038 support this feature will be able to send a request and receive a response
11039 during the 3-way handshake starting from second connection, thus saving one
11040 round-trip after the first connection. This only makes sense with protocols
11041 that use high connection rates and where each round trip matters. This can
11042 possibly cause issues with many firewalls which do not accept data on SYN
11043 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020011044 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
11045 need to build HAProxy with USE_TFO=1 if your libc doesn't define
11046 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020011047
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011048tls-ticket-keys <keyfile>
11049 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
11050 bytes long, encoded with base64 (ex. openssl rand -base64 48). Number of keys
11051 is specified by the TLS_TICKETS_NO build option (default 3) and at least as
11052 many keys need to be present in the file. Last TLS_TICKETS_NO keys will be
11053 used for decryption and the penultimate one for encryption. This enables easy
11054 key rotation by just appending new key to the file and reloading the process.
11055 Keys must be periodically rotated (ex. every 12h) or Perfect Forward Secrecy
11056 is compromised. It is also a good idea to keep the keys off any permanent
11057 storage such as hard drives (hint: use tmpfs and don't swap those files).
11058 Lifetime hint can be changed using tune.ssl.timeout.
11059
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011060transparent
11061 Is an optional keyword which is supported only on certain Linux kernels. It
11062 indicates that the addresses will be bound even if they do not belong to the
11063 local machine, and that packets targeting any of these addresses will be
11064 intercepted just as if the addresses were locally configured. This normally
11065 requires that IP forwarding is enabled. Caution! do not use this with the
11066 default address '*', as it would redirect any traffic for the specified port.
11067 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
11068 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
11069 kernel version. Some distribution kernels include backports of the feature,
11070 so check for support with your vendor.
11071
Willy Tarreau77e3af92012-11-24 15:07:23 +010011072v4v6
11073 Is an optional keyword which is supported only on most recent systems
11074 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
11075 and IPv6 when it uses the default address. Doing so is sometimes necessary
11076 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011077 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010011078
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011079v6only
11080 Is an optional keyword which is supported only on most recent systems
11081 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
11082 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010011083 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
11084 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011085
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011086uid <uid>
11087 Sets the owner of the UNIX sockets to the designated system uid. It can also
11088 be set by default in the global section's "unix-bind" statement. Note that
11089 some platforms simply ignore this. This setting is equivalent to the "user"
11090 setting except that the user numeric ID is used instead of its name. This
11091 setting is ignored by non UNIX sockets.
11092
11093user <user>
11094 Sets the owner of the UNIX sockets to the designated system user. It can also
11095 be set by default in the global section's "unix-bind" statement. Note that
11096 some platforms simply ignore this. This setting is equivalent to the "uid"
11097 setting except that the user name is used instead of its uid. This setting is
11098 ignored by non UNIX sockets.
11099
Emeric Brun1a073b42012-09-28 17:07:34 +020011100verify [none|optional|required]
11101 This setting is only available when support for OpenSSL was built in. If set
11102 to 'none', client certificate is not requested. This is the default. In other
11103 cases, a client certificate is requested. If the client does not provide a
11104 certificate after the request and if 'verify' is set to 'required', then the
11105 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020011106 certificate provided by the client is always verified using CAs from
11107 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
11108 is aborted, regardless of the 'verify' option, unless the error code exactly
11109 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011110
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200111115.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010011112------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011113
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011114The "server" and "default-server" keywords support a certain number of settings
11115which are all passed as arguments on the server line. The order in which those
11116arguments appear does not count, and they are all optional. Some of those
11117settings are single words (booleans) while others expect one or several values
11118after them. In this case, the values must immediately follow the setting name.
11119Except default-server, all those settings must be specified after the server's
11120address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011121
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011122 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011123 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020011124
Frédéric Lécailled2376272017-03-21 18:52:12 +010011125Note that all these settings are supported both by "server" and "default-server"
11126keywords, except "id" which is only supported by "server".
11127
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011128The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011129
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011130addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011131 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010011132 to send health-checks or to probe the agent-check. On some servers, it may be
11133 desirable to dedicate an IP address to specific component able to perform
11134 complex tests which are more suitable to health-checks than the application.
11135 This parameter is ignored if the "check" parameter is not set. See also the
11136 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011137
Simon Hormand60d6912013-11-25 10:46:36 +090011138agent-check
11139 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010011140 health check. An agent health check is performed by making a TCP connection
11141 to the port set by the "agent-port" parameter and reading an ASCII string.
11142 The string is made of a series of words delimited by spaces, tabs or commas
11143 in any order, optionally terminated by '\r' and/or '\n', each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090011144
Willy Tarreau81f5d942013-12-09 20:51:51 +010011145 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090011146 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020011147 weight of a server as configured when haproxy starts. Note that a zero
11148 weight is reported on the stats page as "DRAIN" since it has the same
11149 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090011150
Davor Ocelice9ed2812017-12-25 17:49:28 +010011151 - The string "maxconn:" followed by an integer (no space between). Values
11152 in this format will set the maxconn of a server. The maximum number of
11153 connections advertised needs to be multiplied by the number of load
11154 balancers and different backends that use this health check to get the
11155 total number of connections the server might receive. Example: maxconn:30
Nenad Merdanovic174dd372016-04-24 23:10:06 +020011156
Willy Tarreau81f5d942013-12-09 20:51:51 +010011157 - The word "ready". This will turn the server's administrative state to the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011158 READY mode, thus canceling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090011159
Willy Tarreau81f5d942013-12-09 20:51:51 +010011160 - The word "drain". This will turn the server's administrative state to the
11161 DRAIN mode, thus it will not accept any new connections other than those
11162 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090011163
Willy Tarreau81f5d942013-12-09 20:51:51 +010011164 - The word "maint". This will turn the server's administrative state to the
11165 MAINT mode, thus it will not accept any new connections at all, and health
11166 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090011167
Willy Tarreau81f5d942013-12-09 20:51:51 +010011168 - The words "down", "failed", or "stopped", optionally followed by a
11169 description string after a sharp ('#'). All of these mark the server's
11170 operating state as DOWN, but since the word itself is reported on the stats
11171 page, the difference allows an administrator to know if the situation was
11172 expected or not : the service may intentionally be stopped, may appear up
Davor Ocelice9ed2812017-12-25 17:49:28 +010011173 but fail some validity tests, or may be seen as down (e.g. missing process,
Willy Tarreau81f5d942013-12-09 20:51:51 +010011174 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090011175
Willy Tarreau81f5d942013-12-09 20:51:51 +010011176 - The word "up" sets back the server's operating state as UP if health checks
11177 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090011178
Willy Tarreau81f5d942013-12-09 20:51:51 +010011179 Parameters which are not advertised by the agent are not changed. For
11180 example, an agent might be designed to monitor CPU usage and only report a
11181 relative weight and never interact with the operating status. Similarly, an
11182 agent could be designed as an end-user interface with 3 radio buttons
11183 allowing an administrator to change only the administrative state. However,
11184 it is important to consider that only the agent may revert its own actions,
11185 so if a server is set to DRAIN mode or to DOWN state using the agent, the
11186 agent must implement the other equivalent actions to bring the service into
11187 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090011188
Simon Horman2f1f9552013-11-25 10:46:37 +090011189 Failure to connect to the agent is not considered an error as connectivity
11190 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010011191 parameter. Warning though, it is not a good idea to stop an agent after it
11192 reports "down", since only an agent reporting "up" will be able to turn the
11193 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010011194 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090011195
Willy Tarreau81f5d942013-12-09 20:51:51 +010011196 Requires the "agent-port" parameter to be set. See also the "agent-inter"
Frédéric Lécailled2376272017-03-21 18:52:12 +010011197 and "no-agent-check" parameters.
Simon Hormand60d6912013-11-25 10:46:36 +090011198
James Brown55f9ff12015-10-21 18:19:05 -070011199agent-send <string>
11200 If this option is specified, haproxy will send the given string (verbatim)
11201 to the agent server upon connection. You could, for example, encode
11202 the backend name into this string, which would enable your agent to send
11203 different responses based on the backend. Make sure to include a '\n' if
11204 you want to terminate your request with a newline.
11205
Simon Hormand60d6912013-11-25 10:46:36 +090011206agent-inter <delay>
11207 The "agent-inter" parameter sets the interval between two agent checks
11208 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
11209
11210 Just as with every other time-based parameter, it may be entered in any
11211 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
11212 parameter also serves as a timeout for agent checks "timeout check" is
11213 not set. In order to reduce "resonance" effects when multiple servers are
11214 hosted on the same hardware, the agent and health checks of all servers
11215 are started with a small time offset between them. It is also possible to
11216 add some random noise in the agent and health checks interval using the
11217 global "spread-checks" keyword. This makes sense for instance when a lot
11218 of backends use the same servers.
11219
11220 See also the "agent-check" and "agent-port" parameters.
11221
Misiek768d8602017-01-09 09:52:43 +010011222agent-addr <addr>
11223 The "agent-addr" parameter sets address for agent check.
11224
11225 You can offload agent-check to another target, so you can make single place
11226 managing status and weights of servers defined in haproxy in case you can't
11227 make self-aware and self-managing services. You can specify both IP or
11228 hostname, it will be resolved.
11229
Simon Hormand60d6912013-11-25 10:46:36 +090011230agent-port <port>
11231 The "agent-port" parameter sets the TCP port used for agent checks.
11232
11233 See also the "agent-check" and "agent-inter" parameters.
11234
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011235backup
11236 When "backup" is present on a server line, the server is only used in load
11237 balancing when all other non-backup servers are unavailable. Requests coming
11238 with a persistence cookie referencing the server will always be served
11239 though. By default, only the first operational backup server is used, unless
Frédéric Lécailled2376272017-03-21 18:52:12 +010011240 the "allbackups" option is set in the backend. See also the "no-backup" and
11241 "allbackups" options.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011242
Emeric Brunef42d922012-10-11 16:11:36 +020011243ca-file <cafile>
11244 This setting is only available when support for OpenSSL was built in. It
11245 designates a PEM file from which to load CA certificates used to verify
11246 server's certificate.
11247
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011248check
11249 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010011250 always considered available. If "check" is set, the server is available when
11251 accepting periodic TCP connections, to ensure that it is really able to serve
11252 requests. The default address and port to send the tests to are those of the
11253 server, and the default source is the same as the one defined in the
11254 backend. It is possible to change the address using the "addr" parameter, the
11255 port using the "port" parameter, the source address using the "source"
11256 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090011257 parameters. The request method is define in the backend using the "httpchk",
11258 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
Frédéric Lécailled2376272017-03-21 18:52:12 +010011259 refer to those options and parameters for more information. See also
11260 "no-check" option.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011261
Willy Tarreau6c16adc2012-10-05 00:04:16 +020011262check-send-proxy
11263 This option forces emission of a PROXY protocol line with outgoing health
11264 checks, regardless of whether the server uses send-proxy or not for the
11265 normal traffic. By default, the PROXY protocol is enabled for health checks
11266 if it is already enabled for normal traffic and if no "port" nor "addr"
11267 directive is present. However, if such a directive is present, the
11268 "check-send-proxy" option needs to be used to force the use of the
11269 protocol. See also the "send-proxy" option for more information.
11270
Olivier Houchard9130a962017-10-17 17:33:43 +020011271check-sni
11272 This option allows you to specify the SNI to be used when doing health checks
11273 over SSL.
11274
Willy Tarreau763a95b2012-10-04 23:15:39 +020011275check-ssl
11276 This option forces encryption of all health checks over SSL, regardless of
11277 whether the server uses SSL or not for the normal traffic. This is generally
11278 used when an explicit "port" or "addr" directive is specified and SSL health
11279 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011280 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020011281 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
11282 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011283 All SSL settings are common to health checks and traffic (e.g. ciphers).
Frédéric Lécailled2376272017-03-21 18:52:12 +010011284 See the "ssl" option for more information and "no-check-ssl" to disable
11285 this option.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011286
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011287ciphers <ciphers>
11288 This option sets the string describing the list of cipher algorithms that is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011289 is negotiated during the SSL/TLS handshake with the server. The format of the
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011290 string is defined in "man 1 ciphers". When SSL is used to communicate with
11291 servers on the local network, it is common to see a weaker set of algorithms
11292 than what is used over the internet. Doing so reduces CPU usage on both the
11293 server and haproxy while still keeping it compatible with deployed software.
11294 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
11295 is needed and just connectivity, using DES can be appropriate.
11296
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011297cookie <value>
11298 The "cookie" parameter sets the cookie value assigned to the server to
11299 <value>. This value will be checked in incoming requests, and the first
11300 operational server possessing the same value will be selected. In return, in
11301 cookie insertion or rewrite modes, this value will be assigned to the cookie
11302 sent to the client. There is nothing wrong in having several servers sharing
11303 the same cookie value, and it is in fact somewhat common between normal and
11304 backup servers. See also the "cookie" keyword in backend section.
11305
Emeric Brunef42d922012-10-11 16:11:36 +020011306crl-file <crlfile>
11307 This setting is only available when support for OpenSSL was built in. It
11308 designates a PEM file from which to load certificate revocation list used
11309 to verify server's certificate.
11310
Emeric Bruna7aa3092012-10-26 12:58:00 +020011311crt <cert>
11312 This setting is only available when support for OpenSSL was built in.
11313 It designates a PEM file from which to load both a certificate and the
11314 associated private key. This file can be built by concatenating both PEM
11315 files into one. This certificate will be sent if the server send a client
11316 certificate request.
11317
Willy Tarreau96839092010-03-29 10:02:24 +020011318disabled
11319 The "disabled" keyword starts the server in the "disabled" state. That means
11320 that it is marked down in maintenance mode, and no connection other than the
11321 ones allowed by persist mode will reach it. It is very well suited to setup
11322 new servers, because normal traffic will never reach them, while it is still
11323 possible to test the service by making use of the force-persist mechanism.
Frédéric Lécailled2376272017-03-21 18:52:12 +010011324 See also "enabled" setting.
Willy Tarreau96839092010-03-29 10:02:24 +020011325
Frédéric Lécailled2376272017-03-21 18:52:12 +010011326enabled
11327 This option may be used as 'server' setting to reset any 'disabled'
11328 setting which would have been inherited from 'default-server' directive as
11329 default value.
11330 It may also be used as 'default-server' setting to reset any previous
11331 'default-server' 'disabled' setting.
Willy Tarreau96839092010-03-29 10:02:24 +020011332
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011333error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010011334 If health observing is enabled, the "error-limit" parameter specifies the
11335 number of consecutive errors that triggers event selected by the "on-error"
11336 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011337
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011338 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011339
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011340fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011341 The "fall" parameter states that a server will be considered as dead after
11342 <count> consecutive unsuccessful health checks. This value defaults to 3 if
11343 unspecified. See also the "check", "inter" and "rise" parameters.
11344
Emeric Brun8694b9a2012-10-05 14:39:07 +020011345force-sslv3
11346 This option enforces use of SSLv3 only when SSL is used to communicate with
11347 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011348 high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011349 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011350
11351force-tlsv10
11352 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011353 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011354 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011355
11356force-tlsv11
11357 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011358 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011359 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011360
11361force-tlsv12
11362 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011363 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011364 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011365
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011366force-tlsv13
11367 This option enforces use of TLSv1.3 only when SSL is used to communicate with
11368 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011369 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011370
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011371id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020011372 Set a persistent ID for the server. This ID must be positive and unique for
11373 the proxy. An unused ID will automatically be assigned if unset. The first
11374 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011375
Willy Tarreau6a031d12016-11-07 19:42:35 +010011376init-addr {last | libc | none | <ip>},[...]*
11377 Indicate in what order the server's address should be resolved upon startup
11378 if it uses an FQDN. Attempts are made to resolve the address by applying in
Davor Ocelice9ed2812017-12-25 17:49:28 +010011379 turn each of the methods mentioned in the comma-delimited list. The first
Willy Tarreau6a031d12016-11-07 19:42:35 +010011380 method which succeeds is used. If the end of the list is reached without
11381 finding a working method, an error is thrown. Method "last" suggests to pick
11382 the address which appears in the state file (see "server-state-file"). Method
11383 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
11384 depending on the operating system and build options). Method "none"
11385 specifically indicates that the server should start without any valid IP
11386 address in a down state. It can be useful to ignore some DNS issues upon
11387 startup, waiting for the situation to get fixed later. Finally, an IP address
11388 (IPv4 or IPv6) may be provided. It can be the currently known address of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011389 server (e.g. filled by a configuration generator), or the address of a dummy
Willy Tarreau6a031d12016-11-07 19:42:35 +010011390 server used to catch old sessions and present them with a decent error
11391 message for example. When the "first" load balancing algorithm is used, this
11392 IP address could point to a fake server used to trigger the creation of new
11393 instances on the fly. This option defaults to "last,libc" indicating that the
11394 previous address found in the state file (if any) is used first, otherwise
11395 the libc's resolver is used. This ensures continued compatibility with the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011396 historic behavior.
Willy Tarreau6a031d12016-11-07 19:42:35 +010011397
11398 Example:
11399 defaults
11400 # never fail on address resolution
11401 default-server init-addr last,libc,none
11402
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011403inter <delay>
11404fastinter <delay>
11405downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011406 The "inter" parameter sets the interval between two consecutive health checks
11407 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
11408 It is also possible to use "fastinter" and "downinter" to optimize delays
11409 between checks depending on the server state :
11410
Pieter Baauw44fc9df2015-09-17 21:30:46 +020011411 Server state | Interval used
11412 ----------------------------------------+----------------------------------
11413 UP 100% (non-transitional) | "inter"
11414 ----------------------------------------+----------------------------------
11415 Transitionally UP (going down "fall"), | "fastinter" if set,
11416 Transitionally DOWN (going up "rise"), | "inter" otherwise.
11417 or yet unchecked. |
11418 ----------------------------------------+----------------------------------
11419 DOWN 100% (non-transitional) | "downinter" if set,
11420 | "inter" otherwise.
11421 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010011422
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011423 Just as with every other time-based parameter, they can be entered in any
11424 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
11425 serves as a timeout for health checks sent to servers if "timeout check" is
11426 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090011427 hosted on the same hardware, the agent and health checks of all servers
11428 are started with a small time offset between them. It is also possible to
11429 add some random noise in the agent and health checks interval using the
11430 global "spread-checks" keyword. This makes sense for instance when a lot
11431 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011432
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011433maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011434 The "maxconn" parameter specifies the maximal number of concurrent
11435 connections that will be sent to this server. If the number of incoming
11436 concurrent requests goes higher than this value, they will be queued, waiting
11437 for a connection to be released. This parameter is very important as it can
11438 save fragile servers from going down under extreme loads. If a "minconn"
11439 parameter is specified, the limit becomes dynamic. The default value is "0"
11440 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
11441 the backend's "fullconn" keyword.
11442
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011443maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011444 The "maxqueue" parameter specifies the maximal number of connections which
11445 will wait in the queue for this server. If this limit is reached, next
11446 requests will be redispatched to other servers instead of indefinitely
11447 waiting to be served. This will break persistence but may allow people to
11448 quickly re-log in when the server they try to connect to is dying. The
11449 default value is "0" which means the queue is unlimited. See also the
11450 "maxconn" and "minconn" parameters.
11451
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011452minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011453 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
11454 limit following the backend's load. The server will always accept at least
11455 <minconn> connections, never more than <maxconn>, and the limit will be on
11456 the ramp between both values when the backend has less than <fullconn>
11457 concurrent connections. This makes it possible to limit the load on the
11458 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011459 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011460 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011461
Willy Tarreaud72f0f32015-10-13 14:50:22 +020011462namespace <name>
11463 On Linux, it is possible to specify which network namespace a socket will
11464 belong to. This directive makes it possible to explicitly bind a server to
11465 a namespace different from the default one. Please refer to your operating
11466 system's documentation to find more details about network namespaces.
11467
Frédéric Lécailled2376272017-03-21 18:52:12 +010011468no-agent-check
11469 This option may be used as "server" setting to reset any "agent-check"
11470 setting which would have been inherited from "default-server" directive as
11471 default value.
11472 It may also be used as "default-server" setting to reset any previous
11473 "default-server" "agent-check" setting.
11474
11475no-backup
11476 This option may be used as "server" setting to reset any "backup"
11477 setting which would have been inherited from "default-server" directive as
11478 default value.
11479 It may also be used as "default-server" setting to reset any previous
11480 "default-server" "backup" setting.
11481
11482no-check
11483 This option may be used as "server" setting to reset any "check"
11484 setting which would have been inherited from "default-server" directive as
11485 default value.
11486 It may also be used as "default-server" setting to reset any previous
11487 "default-server" "check" setting.
11488
11489no-check-ssl
11490 This option may be used as "server" setting to reset any "check-ssl"
11491 setting which would have been inherited from "default-server" directive as
11492 default value.
11493 It may also be used as "default-server" setting to reset any previous
11494 "default-server" "check-ssl" setting.
11495
Frédéric Lécailled2376272017-03-21 18:52:12 +010011496no-send-proxy
11497 This option may be used as "server" setting to reset any "send-proxy"
11498 setting which would have been inherited from "default-server" directive as
11499 default value.
11500 It may also be used as "default-server" setting to reset any previous
11501 "default-server" "send-proxy" setting.
11502
11503no-send-proxy-v2
11504 This option may be used as "server" setting to reset any "send-proxy-v2"
11505 setting which would have been inherited from "default-server" directive as
11506 default value.
11507 It may also be used as "default-server" setting to reset any previous
11508 "default-server" "send-proxy-v2" setting.
11509
11510no-send-proxy-v2-ssl
11511 This option may be used as "server" setting to reset any "send-proxy-v2-ssl"
11512 setting which would have been inherited from "default-server" directive as
11513 default value.
11514 It may also be used as "default-server" setting to reset any previous
11515 "default-server" "send-proxy-v2-ssl" setting.
11516
11517no-send-proxy-v2-ssl-cn
11518 This option may be used as "server" setting to reset any "send-proxy-v2-ssl-cn"
11519 setting which would have been inherited from "default-server" directive as
11520 default value.
11521 It may also be used as "default-server" setting to reset any previous
11522 "default-server" "send-proxy-v2-ssl-cn" setting.
11523
11524no-ssl
11525 This option may be used as "server" setting to reset any "ssl"
11526 setting which would have been inherited from "default-server" directive as
11527 default value.
11528 It may also be used as "default-server" setting to reset any previous
11529 "default-server" "ssl" setting.
11530
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010011531no-ssl-reuse
11532 This option disables SSL session reuse when SSL is used to communicate with
11533 the server. It will force the server to perform a full handshake for every
11534 new connection. It's probably only useful for benchmarking, troubleshooting,
11535 and for paranoid users.
11536
Emeric Brun9b3009b2012-10-05 11:55:06 +020011537no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011538 This option disables support for SSLv3 when SSL is used to communicate with
11539 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011540 using any configuration option. Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011541
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011542 Supported in default-server: No
11543
Emeric Brunf9c5c472012-10-11 15:28:34 +020011544no-tls-tickets
11545 This setting is only available when support for OpenSSL was built in. It
11546 disables the stateless session resumption (RFC 5077 TLS Ticket
11547 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011548 session resumption is more expensive in CPU usage for servers. This option
11549 is also available on global statement "ssl-default-server-options".
Frédéric Lécailled2376272017-03-21 18:52:12 +010011550 See also "tls-tickets".
Emeric Brunf9c5c472012-10-11 15:28:34 +020011551
Emeric Brun9b3009b2012-10-05 11:55:06 +020011552no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020011553 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020011554 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11555 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011556 often makes sense to disable it when communicating with local servers. This
11557 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011558 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011559
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011560 Supported in default-server: No
11561
Emeric Brun9b3009b2012-10-05 11:55:06 +020011562no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020011563 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020011564 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11565 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011566 often makes sense to disable it when communicating with local servers. This
11567 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011568 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011569
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011570 Supported in default-server: No
11571
Emeric Brun9b3009b2012-10-05 11:55:06 +020011572no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020011573 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011574 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11575 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011576 often makes sense to disable it when communicating with local servers. This
11577 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011578 Use "ssl-min-ver" and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011579
11580 Supported in default-server: No
11581
11582no-tlsv13
11583 This option disables support for TLSv1.3 when SSL is used to communicate with
11584 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11585 using any configuration option. TLSv1 is more expensive than SSLv3 so it
11586 often makes sense to disable it when communicating with local servers. This
11587 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011588 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011589
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011590 Supported in default-server: No
11591
Frédéric Lécailled2376272017-03-21 18:52:12 +010011592no-verifyhost
11593 This option may be used as "server" setting to reset any "verifyhost"
11594 setting which would have been inherited from "default-server" directive as
11595 default value.
11596 It may also be used as "default-server" setting to reset any previous
11597 "default-server" "verifyhost" setting.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011598
Simon Hormanfa461682011-06-25 09:39:49 +090011599non-stick
11600 Never add connections allocated to this sever to a stick-table.
11601 This may be used in conjunction with backup to ensure that
11602 stick-table persistence is disabled for backup servers.
11603
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011604observe <mode>
11605 This option enables health adjusting based on observing communication with
11606 the server. By default this functionality is disabled and enabling it also
11607 requires to enable health checks. There are two supported modes: "layer4" and
11608 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
11609 significant. In layer7, which is only allowed for http proxies, responses
11610 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010011611 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011612
11613 See also the "check", "on-error" and "error-limit".
11614
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011615on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011616 Select what should happen when enough consecutive errors are detected.
11617 Currently, four modes are available:
11618 - fastinter: force fastinter
11619 - fail-check: simulate a failed check, also forces fastinter (default)
11620 - sudden-death: simulate a pre-fatal failed health check, one more failed
11621 check will mark a server down, forces fastinter
11622 - mark-down: mark the server immediately down and force fastinter
11623
11624 See also the "check", "observe" and "error-limit".
11625
Simon Hormane0d1bfb2011-06-21 14:34:58 +090011626on-marked-down <action>
11627 Modify what occurs when a server is marked down.
11628 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011629 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
11630 all connections to the server are immediately terminated when the server
11631 goes down. It might be used if the health check detects more complex cases
11632 than a simple connection status, and long timeouts would cause the service
11633 to remain unresponsive for too long a time. For instance, a health check
11634 might detect that a database is stuck and that there's no chance to reuse
11635 existing connections anymore. Connections killed this way are logged with
11636 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090011637
11638 Actions are disabled by default
11639
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011640on-marked-up <action>
11641 Modify what occurs when a server is marked up.
11642 Currently one action is available:
11643 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
11644 done only if the server is not in backup state and if it is not disabled
11645 (it must have an effective weight > 0). This can be used sometimes to force
11646 an active server to take all the traffic back after recovery when dealing
Davor Ocelice9ed2812017-12-25 17:49:28 +010011647 with long sessions (e.g. LDAP, SQL, ...). Doing this can cause more trouble
11648 than it tries to solve (e.g. incomplete transactions), so use this feature
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011649 with extreme care. Sessions killed because a server comes up are logged
11650 with an 'U' termination code (for "Up").
11651
11652 Actions are disabled by default
11653
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011654port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011655 Using the "port" parameter, it becomes possible to use a different port to
11656 send health-checks. On some servers, it may be desirable to dedicate a port
11657 to a specific component able to perform complex tests which are more suitable
11658 to health-checks than the application. It is common to run a simple script in
11659 inetd for instance. This parameter is ignored if the "check" parameter is not
11660 set. See also the "addr" parameter.
11661
11662redir <prefix>
11663 The "redir" parameter enables the redirection mode for all GET and HEAD
11664 requests addressing this server. This means that instead of having HAProxy
11665 forward the request to the server, it will send an "HTTP 302" response with
11666 the "Location" header composed of this prefix immediately followed by the
11667 requested URI beginning at the leading '/' of the path component. That means
11668 that no trailing slash should be used after <prefix>. All invalid requests
11669 will be rejected, and all non-GET or HEAD requests will be normally served by
11670 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011671 mangling nor cookie insertion is possible in the response. However, cookies in
Davor Ocelice9ed2812017-12-25 17:49:28 +010011672 requests are still analyzed, making this solution completely usable to direct
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011673 users to a remote location in case of local disaster. Main use consists in
11674 increasing bandwidth for static servers by having the clients directly
11675 connect to them. Note: never use a relative location here, it would cause a
11676 loop between the client and HAProxy!
11677
11678 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
11679
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011680rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011681 The "rise" parameter states that a server will be considered as operational
11682 after <count> consecutive successful health checks. This value defaults to 2
11683 if unspecified. See also the "check", "inter" and "fall" parameters.
11684
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011685resolve-prefer <family>
11686 When DNS resolution is enabled for a server and multiple IP addresses from
11687 different families are returned, HAProxy will prefer using an IP address
11688 from the family mentioned in the "resolve-prefer" parameter.
11689 Available families: "ipv4" and "ipv6"
11690
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020011691 Default value: ipv6
11692
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011693 Example:
11694
11695 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011696
Thierry Fournierac88cfe2016-02-17 22:05:30 +010011697resolve-net <network>[,<network[,...]]
11698 This options prioritize th choice of an ip address matching a network. This is
11699 useful with clouds to prefer a local ip. In some cases, a cloud high
Tim Düsterhus4896c442016-11-29 02:15:19 +010011700 availability service can be announced with many ip addresses on many
Davor Ocelice9ed2812017-12-25 17:49:28 +010011701 different datacenters. The latency between datacenter is not negligible, so
11702 this patch permits to prefer a local datacenter. If no address matches the
Thierry Fournierac88cfe2016-02-17 22:05:30 +010011703 configured network, another address is selected.
11704
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011705 Example:
11706
11707 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010011708
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011709resolvers <id>
11710 Points to an existing "resolvers" section to resolve current server's
11711 hostname.
11712
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011713 Example:
11714
11715 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011716
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011717 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011718
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010011719send-proxy
11720 The "send-proxy" parameter enforces use of the PROXY protocol over any
11721 connection established to this server. The PROXY protocol informs the other
11722 end about the layer 3/4 addresses of the incoming connection, so that it can
11723 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011724 upper layer protocol. For connections accepted by an "accept-proxy" or
11725 "accept-netscaler-cip" listener, the advertised address will be used. Only
11726 TCPv4 and TCPv6 address families are supported. Other families such as
11727 Unix sockets, will report an UNKNOWN family. Servers using this option can
11728 fully be chained to another instance of haproxy listening with an
11729 "accept-proxy" setting. This setting must not be used if the server isn't
11730 aware of the protocol. When health checks are sent to the server, the PROXY
11731 protocol is automatically used when this option is set, unless there is an
11732 explicit "port" or "addr" directive, in which case an explicit
11733 "check-send-proxy" directive would also be needed to use the PROXY protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010011734 See also the "no-send-proxy" option of this section and "accept-proxy" and
11735 "accept-netscaler-cip" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010011736
David Safb76832014-05-08 23:42:08 -040011737send-proxy-v2
11738 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
11739 over any connection established to this server. The PROXY protocol informs
11740 the other end about the layer 3/4 addresses of the incoming connection, so
11741 that it can know the client's address or the public address it accessed to,
Emmanuel Hocdet404d9782017-10-24 10:55:14 +020011742 whatever the upper layer protocol. It also send ALPN information if an alpn
11743 have been negotiated. This setting must not be used if the server isn't aware
11744 of this version of the protocol. See also the "no-send-proxy-v2" option of
11745 this section and send-proxy" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040011746
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010011747proxy-v2-options <option>[,<option>]*
11748 The "proxy-v2-options" parameter add option to send in PROXY protocol version
11749 2 when "send-proxy-v2" is used. Options available are "ssl" (see also
Emmanuel Hocdetfa8d0f12018-02-01 15:53:52 +010011750 send-proxy-v2-ssl), "cert-cn" (see also "send-proxy-v2-ssl-cn"), "ssl-cipher":
11751 name of the used cipher, "cert-sig": signature algorithm of the used
Emmanuel Hocdet253c3b72018-02-01 18:29:59 +010011752 certificate, "cert-key": key algorithm of the used certificate), "authority":
11753 host name value passed by the client (only sni from a tls connection is
Emmanuel Hocdet4399c752018-02-05 15:26:43 +010011754 supported), "crc32c": checksum of the proxy protocol v2 header.
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010011755
David Safb76832014-05-08 23:42:08 -040011756send-proxy-v2-ssl
11757 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
11758 2 over any connection established to this server. The PROXY protocol informs
11759 the other end about the layer 3/4 addresses of the incoming connection, so
11760 that it can know the client's address or the public address it accessed to,
11761 whatever the upper layer protocol. In addition, the SSL information extension
11762 of the PROXY protocol is added to the PROXY protocol header. This setting
11763 must not be used if the server isn't aware of this version of the protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010011764 See also the "no-send-proxy-v2-ssl" option of this section and the
11765 "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040011766
11767send-proxy-v2-ssl-cn
11768 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
11769 2 over any connection established to this server. The PROXY protocol informs
11770 the other end about the layer 3/4 addresses of the incoming connection, so
11771 that it can know the client's address or the public address it accessed to,
11772 whatever the upper layer protocol. In addition, the SSL information extension
11773 of the PROXY protocol, along along with the Common Name from the subject of
11774 the client certificate (if any), is added to the PROXY protocol header. This
11775 setting must not be used if the server isn't aware of this version of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011776 protocol. See also the "no-send-proxy-v2-ssl-cn" option of this section and
11777 the "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040011778
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011779slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011780 The "slowstart" parameter for a server accepts a value in milliseconds which
11781 indicates after how long a server which has just come back up will run at
11782 full speed. Just as with every other time-based parameter, it can be entered
11783 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
11784 linearly from 0 to 100% during this time. The limitation applies to two
11785 parameters :
11786
11787 - maxconn: the number of connections accepted by the server will grow from 1
11788 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
11789
11790 - weight: when the backend uses a dynamic weighted algorithm, the weight
11791 grows linearly from 1 to 100%. In this case, the weight is updated at every
11792 health-check. For this reason, it is important that the "inter" parameter
11793 is smaller than the "slowstart", in order to maximize the number of steps.
11794
11795 The slowstart never applies when haproxy starts, otherwise it would cause
11796 trouble to running servers. It only applies when a server has been previously
11797 seen as failed.
11798
Willy Tarreau732eac42015-07-09 11:40:25 +020011799sni <expression>
11800 The "sni" parameter evaluates the sample fetch expression, converts it to a
11801 string and uses the result as the host name sent in the SNI TLS extension to
11802 the server. A typical use case is to send the SNI received from the client in
11803 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
Willy Tarreau2ab88672017-07-05 18:23:03 +020011804 expression, though alternatives such as req.hdr(host) can also make sense. If
11805 "verify required" is set (which is the recommended setting), the resulting
Willy Tarreauad92a9a2017-07-28 11:38:41 +020011806 name will also be matched against the server certificate's names. See the
11807 "verify" directive for more details.
Willy Tarreau732eac42015-07-09 11:40:25 +020011808
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020011809source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020011810source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020011811source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011812 The "source" parameter sets the source address which will be used when
11813 connecting to the server. It follows the exact same parameters and principle
11814 as the backend "source" keyword, except that it only applies to the server
11815 referencing it. Please consult the "source" keyword for details.
11816
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020011817 Additionally, the "source" statement on a server line allows one to specify a
11818 source port range by indicating the lower and higher bounds delimited by a
11819 dash ('-'). Some operating systems might require a valid IP address when a
11820 source port range is specified. It is permitted to have the same IP/range for
11821 several servers. Doing so makes it possible to bypass the maximum of 64k
11822 total concurrent connections. The limit will then reach 64k connections per
11823 server.
11824
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000011825 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
11826 specifying the source address without port(s).
11827
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011828ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020011829 This option enables SSL ciphering on outgoing connections to the server. It
11830 is critical to verify server certificates using "verify" when using SSL to
11831 connect to servers, otherwise the communication is prone to trivial man in
11832 the-middle attacks rendering SSL useless. When this option is used, health
11833 checks are automatically sent in SSL too unless there is a "port" or an
11834 "addr" directive indicating the check should be sent to a different location.
Frédéric Lécailled2376272017-03-21 18:52:12 +010011835 See the "no-ssl" to disable "ssl" option and "check-ssl" option to force
11836 SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011837
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011838ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11839 This option enforces use of <version> or lower when SSL is used to communicate
11840 with the server. This option is also available on global statement
11841 "ssl-default-server-options". See also "ssl-min-ver".
11842
11843ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11844 This option enforces use of <version> or upper when SSL is used to communicate
11845 with the server. This option is also available on global statement
11846 "ssl-default-server-options". See also "ssl-max-ver".
11847
Frédéric Lécailled2376272017-03-21 18:52:12 +010011848ssl-reuse
11849 This option may be used as "server" setting to reset any "no-ssl-reuse"
11850 setting which would have been inherited from "default-server" directive as
11851 default value.
11852 It may also be used as "default-server" setting to reset any previous
11853 "default-server" "no-ssl-reuse" setting.
11854
11855stick
11856 This option may be used as "server" setting to reset any "non-stick"
11857 setting which would have been inherited from "default-server" directive as
11858 default value.
11859 It may also be used as "default-server" setting to reset any previous
11860 "default-server" "non-stick" setting.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011861
Willy Tarreau163d4622015-10-13 16:16:41 +020011862tcp-ut <delay>
11863 Sets the TCP User Timeout for all outgoing connections to this server. This
11864 option is available on Linux since version 2.6.37. It allows haproxy to
11865 configure a timeout for sockets which contain data not receiving an
Davor Ocelice9ed2812017-12-25 17:49:28 +010011866 acknowledgment for the configured delay. This is especially useful on
Willy Tarreau163d4622015-10-13 16:16:41 +020011867 long-lived connections experiencing long idle periods such as remote
11868 terminals or database connection pools, where the client and server timeouts
11869 must remain high to allow a long period of idle, but where it is important to
11870 detect that the server has disappeared in order to release all resources
11871 associated with its connection (and the client's session). One typical use
11872 case is also to force dead server connections to die when health checks are
11873 too slow or during a soft reload since health checks are then disabled. The
11874 argument is a delay expressed in milliseconds by default. This only works for
11875 regular TCP connections, and is ignored for other protocols.
11876
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011877track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020011878 This option enables ability to set the current state of the server by tracking
11879 another one. It is possible to track a server which itself tracks another
11880 server, provided that at the end of the chain, a server has health checks
11881 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011882 used, it has to be enabled on both proxies.
11883
Frédéric Lécailled2376272017-03-21 18:52:12 +010011884tls-tickets
11885 This option may be used as "server" setting to reset any "no-tls-tickets"
11886 setting which would have been inherited from "default-server" directive as
11887 default value.
11888 It may also be used as "default-server" setting to reset any previous
11889 "default-server" "no-tlsv-tickets" setting.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011890
Emeric Brunef42d922012-10-11 16:11:36 +020011891verify [none|required]
11892 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010011893 to 'none', server certificate is not verified. In the other case, The
Willy Tarreauad92a9a2017-07-28 11:38:41 +020011894 certificate provided by the server is verified using CAs from 'ca-file' and
11895 optional CRLs from 'crl-file' after having checked that the names provided in
Davor Ocelice9ed2812017-12-25 17:49:28 +010011896 the certificate's subject and subjectAlternateNames attributes match either
Willy Tarreauad92a9a2017-07-28 11:38:41 +020011897 the name passed using the "sni" directive, or if not provided, the static
11898 host name passed using the "verifyhost" directive. When no name is found, the
11899 certificate's names are ignored. For this reason, without SNI it's important
11900 to use "verifyhost". On verification failure the handshake is aborted. It is
11901 critically important to verify server certificates when using SSL to connect
11902 to servers, otherwise the communication is prone to trivial man-in-the-middle
11903 attacks rendering SSL totally useless. Unless "ssl_server_verify" appears in
11904 the global section, "verify" is set to "required" by default.
Emeric Brunef42d922012-10-11 16:11:36 +020011905
Evan Broderbe554312013-06-27 00:05:25 -070011906verifyhost <hostname>
11907 This setting is only available when support for OpenSSL was built in, and
Willy Tarreauad92a9a2017-07-28 11:38:41 +020011908 only takes effect if 'verify required' is also specified. This directive sets
11909 a default static hostname to check the server's certificate against when no
11910 SNI was used to connect to the server. If SNI is not used, this is the only
11911 way to enable hostname verification. This static hostname, when set, will
11912 also be used for health checks (which cannot provide an SNI value). If none
11913 of the hostnames in the certificate match the specified hostname, the
11914 handshake is aborted. The hostnames in the server-provided certificate may
11915 include wildcards. See also "verify", "sni" and "no-verifyhost" options.
Evan Broderbe554312013-06-27 00:05:25 -070011916
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011917weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011918 The "weight" parameter is used to adjust the server's weight relative to
11919 other servers. All servers will receive a load proportional to their weight
11920 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020011921 load. The default weight is 1, and the maximal value is 256. A value of 0
11922 means the server will not participate in load-balancing but will still accept
11923 persistent connections. If this parameter is used to distribute the load
11924 according to server's capacity, it is recommended to start with values which
11925 can both grow and shrink, for instance between 10 and 100 to leave enough
11926 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011927
11928
Cyril Bonté46175dd2015-07-02 22:45:32 +0200119295.3. Server IP address resolution using DNS
11930-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011931
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011932HAProxy allows using a host name on the server line to retrieve its IP address
11933using name servers. By default, HAProxy resolves the name when parsing the
11934configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011935This is not sufficient in some cases, such as in Amazon where a server's IP
11936can change after a reboot or an ELB Virtual IP can change based on current
11937workload.
11938This chapter describes how HAProxy can be configured to process server's name
11939resolution at run time.
11940Whether run time server name resolution has been enable or not, HAProxy will
11941carry on doing the first resolution when parsing the configuration.
11942
11943
Cyril Bonté46175dd2015-07-02 22:45:32 +0200119445.3.1. Global overview
11945----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011946
11947As we've seen in introduction, name resolution in HAProxy occurs at two
11948different steps of the process life:
11949
11950 1. when starting up, HAProxy parses the server line definition and matches a
11951 host name. It uses libc functions to get the host name resolved. This
11952 resolution relies on /etc/resolv.conf file.
11953
Christopher Faulet67957bd2017-09-27 11:00:59 +020011954 2. at run time, HAProxy performs periodically name resolutions for servers
11955 requiring DNS resolutions.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011956
11957A few other events can trigger a name resolution at run time:
11958 - when a server's health check ends up in a connection timeout: this may be
11959 because the server has a new IP address. So we need to trigger a name
11960 resolution to know this new IP.
11961
Christopher Faulet67957bd2017-09-27 11:00:59 +020011962When using resolvers, the server name can either be a hostname, or a SRV label.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011963HAProxy considers anything that starts with an underscore as a SRV label. If a
Christopher Faulet67957bd2017-09-27 11:00:59 +020011964SRV label is specified, then the corresponding SRV records will be retrieved
11965from the DNS server, and the provided hostnames will be used. The SRV label
11966will be checked periodically, and if any server are added or removed, haproxy
11967will automatically do the same.
Olivier Houchardecfa18d2017-08-07 17:30:03 +020011968
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011969A few things important to notice:
11970 - all the name servers are queried in the mean time. HAProxy will process the
11971 first valid response.
11972
11973 - a resolution is considered as invalid (NX, timeout, refused), when all the
11974 servers return an error.
11975
11976
Cyril Bonté46175dd2015-07-02 22:45:32 +0200119775.3.2. The resolvers section
11978----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011979
11980This section is dedicated to host information related to name resolution in
Christopher Faulet67957bd2017-09-27 11:00:59 +020011981HAProxy. There can be as many as resolvers section as needed. Each section can
11982contain many name servers.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011983
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011984When multiple name servers are configured in a resolvers section, then HAProxy
11985uses the first valid response. In case of invalid responses, only the last one
11986is treated. Purpose is to give the chance to a slow server to deliver a valid
11987answer after a fast faulty or outdated server.
11988
11989When each server returns a different error type, then only the last error is
Christopher Faulet67957bd2017-09-27 11:00:59 +020011990used by HAProxy. The following processing is applied on this error:
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011991
Christopher Faulet67957bd2017-09-27 11:00:59 +020011992 1. HAProxy retries the same DNS query with a new query type. The A queries are
11993 switch to AAAA or the opposite. SRV queries are not concerned here. Timeout
11994 errors are also excluded.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011995
Christopher Faulet67957bd2017-09-27 11:00:59 +020011996 2. When the fallback on the query type was done (or not applicable), HAProxy
11997 retries the original DNS query, with the preferred query type.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011998
Christopher Faulet67957bd2017-09-27 11:00:59 +020011999 3. HAProxy retries previous steps <resolve_retires> times. If no valid
12000 response is received after that, it stops the DNS resolution and reports
12001 the error.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012002
Christopher Faulet67957bd2017-09-27 11:00:59 +020012003For example, with 2 name servers configured in a resolvers section, the
12004following scenarios are possible:
12005
12006 - First response is valid and is applied directly, second response is
12007 ignored
12008
12009 - First response is invalid and second one is valid, then second response is
12010 applied
12011
12012 - First response is a NX domain and second one a truncated response, then
12013 HAProxy retries the query with a new type
12014
12015 - First response is a NX domain and second one is a timeout, then HAProxy
12016 retries the query with a new type
12017
12018 - Query timed out for both name servers, then HAProxy retries it with the
12019 same query type
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012020
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012021As a DNS server may not answer all the IPs in one DNS request, haproxy keeps
12022a cache of previous answers, an answer will be considered obsolete after
Christopher Faulet67957bd2017-09-27 11:00:59 +020012023<hold obsolete> seconds without the IP returned.
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012024
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012025
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012026resolvers <resolvers id>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012027 Creates a new name server list labeled <resolvers id>
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012028
12029A resolvers section accept the following parameters:
12030
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012031accepted_payload_size <nb>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012032 Defines the maximum payload size accepted by HAProxy and announced to all the
Christopher Faulet67957bd2017-09-27 11:00:59 +020012033 name servers configured in this resolvers section.
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012034 <nb> is in bytes. If not set, HAProxy announces 512. (minimal value defined
12035 by RFC 6891)
12036
Baptiste Assmann9d8dbbc2017-08-18 23:35:08 +020012037 Note: the maximum allowed value is 8192.
12038
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012039nameserver <id> <ip>:<port>
12040 DNS server description:
12041 <id> : label of the server, should be unique
12042 <ip> : IP address of the server
12043 <port> : port where the DNS service actually runs
12044
Ben Draut44e609b2018-05-29 15:40:08 -060012045parse-resolv-conf
12046 Adds all nameservers found in /etc/resolv.conf to this resolvers nameservers
12047 list. Ordered as if each nameserver in /etc/resolv.conf was individually
12048 placed in the resolvers section in place of this directive.
12049
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012050hold <status> <period>
12051 Defines <period> during which the last name resolution should be kept based
12052 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012053 <status> : last name resolution status. Acceptable values are "nx",
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012054 "other", "refused", "timeout", "valid", "obsolete".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012055 <period> : interval between two successive name resolution when the last
12056 answer was in <status>. It follows the HAProxy time format.
12057 <period> is in milliseconds by default.
12058
Baptiste Assmann686408b2017-08-18 10:15:42 +020012059 Default value is 10s for "valid", 0s for "obsolete" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012060
Christopher Faulet67957bd2017-09-27 11:00:59 +020012061resolution_pool_size <nb> (deprecated)
Baptiste Assmann201c07f2017-05-22 15:17:15 +020012062 Defines the number of resolutions available in the pool for this resolvers.
12063 If not defines, it defaults to 64. If your configuration requires more than
12064 <nb>, then HAProxy will return an error when parsing the configuration.
12065
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012066resolve_retries <nb>
12067 Defines the number <nb> of queries to send to resolve a server name before
12068 giving up.
12069 Default value: 3
12070
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012071 A retry occurs on name server timeout or when the full sequence of DNS query
12072 type failover is over and we need to start up from the default ANY query
12073 type.
12074
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012075timeout <event> <time>
12076 Defines timeouts related to name resolution
12077 <event> : the event on which the <time> timeout period applies to.
12078 events available are:
Christopher Faulet67957bd2017-09-27 11:00:59 +020012079 - resolve : default time to trigger name resolutions when no
12080 other time applied.
12081 Default value: 1s
12082 - retry : time between two DNS queries, when no valid response
12083 have been received.
12084 Default value: 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012085 <time> : time related to the event. It follows the HAProxy time format.
12086 <time> is expressed in milliseconds.
12087
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012088 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012089
12090 resolvers mydns
12091 nameserver dns1 10.0.0.1:53
12092 nameserver dns2 10.0.0.2:53
Ben Draut44e609b2018-05-29 15:40:08 -060012093 parse-resolv-conf
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012094 resolve_retries 3
Christopher Faulet67957bd2017-09-27 11:00:59 +020012095 timeout resolve 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012096 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012097 hold other 30s
12098 hold refused 30s
12099 hold nx 30s
12100 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012101 hold valid 10s
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012102 hold obsolete 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012103
12104
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200121056. HTTP header manipulation
12106---------------------------
12107
12108In HTTP mode, it is possible to rewrite, add or delete some of the request and
12109response headers based on regular expressions. It is also possible to block a
12110request or a response if a particular header matches a regular expression,
12111which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010012112against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012113
Willy Tarreau70dffda2014-01-30 03:07:23 +010012114If HAProxy encounters an "Informational Response" (status code 1xx), it is able
12115to process all rsp* rules which can allow, deny, rewrite or delete a header,
12116but it will refuse to add a header to any such messages as this is not
12117HTTP-compliant. The reason for still processing headers in such responses is to
12118stop and/or fix any possible information leak which may happen, for instance
12119because another downstream equipment would unconditionally add a header, or if
12120a server name appears there. When such messages are seen, normal processing
12121still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020012122
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012123This section covers common usage of the following keywords, described in detail
12124in section 4.2 :
12125
12126 - reqadd <string>
12127 - reqallow <search>
12128 - reqiallow <search>
12129 - reqdel <search>
12130 - reqidel <search>
12131 - reqdeny <search>
12132 - reqideny <search>
12133 - reqpass <search>
12134 - reqipass <search>
12135 - reqrep <search> <replace>
12136 - reqirep <search> <replace>
12137 - reqtarpit <search>
12138 - reqitarpit <search>
12139 - rspadd <string>
12140 - rspdel <search>
12141 - rspidel <search>
12142 - rspdeny <search>
12143 - rspideny <search>
12144 - rsprep <search> <replace>
12145 - rspirep <search> <replace>
12146
12147With all these keywords, the same conventions are used. The <search> parameter
12148is a POSIX extended regular expression (regex) which supports grouping through
12149parenthesis (without the backslash). Spaces and other delimiters must be
12150prefixed with a backslash ('\') to avoid confusion with a field delimiter.
12151Other characters may be prefixed with a backslash to change their meaning :
12152
12153 \t for a tab
12154 \r for a carriage return (CR)
12155 \n for a new line (LF)
12156 \ to mark a space and differentiate it from a delimiter
12157 \# to mark a sharp and differentiate it from a comment
12158 \\ to use a backslash in a regex
12159 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
12160 \xXX to write the ASCII hex code XX as in the C language
12161
12162The <replace> parameter contains the string to be used to replace the largest
12163portion of text matching the regex. It can make use of the special characters
12164above, and can reference a substring which is delimited by parenthesis in the
12165regex, by writing a backslash ('\') immediately followed by one digit from 0 to
121669 indicating the group position (0 designating the entire line). This practice
12167is very common to users of the "sed" program.
12168
12169The <string> parameter represents the string which will systematically be added
12170after the last header line. It can also use special character sequences above.
12171
12172Notes related to these keywords :
12173---------------------------------
12174 - these keywords are not always convenient to allow/deny based on header
12175 contents. It is strongly recommended to use ACLs with the "block" keyword
12176 instead, resulting in far more flexible and manageable rules.
12177
12178 - lines are always considered as a whole. It is not possible to reference
12179 a header name only or a value only. This is important because of the way
12180 headers are written (notably the number of spaces after the colon).
12181
12182 - the first line is always considered as a header, which makes it possible to
12183 rewrite or filter HTTP requests URIs or response codes, but in turn makes
12184 it harder to distinguish between headers and request line. The regex prefix
12185 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
12186 ^[^ \t:]*: matches any header name followed by a colon.
12187
12188 - for performances reasons, the number of characters added to a request or to
12189 a response is limited at build time to values between 1 and 4 kB. This
12190 should normally be far more than enough for most usages. If it is too short
12191 on occasional usages, it is possible to gain some space by removing some
12192 useless headers before adding new ones.
12193
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012194 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012195 without the 'i' letter except that they ignore case when matching patterns.
12196
12197 - when a request passes through a frontend then a backend, all req* rules
12198 from the frontend will be evaluated, then all req* rules from the backend
12199 will be evaluated. The reverse path is applied to responses.
12200
12201 - req* statements are applied after "block" statements, so that "block" is
12202 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010012203 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012204
12205
Willy Tarreau74ca5042013-06-11 23:12:07 +0200122067. Using ACLs and fetching samples
12207----------------------------------
12208
Davor Ocelice9ed2812017-12-25 17:49:28 +010012209HAProxy is capable of extracting data from request or response streams, from
Willy Tarreau74ca5042013-06-11 23:12:07 +020012210client or server information, from tables, environmental information etc...
12211The action of extracting such data is called fetching a sample. Once retrieved,
12212these samples may be used for various purposes such as a key to a stick-table,
12213but most common usages consist in matching them against predefined constant
12214data called patterns.
12215
12216
122177.1. ACL basics
12218---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012219
12220The use of Access Control Lists (ACL) provides a flexible solution to perform
12221content switching and generally to take decisions based on content extracted
12222from the request, the response or any environmental status. The principle is
12223simple :
12224
Willy Tarreau74ca5042013-06-11 23:12:07 +020012225 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012226 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020012227 - apply one or multiple pattern matching methods on this sample
12228 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012229
Willy Tarreau74ca5042013-06-11 23:12:07 +020012230The actions generally consist in blocking a request, selecting a backend, or
12231adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012232
12233In order to define a test, the "acl" keyword is used. The syntax is :
12234
Willy Tarreau74ca5042013-06-11 23:12:07 +020012235 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012236
12237This creates a new ACL <aclname> or completes an existing one with new tests.
12238Those tests apply to the portion of request/response specified in <criterion>
12239and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012240an operator which may be specified before the set of values. Optionally some
12241conversion operators may be applied to the sample, and they will be specified
12242as a comma-delimited list of keywords just after the first keyword. The values
12243are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012244
12245ACL names must be formed from upper and lower case letters, digits, '-' (dash),
12246'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
12247which means that "my_acl" and "My_Acl" are two different ACLs.
12248
12249There is no enforced limit to the number of ACLs. The unused ones do not affect
12250performance, they just consume a small amount of memory.
12251
Willy Tarreau74ca5042013-06-11 23:12:07 +020012252The criterion generally is the name of a sample fetch method, or one of its ACL
12253specific declinations. The default test method is implied by the output type of
12254this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012255methods of a same sample fetch method. The sample fetch methods are the only
12256ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012257
12258Sample fetch methods return data which can be of the following types :
12259 - boolean
12260 - integer (signed or unsigned)
12261 - IPv4 or IPv6 address
12262 - string
12263 - data block
12264
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012265Converters transform any of these data into any of these. For example, some
12266converters might convert a string to a lower-case string while other ones
12267would turn a string to an IPv4 address, or apply a netmask to an IP address.
12268The resulting sample is of the type of the last converter applied to the list,
12269which defaults to the type of the sample fetch method.
12270
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012271Each sample or converter returns data of a specific type, specified with its
12272keyword in this documentation. When an ACL is declared using a standard sample
12273fetch method, certain types automatically involved a default matching method
12274which are summarized in the table below :
12275
12276 +---------------------+-----------------+
12277 | Sample or converter | Default |
12278 | output type | matching method |
12279 +---------------------+-----------------+
12280 | boolean | bool |
12281 +---------------------+-----------------+
12282 | integer | int |
12283 +---------------------+-----------------+
12284 | ip | ip |
12285 +---------------------+-----------------+
12286 | string | str |
12287 +---------------------+-----------------+
12288 | binary | none, use "-m" |
12289 +---------------------+-----------------+
12290
12291Note that in order to match a binary samples, it is mandatory to specify a
12292matching method, see below.
12293
Willy Tarreau74ca5042013-06-11 23:12:07 +020012294The ACL engine can match these types against patterns of the following types :
12295 - boolean
12296 - integer or integer range
12297 - IP address / network
12298 - string (exact, substring, suffix, prefix, subdir, domain)
12299 - regular expression
12300 - hex block
12301
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012302The following ACL flags are currently supported :
12303
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012304 -i : ignore case during matching of all subsequent patterns.
12305 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012306 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010012307 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010012308 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010012309 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020012310 -- : force end of flags. Useful when a string looks like one of the flags.
12311
Willy Tarreau74ca5042013-06-11 23:12:07 +020012312The "-f" flag is followed by the name of a file from which all lines will be
12313read as individual values. It is even possible to pass multiple "-f" arguments
12314if the patterns are to be loaded from multiple files. Empty lines as well as
12315lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
12316will be stripped. If it is absolutely necessary to insert a valid pattern
12317beginning with a sharp, just prefix it with a space so that it is not taken for
12318a comment. Depending on the data type and match method, haproxy may load the
12319lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
12320exact string matching. In this case, duplicates will automatically be removed.
12321
Thierry FOURNIER9860c412014-01-29 14:23:29 +010012322The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
12323parsed as two column file. The first column contains the patterns used by the
12324ACL, and the second column contain the samples. The sample can be used later by
12325a map. This can be useful in some rare cases where an ACL would just be used to
12326check for the existence of a pattern in a map before a mapping is applied.
12327
Thierry FOURNIER3534d882014-01-20 17:01:44 +010012328The "-u" flag forces the unique id of the ACL. This unique id is used with the
12329socket interface to identify ACL and dynamically change its values. Note that a
12330file is always identified by its name even if an id is set.
12331
Willy Tarreau74ca5042013-06-11 23:12:07 +020012332Also, note that the "-i" flag applies to subsequent entries and not to entries
12333loaded from files preceding it. For instance :
12334
12335 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
12336
12337In this example, each line of "exact-ua.lst" will be exactly matched against
12338the "user-agent" header of the request. Then each line of "generic-ua" will be
12339case-insensitively matched. Then the word "test" will be insensitively matched
12340as well.
12341
12342The "-m" flag is used to select a specific pattern matching method on the input
12343sample. All ACL-specific criteria imply a pattern matching method and generally
12344do not need this flag. However, this flag is useful with generic sample fetch
12345methods to describe how they're going to be matched against the patterns. This
12346is required for sample fetches which return data type for which there is no
Davor Ocelice9ed2812017-12-25 17:49:28 +010012347obvious matching method (e.g. string or binary). When "-m" is specified and
Willy Tarreau74ca5042013-06-11 23:12:07 +020012348followed by a pattern matching method name, this method is used instead of the
12349default one for the criterion. This makes it possible to match contents in ways
12350that were not initially planned, or with sample fetch methods which return a
12351string. The matching method also affects the way the patterns are parsed.
12352
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010012353The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
12354By default, if the parser cannot parse ip address it considers that the parsed
12355string is maybe a domain name and try dns resolution. The flag "-n" disable this
12356resolution. It is useful for detecting malformed ip lists. Note that if the DNS
12357server is not reachable, the haproxy configuration parsing may last many minutes
12358waiting fir the timeout. During this time no error messages are displayed. The
12359flag "-n" disable this behavior. Note also that during the runtime, this
12360function is disabled for the dynamic acl modifications.
12361
Willy Tarreau74ca5042013-06-11 23:12:07 +020012362There are some restrictions however. Not all methods can be used with all
12363sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
12364be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020012365
12366 - "found" : only check if the requested sample could be found in the stream,
12367 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020012368 to pass any pattern to avoid confusion. This matching method is
12369 particularly useful to detect presence of certain contents such
12370 as headers, cookies, etc... even if they are empty and without
12371 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012372
12373 - "bool" : check the value as a boolean. It can only be applied to fetches
12374 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012375 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012376
12377 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020012378 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012379
12380 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020012381 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012382
Davor Ocelice9ed2812017-12-25 17:49:28 +010012383 - "bin" : match the contents against a hexadecimal string representing a
Willy Tarreau5adeda12013-03-31 22:13:34 +020012384 binary sequence. This may be used with binary or string samples.
12385
12386 - "len" : match the sample's length as an integer. This may be used with
12387 binary or string samples.
12388
Willy Tarreau74ca5042013-06-11 23:12:07 +020012389 - "str" : exact match : match the contents against a string. This may be
12390 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012391
Willy Tarreau74ca5042013-06-11 23:12:07 +020012392 - "sub" : substring match : check that the contents contain at least one of
12393 the provided string patterns. This may be used with binary or
12394 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012395
Willy Tarreau74ca5042013-06-11 23:12:07 +020012396 - "reg" : regex match : match the contents against a list of regular
12397 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012398
Willy Tarreau74ca5042013-06-11 23:12:07 +020012399 - "beg" : prefix match : check that the contents begin like the provided
12400 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012401
Willy Tarreau74ca5042013-06-11 23:12:07 +020012402 - "end" : suffix match : check that the contents end like the provided
12403 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012404
Willy Tarreau74ca5042013-06-11 23:12:07 +020012405 - "dir" : subdir match : check that a slash-delimited portion of the
12406 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012407 This may be used with binary or string samples.
12408
Willy Tarreau74ca5042013-06-11 23:12:07 +020012409 - "dom" : domain match : check that a dot-delimited portion of the contents
12410 exactly match one of the provided string patterns. This may be
12411 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012412
12413For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
12414request, it is possible to do :
12415
12416 acl jsess_present cook(JSESSIONID) -m found
12417
12418In order to apply a regular expression on the 500 first bytes of data in the
12419buffer, one would use the following acl :
12420
12421 acl script_tag payload(0,500) -m reg -i <script>
12422
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012423On systems where the regex library is much slower when using "-i", it is
12424possible to convert the sample to lowercase before matching, like this :
12425
12426 acl script_tag payload(0,500),lower -m reg <script>
12427
Willy Tarreau74ca5042013-06-11 23:12:07 +020012428All ACL-specific criteria imply a default matching method. Most often, these
12429criteria are composed by concatenating the name of the original sample fetch
12430method and the matching method. For example, "hdr_beg" applies the "beg" match
12431to samples retrieved using the "hdr" fetch method. Since all ACL-specific
12432criteria rely on a sample fetch method, it is always possible instead to use
12433the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012434
Willy Tarreau74ca5042013-06-11 23:12:07 +020012435If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012436the matching method is simply applied to the underlying sample fetch method.
12437For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012438
Willy Tarreau74ca5042013-06-11 23:12:07 +020012439 acl short_form hdr_beg(host) www.
12440 acl alternate1 hdr_beg(host) -m beg www.
12441 acl alternate2 hdr_dom(host) -m beg www.
12442 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012443
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012444
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012445The table below summarizes the compatibility matrix between sample or converter
12446types and the pattern types to fetch against. It indicates for each compatible
12447combination the name of the matching method to be used, surrounded with angle
12448brackets ">" and "<" when the method is the default one and will work by
12449default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010012450
Willy Tarreau74ca5042013-06-11 23:12:07 +020012451 +-------------------------------------------------+
12452 | Input sample type |
12453 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012454 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012455 +----------------------+---------+---------+---------+---------+---------+
12456 | none (presence only) | found | found | found | found | found |
12457 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012458 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012459 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012460 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012461 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012462 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012463 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012464 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012465 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012466 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012467 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012468 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012469 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012470 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012471 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012472 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012473 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012474 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012475 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012476 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012477 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012478 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012479 +----------------------+---------+---------+---------+---------+---------+
12480 | hex block | | | | bin | bin |
12481 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020012482
12483
Willy Tarreau74ca5042013-06-11 23:12:07 +0200124847.1.1. Matching booleans
12485------------------------
12486
12487In order to match a boolean, no value is needed and all values are ignored.
12488Boolean matching is used by default for all fetch methods of type "boolean".
12489When boolean matching is used, the fetched value is returned as-is, which means
12490that a boolean "true" will always match and a boolean "false" will never match.
12491
12492Boolean matching may also be enforced using "-m bool" on fetch methods which
12493return an integer value. Then, integer value 0 is converted to the boolean
12494"false" and all other values are converted to "true".
12495
Willy Tarreau6a06a402007-07-15 20:15:28 +020012496
Willy Tarreau74ca5042013-06-11 23:12:07 +0200124977.1.2. Matching integers
12498------------------------
12499
12500Integer matching applies by default to integer fetch methods. It can also be
12501enforced on boolean fetches using "-m int". In this case, "false" is converted
12502to the integer 0, and "true" is converted to the integer 1.
12503
12504Integer matching also supports integer ranges and operators. Note that integer
12505matching only applies to positive values. A range is a value expressed with a
12506lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012507
12508For instance, "1024:65535" is a valid range to represent a range of
12509unprivileged ports, and "1024:" would also work. "0:1023" is a valid
12510representation of privileged ports, and ":1023" would also work.
12511
Willy Tarreau62644772008-07-16 18:36:06 +020012512As a special case, some ACL functions support decimal numbers which are in fact
12513two integers separated by a dot. This is used with some version checks for
12514instance. All integer properties apply to those decimal numbers, including
12515ranges and operators.
12516
Willy Tarreau6a06a402007-07-15 20:15:28 +020012517For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010012518operators with ranges does not make much sense and is strongly discouraged.
12519Similarly, it does not make much sense to perform order comparisons with a set
12520of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012521
Willy Tarreau0ba27502007-12-24 16:55:16 +010012522Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020012523
12524 eq : true if the tested value equals at least one value
12525 ge : true if the tested value is greater than or equal to at least one value
12526 gt : true if the tested value is greater than at least one value
12527 le : true if the tested value is less than or equal to at least one value
12528 lt : true if the tested value is less than at least one value
12529
Willy Tarreau0ba27502007-12-24 16:55:16 +010012530For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020012531
12532 acl negative-length hdr_val(content-length) lt 0
12533
Willy Tarreau62644772008-07-16 18:36:06 +020012534This one matches SSL versions between 3.0 and 3.1 (inclusive) :
12535
12536 acl sslv3 req_ssl_ver 3:3.1
12537
Willy Tarreau6a06a402007-07-15 20:15:28 +020012538
Willy Tarreau74ca5042013-06-11 23:12:07 +0200125397.1.3. Matching strings
12540-----------------------
12541
12542String matching applies to string or binary fetch methods, and exists in 6
12543different forms :
12544
12545 - exact match (-m str) : the extracted string must exactly match the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012546 patterns;
Willy Tarreau74ca5042013-06-11 23:12:07 +020012547
12548 - substring match (-m sub) : the patterns are looked up inside the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012549 extracted string, and the ACL matches if any of them is found inside;
Willy Tarreau74ca5042013-06-11 23:12:07 +020012550
12551 - prefix match (-m beg) : the patterns are compared with the beginning of
12552 the extracted string, and the ACL matches if any of them matches.
12553
12554 - suffix match (-m end) : the patterns are compared with the end of the
12555 extracted string, and the ACL matches if any of them matches.
12556
Baptiste Assmann33db6002016-03-06 23:32:10 +010012557 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020012558 string, delimited with slashes ("/"), and the ACL matches if any of them
12559 matches.
12560
12561 - domain match (-m dom) : the patterns are looked up inside the extracted
12562 string, delimited with dots ("."), and the ACL matches if any of them
12563 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012564
12565String matching applies to verbatim strings as they are passed, with the
12566exception of the backslash ("\") which makes it possible to escape some
12567characters such as the space. If the "-i" flag is passed before the first
12568string, then the matching will be performed ignoring the case. In order
12569to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010012570before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020012571
12572
Willy Tarreau74ca5042013-06-11 23:12:07 +0200125737.1.4. Matching regular expressions (regexes)
12574---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020012575
12576Just like with string matching, regex matching applies to verbatim strings as
12577they are passed, with the exception of the backslash ("\") which makes it
12578possible to escape some characters such as the space. If the "-i" flag is
12579passed before the first regex, then the matching will be performed ignoring
12580the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010012581the "--" flag before the first string. Same principle applies of course to
12582match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020012583
12584
Willy Tarreau74ca5042013-06-11 23:12:07 +0200125857.1.5. Matching arbitrary data blocks
12586-------------------------------------
12587
12588It is possible to match some extracted samples against a binary block which may
12589not safely be represented as a string. For this, the patterns must be passed as
12590a series of hexadecimal digits in an even number, when the match method is set
12591to binary. Each sequence of two digits will represent a byte. The hexadecimal
12592digits may be used upper or lower case.
12593
12594Example :
12595 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
12596 acl hello payload(0,6) -m bin 48656c6c6f0a
12597
12598
125997.1.6. Matching IPv4 and IPv6 addresses
12600---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020012601
12602IPv4 addresses values can be specified either as plain addresses or with a
12603netmask appended, in which case the IPv4 address matches whenever it is
12604within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010012605host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010012606difficult to read and debug configurations. If hostnames are used, you should
12607at least ensure that they are present in /etc/hosts so that the configuration
12608does not depend on any random DNS match at the moment the configuration is
12609parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012610
Daniel Schnellereba56342016-04-13 00:26:52 +020012611The dotted IPv4 address notation is supported in both regular as well as the
12612abbreviated form with all-0-octets omitted:
12613
12614 +------------------+------------------+------------------+
12615 | Example 1 | Example 2 | Example 3 |
12616 +------------------+------------------+------------------+
12617 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
12618 | 192.168.1 | 10.12 | 127.1 |
12619 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
12620 | 192.168.1/22 | 10.12/8 | 127.1/8 |
12621 +------------------+------------------+------------------+
12622
12623Notice that this is different from RFC 4632 CIDR address notation in which
12624192.168.42/24 would be equivalent to 192.168.42.0/24.
12625
Willy Tarreauceb4ac92012-04-28 00:41:46 +020012626IPv6 may be entered in their usual form, with or without a netmask appended.
12627Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
12628trouble with randomly resolved IP addresses, host names are never allowed in
12629IPv6 patterns.
12630
12631HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
12632following situations :
12633 - tested address is IPv4, pattern address is IPv4, the match applies
12634 in IPv4 using the supplied mask if any.
12635 - tested address is IPv6, pattern address is IPv6, the match applies
12636 in IPv6 using the supplied mask if any.
12637 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
12638 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
12639 ::IPV4 or ::ffff:IPV4, otherwise it fails.
12640 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
12641 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
12642 applied in IPv6 using the supplied IPv6 mask.
12643
Willy Tarreau74ca5042013-06-11 23:12:07 +020012644
126457.2. Using ACLs to form conditions
12646----------------------------------
12647
12648Some actions are only performed upon a valid condition. A condition is a
12649combination of ACLs with operators. 3 operators are supported :
12650
12651 - AND (implicit)
12652 - OR (explicit with the "or" keyword or the "||" operator)
12653 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020012654
Willy Tarreau74ca5042013-06-11 23:12:07 +020012655A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020012656
Willy Tarreau74ca5042013-06-11 23:12:07 +020012657 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020012658
Willy Tarreau74ca5042013-06-11 23:12:07 +020012659Such conditions are generally used after an "if" or "unless" statement,
12660indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020012661
Willy Tarreau74ca5042013-06-11 23:12:07 +020012662For instance, to block HTTP requests to the "*" URL with methods other than
12663"OPTIONS", as well as POST requests without content-length, and GET or HEAD
12664requests with a content-length greater than 0, and finally every request which
12665is not either GET/HEAD/POST/OPTIONS !
12666
12667 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030012668 http-request deny if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
12669 http-request deny if METH_GET HTTP_CONTENT
12670 http-request deny unless METH_GET or METH_POST or METH_OPTIONS
Willy Tarreau74ca5042013-06-11 23:12:07 +020012671
12672To select a different backend for requests to static contents on the "www" site
12673and to every request on the "img", "video", "download" and "ftp" hosts :
12674
12675 acl url_static path_beg /static /images /img /css
12676 acl url_static path_end .gif .png .jpg .css .js
12677 acl host_www hdr_beg(host) -i www
12678 acl host_static hdr_beg(host) -i img. video. download. ftp.
12679
Davor Ocelice9ed2812017-12-25 17:49:28 +010012680 # now use backend "static" for all static-only hosts, and for static URLs
Willy Tarreau74ca5042013-06-11 23:12:07 +020012681 # of host "www". Use backend "www" for the rest.
12682 use_backend static if host_static or host_www url_static
12683 use_backend www if host_www
12684
12685It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
12686expressions that are built on the fly without needing to be declared. They must
12687be enclosed between braces, with a space before and after each brace (because
12688the braces must be seen as independent words). Example :
12689
12690 The following rule :
12691
12692 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030012693 http-request deny if METH_POST missing_cl
Willy Tarreau74ca5042013-06-11 23:12:07 +020012694
12695 Can also be written that way :
12696
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030012697 http-request deny if METH_POST { hdr_cnt(Content-length) eq 0 }
Willy Tarreau74ca5042013-06-11 23:12:07 +020012698
12699It is generally not recommended to use this construct because it's a lot easier
12700to leave errors in the configuration when written that way. However, for very
12701simple rules matching only one source IP address for instance, it can make more
12702sense to use them than to declare ACLs with random names. Another example of
12703good use is the following :
12704
12705 With named ACLs :
12706
12707 acl site_dead nbsrv(dynamic) lt 2
12708 acl site_dead nbsrv(static) lt 2
12709 monitor fail if site_dead
12710
12711 With anonymous ACLs :
12712
12713 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
12714
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030012715See section 4.2 for detailed help on the "http-request deny" and "use_backend"
12716keywords.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012717
12718
127197.3. Fetching samples
12720---------------------
12721
12722Historically, sample fetch methods were only used to retrieve data to match
12723against patterns using ACLs. With the arrival of stick-tables, a new class of
12724sample fetch methods was created, most often sharing the same syntax as their
12725ACL counterpart. These sample fetch methods are also known as "fetches". As
12726of now, ACLs and fetches have converged. All ACL fetch methods have been made
12727available as fetch methods, and ACLs may use any sample fetch method as well.
12728
12729This section details all available sample fetch methods and their output type.
12730Some sample fetch methods have deprecated aliases that are used to maintain
12731compatibility with existing configurations. They are then explicitly marked as
12732deprecated and should not be used in new setups.
12733
12734The ACL derivatives are also indicated when available, with their respective
12735matching methods. These ones all have a well defined default pattern matching
12736method, so it is never necessary (though allowed) to pass the "-m" option to
12737indicate how the sample will be matched using ACLs.
12738
12739As indicated in the sample type versus matching compatibility matrix above,
12740when using a generic sample fetch method in an ACL, the "-m" option is
12741mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
12742the same keyword exists as an ACL keyword and as a standard fetch method, the
12743ACL engine will automatically pick the ACL-only one by default.
12744
12745Some of these keywords support one or multiple mandatory arguments, and one or
12746multiple optional arguments. These arguments are strongly typed and are checked
12747when the configuration is parsed so that there is no risk of running with an
Davor Ocelice9ed2812017-12-25 17:49:28 +010012748incorrect argument (e.g. an unresolved backend name). Fetch function arguments
12749are passed between parenthesis and are delimited by commas. When an argument
Willy Tarreau74ca5042013-06-11 23:12:07 +020012750is optional, it will be indicated below between square brackets ('[ ]'). When
12751all arguments are optional, the parenthesis may be omitted.
12752
12753Thus, the syntax of a standard sample fetch method is one of the following :
12754 - name
12755 - name(arg1)
12756 - name(arg1,arg2)
12757
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012758
127597.3.1. Converters
12760-----------------
12761
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012762Sample fetch methods may be combined with transformations to be applied on top
12763of the fetched sample (also called "converters"). These combinations form what
12764is called "sample expressions" and the result is a "sample". Initially this
12765was only supported by "stick on" and "stick store-request" directives but this
Davor Ocelice9ed2812017-12-25 17:49:28 +010012766has now be extended to all places where samples may be used (ACLs, log-format,
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012767unique-id-format, add-header, ...).
12768
12769These transformations are enumerated as a series of specific keywords after the
12770sample fetch method. These keywords may equally be appended immediately after
12771the fetch keyword's argument, delimited by a comma. These keywords can also
Davor Ocelice9ed2812017-12-25 17:49:28 +010012772support some arguments (e.g. a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010012773
Willy Tarreau97707872015-01-27 15:12:13 +010012774A certain category of converters are bitwise and arithmetic operators which
12775support performing basic operations on integers. Some bitwise operations are
12776supported (and, or, xor, cpl) and some arithmetic operations are supported
12777(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
12778bool) which make it possible to report a match without having to write an ACL.
12779
Willy Tarreau74ca5042013-06-11 23:12:07 +020012780The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010012781
Ben Shillitof25e8e52016-12-02 14:25:37 +00001278251d.single(<prop>[,<prop>*])
12783 Returns values for the properties requested as a string, where values are
12784 separated by the delimiter specified with "51degrees-property-separator".
12785 The device is identified using the User-Agent header passed to the
12786 converter. The function can be passed up to five property names, and if a
12787 property name can't be found, the value "NoData" is returned.
12788
12789 Example :
Davor Ocelice9ed2812017-12-25 17:49:28 +010012790 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request,
12791 # containing values for the three properties requested by using the
Ben Shillitof25e8e52016-12-02 14:25:37 +000012792 # User-Agent passed to the converter.
12793 frontend http-in
12794 bind *:8081
12795 default_backend servers
12796 http-request set-header X-51D-DeviceTypeMobileTablet \
12797 %[req.fhdr(User-Agent),51d.single(DeviceType,IsMobile,IsTablet)]
12798
Willy Tarreau97707872015-01-27 15:12:13 +010012799add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012800 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012801 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012802 name. The name of the variable starts with an indication about its scope. The
12803 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012804 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012805 "sess" : the variable is shared with the whole session
12806 "txn" : the variable is shared with the transaction (request and response)
12807 "req" : the variable is shared only during request processing
12808 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012809 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012810 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012811
12812and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012813 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012814 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010012815 numeric value or a variable name. The name of the variable starts with an
12816 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012817 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012818 "sess" : the variable is shared with the whole session
12819 "txn" : the variable is shared with the transaction (request and response)
12820 "req" : the variable is shared only during request processing
12821 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012822 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012823 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012824
Holger Just1bfc24b2017-05-06 00:56:53 +020012825b64dec
12826 Converts (decodes) a base64 encoded input string to its binary
12827 representation. It performs the inverse operation of base64().
12828
Emeric Brun53d1a982014-04-30 18:21:37 +020012829base64
12830 Converts a binary input sample to a base64 string. It is used to log or
Davor Ocelice9ed2812017-12-25 17:49:28 +010012831 transfer binary content in a way that can be reliably transferred (e.g.
Emeric Brun53d1a982014-04-30 18:21:37 +020012832 an SSL ID can be copied in a header).
12833
Willy Tarreau97707872015-01-27 15:12:13 +010012834bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012835 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010012836 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010012837 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010012838 presence of a flag).
12839
Emeric Brun54c4ac82014-11-03 15:32:43 +010012840bytes(<offset>[,<length>])
12841 Extracts some bytes from an input binary sample. The result is a binary
12842 sample starting at an offset (in bytes) of the original sample and
Tim Düsterhus4896c442016-11-29 02:15:19 +010012843 optionally truncated at the given length.
Emeric Brun54c4ac82014-11-03 15:32:43 +010012844
Willy Tarreau280f42b2018-02-19 15:34:12 +010012845concat([<start>],[<var>],[<end>])
12846 Concatenates up to 3 fields after the current sample which is then turned to
12847 a string. The first one, <start>, is a constant string, that will be appended
12848 immediately after the existing sample. It may be omitted if not used. The
12849 second one, <var>, is a variable name. The variable will be looked up, its
12850 contents converted to a string, and it will be appended immediately after the
12851 <first> part. If the variable is not found, nothing is appended. It may be
12852 omitted as well. The third field, <end> is a constant string that will be
12853 appended after the variable. It may also be omitted. Together, these elements
12854 allow to concatenate variables with delimiters to an existing set of
12855 variables. This can be used to build new variables made of a succession of
12856 other variables, such as colon-delimited varlues. Note that due to the config
12857 parser, it is not possible to use a comma nor a closing parenthesis as
12858 delimitors.
12859
12860 Example:
12861 tcp-request session set-var(sess.src) src
12862 tcp-request session set-var(sess.dn) ssl_c_s_dn
12863 tcp-request session set-var(txn.sig) str(),concat(<ip=,sess.ip,>),concat(<dn=,sess.dn,>)
12864 http-request set-header x-hap-sig %[var(txn.sig)]
12865
Willy Tarreau97707872015-01-27 15:12:13 +010012866cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012867 Takes the input value of type signed integer, applies a ones-complement
12868 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010012869
Willy Tarreau80599772015-01-20 19:35:24 +010012870crc32([<avalanche>])
12871 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
12872 hash function. Optionally, it is possible to apply a full avalanche hash
12873 function to the output if the optional <avalanche> argument equals 1. This
12874 converter uses the same functions as used by the various hash-based load
12875 balancing algorithms, so it will provide exactly the same results. It is
12876 provided for compatibility with other software which want a CRC32 to be
12877 computed on some input keys, so it follows the most common implementation as
12878 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
12879 but may provide a better or at least less predictable distribution. It must
12880 not be used for security purposes as a 32-bit hash is trivial to break. See
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010012881 also "djb2", "sdbm", "wt6", "crc32c" and the "hash-type" directive.
12882
12883crc32c([<avalanche>])
12884 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32C
12885 hash function. Optionally, it is possible to apply a full avalanche hash
12886 function to the output if the optional <avalanche> argument equals 1. This
12887 converter uses the same functions as described in RFC4960, Appendix B [8].
12888 It is provided for compatibility with other software which want a CRC32C to be
12889 computed on some input keys. It is slower than the other algorithms and it must
12890 not be used for security purposes as a 32-bit hash is trivial to break. See
12891 also "djb2", "sdbm", "wt6", "crc32" and the "hash-type" directive.
Willy Tarreau80599772015-01-20 19:35:24 +010012892
David Carlier29b3ca32015-09-25 14:09:21 +010012893da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020012894 Asks the DeviceAtlas converter to identify the User Agent string passed on
12895 input, and to emit a string made of the concatenation of the properties
12896 enumerated in argument, delimited by the separator defined by the global
12897 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000012898 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020012899 configuration language.
12900
12901 Example:
12902 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020012903 bind *:8881
12904 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000012905 http-request set-header X-DeviceAtlas-Data %[req.fhdr(User-Agent),da-csv(primaryHardwareType,osName,osVersion,browserName,browserVersion,browserRenderingEngine)]
David Carlier4542b102015-06-01 13:54:29 +020012906
Thierry FOURNIER9687c772015-05-07 15:46:29 +020012907debug
12908 This converter is used as debug tool. It dumps on screen the content and the
12909 type of the input sample. The sample is returned as is on its output. This
12910 converter only exists when haproxy was built with debugging enabled.
12911
Willy Tarreau97707872015-01-27 15:12:13 +010012912div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012913 Divides the input value of type signed integer by <value>, and returns the
12914 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012915 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010012916 variable name. The name of the variable starts with an indication about its
12917 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012918 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012919 "sess" : the variable is shared with the whole session
12920 "txn" : the variable is shared with the transaction (request and response)
12921 "req" : the variable is shared only during request processing
12922 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012923 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012924 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012925
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012926djb2([<avalanche>])
12927 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
12928 hash function. Optionally, it is possible to apply a full avalanche hash
12929 function to the output if the optional <avalanche> argument equals 1. This
12930 converter uses the same functions as used by the various hash-based load
12931 balancing algorithms, so it will provide exactly the same results. It is
12932 mostly intended for debugging, but can be used as a stick-table entry to
12933 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010012934 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6", "crc32c",
12935 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012936
Willy Tarreau97707872015-01-27 15:12:13 +010012937even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012938 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010012939 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
12940
Marcin Deranek9631a282018-04-16 14:30:46 +020012941field(<index>,<delimiters>[,<count>])
12942 Extracts the substring at the given index counting from the beginning
12943 (positive index) or from the end (negative index) considering given delimiters
12944 from an input string. Indexes start at 1 or -1 and delimiters are a string
12945 formatted list of chars. Optionally you can specify <count> of fields to
12946 extract (default: 1). Value of 0 indicates extraction of all remaining
12947 fields.
12948
12949 Example :
12950 str(f1_f2_f3__f5),field(5,_) # f5
12951 str(f1_f2_f3__f5),field(2,_,0) # f2_f3__f5
12952 str(f1_f2_f3__f5),field(2,_,2) # f2_f3
12953 str(f1_f2_f3__f5),field(-2,_,3) # f2_f3_
12954 str(f1_f2_f3__f5),field(-3,_,0) # f1_f2_f3
Emeric Brunf399b0d2014-11-03 17:07:03 +010012955
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012956hex
Davor Ocelice9ed2812017-12-25 17:49:28 +010012957 Converts a binary input sample to a hex string containing two hex digits per
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012958 input byte. It is used to log or transfer hex dumps of some binary input data
Davor Ocelice9ed2812017-12-25 17:49:28 +010012959 in a way that can be reliably transferred (e.g. an SSL ID can be copied in a
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012960 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010012961
Dragan Dosen3f957b22017-10-24 09:27:34 +020012962hex2i
12963 Converts a hex string containing two hex digits per input byte to an
12964 integer. If the input value can not be converted, then zero is returned.
12965
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012966http_date([<offset>])
12967 Converts an integer supposed to contain a date since epoch to a string
12968 representing this date in a format suitable for use in HTTP header fields. If
12969 an offset value is specified, then it is a number of seconds that is added to
12970 the date before the conversion is operated. This is particularly useful to
12971 emit Date header fields, Expires values in responses when combined with a
12972 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012973
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012974in_table(<table>)
12975 Uses the string representation of the input sample to perform a look up in
12976 the specified table. If the key is not found in the table, a boolean false
12977 is returned. Otherwise a boolean true is returned. This can be used to verify
Davor Ocelice9ed2812017-12-25 17:49:28 +010012978 the presence of a certain key in a table tracking some elements (e.g. whether
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012979 or not a source IP address or an Authorization header was already seen).
12980
Tim Duesterhus1478aa72018-01-25 16:24:51 +010012981ipmask(<mask4>, [<mask6>])
12982 Apply a mask to an IP address, and use the result for lookups and storage.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012983 This can be used to make all hosts within a certain mask to share the same
Tim Duesterhus1478aa72018-01-25 16:24:51 +010012984 table entries and as such use the same server. The mask4 can be passed in
12985 dotted form (e.g. 255.255.255.0) or in CIDR form (e.g. 24). The mask6 can
12986 be passed in quadruplet form (e.g. ffff:ffff::) or in CIDR form (e.g. 64).
12987 If no mask6 is given IPv6 addresses will fail to convert for backwards
12988 compatibility reasons.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012989
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020012990json([<input-code>])
Davor Ocelice9ed2812017-12-25 17:49:28 +010012991 Escapes the input string and produces an ASCII output string ready to use as a
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020012992 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020012993 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020012994 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
12995 of errors:
12996 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
12997 bytes, ...)
12998 - invalid range (the decoded value is within a UTF-8 prohibited range),
12999 - code overlong (the value is encoded with more bytes than necessary).
13000
13001 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
13002 character is greater than 0xffff because the JSON string escape specification
13003 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
13004 in 4 variants designated by a combination of two suffix letters : "p" for
13005 "permissive" and "s" for "silently ignore". The behaviors of the decoders
13006 are :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013007 - "ascii" : never fails;
13008 - "utf8" : fails on any detected errors;
13009 - "utf8s" : never fails, but removes characters corresponding to errors;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013010 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
Davor Ocelice9ed2812017-12-25 17:49:28 +010013011 error;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013012 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
13013 characters corresponding to the other errors.
13014
13015 This converter is particularly useful for building properly escaped JSON for
Davor Ocelice9ed2812017-12-25 17:49:28 +010013016 logging to servers which consume JSON-formatted traffic logs.
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013017
13018 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013019 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013020 capture request header user-agent len 150
13021 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013022
13023 Input request from client 127.0.0.1:
13024 GET / HTTP/1.0
13025 User-Agent: Very "Ugly" UA 1/2
13026
13027 Output log:
13028 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
13029
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013030language(<value>[,<default>])
13031 Returns the value with the highest q-factor from a list as extracted from the
13032 "accept-language" header using "req.fhdr". Values with no q-factor have a
13033 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
13034 belong to the list of semi-colon delimited <values> will be considered. The
13035 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
13036 given list and a default value is provided, it is returned. Note that language
13037 names may have a variant after a dash ('-'). If this variant is present in the
13038 list, it will be matched, but if it is not, only the base language is checked.
13039 The match is case-sensitive, and the output string is always one of those
Davor Ocelice9ed2812017-12-25 17:49:28 +010013040 provided in arguments. The ordering of arguments is meaningless, only the
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013041 ordering of the values in the request counts, as the first value among
13042 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020013043
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013044 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013045
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013046 # this configuration switches to the backend matching a
13047 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013048
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013049 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
13050 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
13051 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
13052 use_backend spanish if es
13053 use_backend french if fr
13054 use_backend english if en
13055 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020013056
Willy Tarreau60a2ee72017-12-15 07:13:48 +010013057length
Etienne Carriereed0d24e2017-12-13 13:41:34 +010013058 Get the length of the string. This can only be placed after a string
13059 sample fetch function or after a transformation keyword returning a string
13060 type. The result is of type integer.
13061
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013062lower
13063 Convert a string sample to lower case. This can only be placed after a string
13064 sample fetch function or after a transformation keyword returning a string
13065 type. The result is of type string.
13066
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013067ltime(<format>[,<offset>])
13068 Converts an integer supposed to contain a date since epoch to a string
13069 representing this date in local time using a format defined by the <format>
13070 string using strftime(3). The purpose is to allow any date format to be used
13071 in logs. An optional <offset> in seconds may be applied to the input date
13072 (positive or negative). See the strftime() man page for the format supported
13073 by your operating system. See also the utime converter.
13074
13075 Example :
13076
13077 # Emit two colons, one with the local time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010013078 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013079 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
13080
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013081map(<map_file>[,<default_value>])
13082map_<match_type>(<map_file>[,<default_value>])
13083map_<match_type>_<output_type>(<map_file>[,<default_value>])
13084 Search the input value from <map_file> using the <match_type> matching method,
13085 and return the associated value converted to the type <output_type>. If the
13086 input value cannot be found in the <map_file>, the converter returns the
13087 <default_value>. If the <default_value> is not set, the converter fails and
13088 acts as if no input value could be fetched. If the <match_type> is not set, it
13089 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
13090 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
13091 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013092
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013093 It is important to avoid overlapping between the keys : IP addresses and
13094 strings are stored in trees, so the first of the finest match will be used.
13095 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013096
Tim Düsterhus4896c442016-11-29 02:15:19 +010013097 The following array contains the list of all map functions available sorted by
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013098 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013099
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013100 input type | match method | output type str | output type int | output type ip
13101 -----------+--------------+-----------------+-----------------+---------------
13102 str | str | map_str | map_str_int | map_str_ip
13103 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020013104 str | beg | map_beg | map_beg_int | map_end_ip
13105 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013106 str | sub | map_sub | map_sub_int | map_sub_ip
13107 -----------+--------------+-----------------+-----------------+---------------
13108 str | dir | map_dir | map_dir_int | map_dir_ip
13109 -----------+--------------+-----------------+-----------------+---------------
13110 str | dom | map_dom | map_dom_int | map_dom_ip
13111 -----------+--------------+-----------------+-----------------+---------------
13112 str | end | map_end | map_end_int | map_end_ip
13113 -----------+--------------+-----------------+-----------------+---------------
Ruoshan Huang3c5e3742016-12-02 16:25:31 +080013114 str | reg | map_reg | map_reg_int | map_reg_ip
13115 -----------+--------------+-----------------+-----------------+---------------
13116 str | reg | map_regm | map_reg_int | map_reg_ip
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013117 -----------+--------------+-----------------+-----------------+---------------
13118 int | int | map_int | map_int_int | map_int_ip
13119 -----------+--------------+-----------------+-----------------+---------------
13120 ip | ip | map_ip | map_ip_int | map_ip_ip
13121 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013122
Thierry Fournier8feaa662016-02-10 22:55:20 +010013123 The special map called "map_regm" expect matching zone in the regular
13124 expression and modify the output replacing back reference (like "\1") by
13125 the corresponding match text.
13126
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013127 The file contains one key + value per line. Lines which start with '#' are
13128 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
13129 is then the first "word" (series of non-space/tabs characters), and the value
13130 is what follows this series of space/tab till the end of the line excluding
13131 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013132
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013133 Example :
13134
13135 # this is a comment and is ignored
13136 2.22.246.0/23 United Kingdom \n
13137 <-><-----------><--><------------><---->
13138 | | | | `- trailing spaces ignored
13139 | | | `---------- value
13140 | | `-------------------- middle spaces ignored
13141 | `---------------------------- key
13142 `------------------------------------ leading spaces ignored
13143
Willy Tarreau97707872015-01-27 15:12:13 +010013144mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013145 Divides the input value of type signed integer by <value>, and returns the
13146 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013147 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013148 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013149 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013150 "sess" : the variable is shared with the whole session
13151 "txn" : the variable is shared with the transaction (request and response)
13152 "req" : the variable is shared only during request processing
13153 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013154 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013155 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013156
13157mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013158 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020013159 the product as an signed integer. In case of overflow, the largest possible
13160 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013161 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013162 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013163 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013164 "sess" : the variable is shared with the whole session
13165 "txn" : the variable is shared with the transaction (request and response)
13166 "req" : the variable is shared only during request processing
13167 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013168 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013169 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013170
Nenad Merdanovicb7e7c472017-03-12 21:56:55 +010013171nbsrv
13172 Takes an input value of type string, interprets it as a backend name and
13173 returns the number of usable servers in that backend. Can be used in places
13174 where we want to look up a backend from a dynamic name, like a result of a
13175 map lookup.
13176
Willy Tarreau97707872015-01-27 15:12:13 +010013177neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013178 Takes the input value of type signed integer, computes the opposite value,
13179 and returns the remainder as an signed integer. 0 is identity. This operator
13180 is provided for reversed subtracts : in order to subtract the input from a
13181 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010013182
13183not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013184 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013185 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013186 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013187 absence of a flag).
13188
13189odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013190 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010013191 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
13192
13193or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013194 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013195 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013196 numeric value or a variable name. The name of the variable starts with an
13197 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013198 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013199 "sess" : the variable is shared with the whole session
13200 "txn" : the variable is shared with the transaction (request and response)
13201 "req" : the variable is shared only during request processing
13202 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013203 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013204 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013205
Willy Tarreauc4dc3502015-01-23 20:39:28 +010013206regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010013207 Applies a regex-based substitution to the input string. It does the same
13208 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
13209 default it will replace in the input string the first occurrence of the
13210 largest part matching the regular expression <regex> with the substitution
13211 string <subst>. It is possible to replace all occurrences instead by adding
13212 the flag "g" in the third argument <flags>. It is also possible to make the
13213 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
13214 string, it is made up from the concatenation of all desired flags. Thus if
13215 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
13216 It is important to note that due to the current limitations of the
Baptiste Assmann66025d82016-03-06 23:36:48 +010013217 configuration parser, some characters such as closing parenthesis, closing
13218 square brackets or comma are not possible to use in the arguments. The first
13219 use of this converter is to replace certain characters or sequence of
13220 characters with other ones.
Willy Tarreau7eda8492015-01-20 19:47:06 +010013221
13222 Example :
13223
13224 # de-duplicate "/" in header "x-path".
13225 # input: x-path: /////a///b/c/xzxyz/
13226 # output: x-path: /a/b/c/xzxyz/
13227 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
13228
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013229capture-req(<id>)
13230 Capture the string entry in the request slot <id> and returns the entry as
13231 is. If the slot doesn't exist, the capture fails silently.
13232
13233 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020013234 "http-response capture", "capture.req.hdr" and
13235 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013236
13237capture-res(<id>)
13238 Capture the string entry in the response slot <id> and returns the entry as
13239 is. If the slot doesn't exist, the capture fails silently.
13240
13241 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020013242 "http-response capture", "capture.req.hdr" and
13243 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013244
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013245sdbm([<avalanche>])
13246 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
13247 hash function. Optionally, it is possible to apply a full avalanche hash
13248 function to the output if the optional <avalanche> argument equals 1. This
13249 converter uses the same functions as used by the various hash-based load
13250 balancing algorithms, so it will provide exactly the same results. It is
13251 mostly intended for debugging, but can be used as a stick-table entry to
13252 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013253 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6", "crc32c",
13254 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013255
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013256set-var(<var name>)
Davor Ocelice9ed2812017-12-25 17:49:28 +010013257 Sets a variable with the input content and returns the content on the output
13258 as-is. The variable keeps the value and the associated input type. The name of
13259 the variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013260 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013261 "sess" : the variable is shared with the whole session
13262 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013263 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010013264 "req" : the variable is shared only during request processing,
13265 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013266 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013267 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013268
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020013269sha1
13270 Converts a binary input sample to a SHA1 digest. The result is a binary
13271 sample with length of 20 bytes.
13272
Tim Duesterhusca097c12018-04-27 21:18:45 +020013273strcmp(<var>)
13274 Compares the contents of <var> with the input value of type string. Returns
13275 the result as a signed integer compatible with strcmp(3): 0 if both strings
13276 are identical. A value less than 0 if the left string is lexicographically
13277 smaller than the right string or if the left string is shorter. A value greater
13278 than 0 otherwise (right string greater than left string or the right string is
13279 shorter).
13280
13281 Example :
13282
13283 http-request set-var(txn.host) hdr(host)
13284 # Check whether the client is attempting domain fronting.
13285 acl ssl_sni_http_host_match ssl_fc_sni,strcmp(txn.host) eq 0
13286
13287
Willy Tarreau97707872015-01-27 15:12:13 +010013288sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013289 Subtracts <value> from the input value of type signed integer, and returns
13290 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013291 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010013292 or a variable name. The name of the variable starts with an indication about
13293 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013294 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013295 "sess" : the variable is shared with the whole session
13296 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013297 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010013298 "req" : the variable is shared only during request processing,
13299 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013300 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013301 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013302
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013303table_bytes_in_rate(<table>)
13304 Uses the string representation of the input sample to perform a look up in
13305 the specified table. If the key is not found in the table, integer value zero
13306 is returned. Otherwise the converter returns the average client-to-server
13307 bytes rate associated with the input sample in the designated table, measured
13308 in amount of bytes over the period configured in the table. See also the
13309 sc_bytes_in_rate sample fetch keyword.
13310
13311
13312table_bytes_out_rate(<table>)
13313 Uses the string representation of the input sample to perform a look up in
13314 the specified table. If the key is not found in the table, integer value zero
13315 is returned. Otherwise the converter returns the average server-to-client
13316 bytes rate associated with the input sample in the designated table, measured
13317 in amount of bytes over the period configured in the table. See also the
13318 sc_bytes_out_rate sample fetch keyword.
13319
13320table_conn_cnt(<table>)
13321 Uses the string representation of the input sample to perform a look up in
13322 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013323 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013324 connections associated with the input sample in the designated table. See
13325 also the sc_conn_cnt sample fetch keyword.
13326
13327table_conn_cur(<table>)
13328 Uses the string representation of the input sample to perform a look up in
13329 the specified table. If the key is not found in the table, integer value zero
13330 is returned. Otherwise the converter returns the current amount of concurrent
13331 tracked connections associated with the input sample in the designated table.
13332 See also the sc_conn_cur sample fetch keyword.
13333
13334table_conn_rate(<table>)
13335 Uses the string representation of the input sample to perform a look up in
13336 the specified table. If the key is not found in the table, integer value zero
13337 is returned. Otherwise the converter returns the average incoming connection
13338 rate associated with the input sample in the designated table. See also the
13339 sc_conn_rate sample fetch keyword.
13340
Thierry FOURNIER236657b2015-08-19 08:25:14 +020013341table_gpt0(<table>)
13342 Uses the string representation of the input sample to perform a look up in
13343 the specified table. If the key is not found in the table, boolean value zero
13344 is returned. Otherwise the converter returns the current value of the first
13345 general purpose tag associated with the input sample in the designated table.
13346 See also the sc_get_gpt0 sample fetch keyword.
13347
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013348table_gpc0(<table>)
13349 Uses the string representation of the input sample to perform a look up in
13350 the specified table. If the key is not found in the table, integer value zero
13351 is returned. Otherwise the converter returns the current value of the first
13352 general purpose counter associated with the input sample in the designated
13353 table. See also the sc_get_gpc0 sample fetch keyword.
13354
13355table_gpc0_rate(<table>)
13356 Uses the string representation of the input sample to perform a look up in
13357 the specified table. If the key is not found in the table, integer value zero
13358 is returned. Otherwise the converter returns the frequency which the gpc0
13359 counter was incremented over the configured period in the table, associated
13360 with the input sample in the designated table. See also the sc_get_gpc0_rate
13361 sample fetch keyword.
13362
Frédéric Lécaille6778b272018-01-29 15:22:53 +010013363table_gpc1(<table>)
13364 Uses the string representation of the input sample to perform a look up in
13365 the specified table. If the key is not found in the table, integer value zero
13366 is returned. Otherwise the converter returns the current value of the second
13367 general purpose counter associated with the input sample in the designated
13368 table. See also the sc_get_gpc1 sample fetch keyword.
13369
13370table_gpc1_rate(<table>)
13371 Uses the string representation of the input sample to perform a look up in
13372 the specified table. If the key is not found in the table, integer value zero
13373 is returned. Otherwise the converter returns the frequency which the gpc1
13374 counter was incremented over the configured period in the table, associated
13375 with the input sample in the designated table. See also the sc_get_gpc1_rate
13376 sample fetch keyword.
13377
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013378table_http_err_cnt(<table>)
13379 Uses the string representation of the input sample to perform a look up in
13380 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013381 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013382 errors associated with the input sample in the designated table. See also the
13383 sc_http_err_cnt sample fetch keyword.
13384
13385table_http_err_rate(<table>)
13386 Uses the string representation of the input sample to perform a look up in
13387 the specified table. If the key is not found in the table, integer value zero
13388 is returned. Otherwise the average rate of HTTP errors associated with the
13389 input sample in the designated table, measured in amount of errors over the
13390 period configured in the table. See also the sc_http_err_rate sample fetch
13391 keyword.
13392
13393table_http_req_cnt(<table>)
13394 Uses the string representation of the input sample to perform a look up in
13395 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013396 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013397 requests associated with the input sample in the designated table. See also
13398 the sc_http_req_cnt sample fetch keyword.
13399
13400table_http_req_rate(<table>)
13401 Uses the string representation of the input sample to perform a look up in
13402 the specified table. If the key is not found in the table, integer value zero
13403 is returned. Otherwise the average rate of HTTP requests associated with the
13404 input sample in the designated table, measured in amount of requests over the
13405 period configured in the table. See also the sc_http_req_rate sample fetch
13406 keyword.
13407
13408table_kbytes_in(<table>)
13409 Uses the string representation of the input sample to perform a look up in
13410 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013411 is returned. Otherwise the converter returns the cumulative number of client-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013412 to-server data associated with the input sample in the designated table,
13413 measured in kilobytes. The test is currently performed on 32-bit integers,
13414 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
13415 keyword.
13416
13417table_kbytes_out(<table>)
13418 Uses the string representation of the input sample to perform a look up in
13419 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013420 is returned. Otherwise the converter returns the cumulative number of server-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013421 to-client data associated with the input sample in the designated table,
13422 measured in kilobytes. The test is currently performed on 32-bit integers,
13423 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
13424 keyword.
13425
13426table_server_id(<table>)
13427 Uses the string representation of the input sample to perform a look up in
13428 the specified table. If the key is not found in the table, integer value zero
13429 is returned. Otherwise the converter returns the server ID associated with
13430 the input sample in the designated table. A server ID is associated to a
13431 sample by a "stick" rule when a connection to a server succeeds. A server ID
13432 zero means that no server is associated with this key.
13433
13434table_sess_cnt(<table>)
13435 Uses the string representation of the input sample to perform a look up in
13436 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013437 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013438 sessions associated with the input sample in the designated table. Note that
13439 a session here refers to an incoming connection being accepted by the
13440 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
13441 keyword.
13442
13443table_sess_rate(<table>)
13444 Uses the string representation of the input sample to perform a look up in
13445 the specified table. If the key is not found in the table, integer value zero
13446 is returned. Otherwise the converter returns the average incoming session
13447 rate associated with the input sample in the designated table. Note that a
13448 session here refers to an incoming connection being accepted by the
13449 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
13450 keyword.
13451
13452table_trackers(<table>)
13453 Uses the string representation of the input sample to perform a look up in
13454 the specified table. If the key is not found in the table, integer value zero
13455 is returned. Otherwise the converter returns the current amount of concurrent
13456 connections tracking the same key as the input sample in the designated
13457 table. It differs from table_conn_cur in that it does not rely on any stored
13458 information but on the table's reference count (the "use" value which is
13459 returned by "show table" on the CLI). This may sometimes be more suited for
13460 layer7 tracking. It can be used to tell a server how many concurrent
13461 connections there are from a given address for example. See also the
13462 sc_trackers sample fetch keyword.
13463
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013464upper
13465 Convert a string sample to upper case. This can only be placed after a string
13466 sample fetch function or after a transformation keyword returning a string
13467 type. The result is of type string.
13468
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020013469url_dec
13470 Takes an url-encoded string provided as input and returns the decoded
13471 version as output. The input and the output are of type string.
13472
Christopher Faulet85d79c92016-11-09 16:54:56 +010013473unset-var(<var name>)
13474 Unsets a variable if the input content is defined. The name of the variable
13475 starts with an indication about its scope. The scopes allowed are:
13476 "proc" : the variable is shared with the whole process
13477 "sess" : the variable is shared with the whole session
13478 "txn" : the variable is shared with the transaction (request and
13479 response),
13480 "req" : the variable is shared only during request processing,
13481 "res" : the variable is shared only during response processing.
13482 This prefix is followed by a name. The separator is a '.'. The name may only
13483 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
13484
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013485utime(<format>[,<offset>])
13486 Converts an integer supposed to contain a date since epoch to a string
13487 representing this date in UTC time using a format defined by the <format>
13488 string using strftime(3). The purpose is to allow any date format to be used
13489 in logs. An optional <offset> in seconds may be applied to the input date
13490 (positive or negative). See the strftime() man page for the format supported
13491 by your operating system. See also the ltime converter.
13492
13493 Example :
13494
13495 # Emit two colons, one with the UTC time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010013496 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013497 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
13498
Marcin Deranek9631a282018-04-16 14:30:46 +020013499word(<index>,<delimiters>[,<count>])
13500 Extracts the nth word counting from the beginning (positive index) or from
13501 the end (negative index) considering given delimiters from an input string.
13502 Indexes start at 1 or -1 and delimiters are a string formatted list of chars.
13503 Optionally you can specify <count> of words to extract (default: 1).
13504 Value of 0 indicates extraction of all remaining words.
13505
13506 Example :
13507 str(f1_f2_f3__f5),word(4,_) # f5
13508 str(f1_f2_f3__f5),word(2,_,0) # f2_f3__f5
13509 str(f1_f2_f3__f5),word(3,_,2) # f3__f5
13510 str(f1_f2_f3__f5),word(-2,_,3) # f1_f2_f3
13511 str(f1_f2_f3__f5),word(-3,_,0) # f1_f2
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010013512
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013513wt6([<avalanche>])
13514 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
13515 hash function. Optionally, it is possible to apply a full avalanche hash
13516 function to the output if the optional <avalanche> argument equals 1. This
13517 converter uses the same functions as used by the various hash-based load
13518 balancing algorithms, so it will provide exactly the same results. It is
13519 mostly intended for debugging, but can be used as a stick-table entry to
13520 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013521 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", "crc32c",
13522 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013523
Willy Tarreau97707872015-01-27 15:12:13 +010013524xor(<value>)
13525 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013526 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013527 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013528 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013529 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013530 "sess" : the variable is shared with the whole session
13531 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013532 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010013533 "req" : the variable is shared only during request processing,
13534 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013535 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013536 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013537
Thierry FOURNIER01e09742016-12-26 11:46:11 +010013538xxh32([<seed>])
13539 Hashes a binary input sample into an unsigned 32-bit quantity using the 32-bit
13540 variant of the XXHash hash function. This hash supports a seed which defaults
13541 to zero but a different value maybe passed as the <seed> argument. This hash
13542 is known to be very good and very fast so it can be used to hash URLs and/or
13543 URL parameters for use as stick-table keys to collect statistics with a low
13544 collision rate, though care must be taken as the algorithm is not considered
13545 as cryptographically secure.
13546
13547xxh64([<seed>])
13548 Hashes a binary input sample into a signed 64-bit quantity using the 64-bit
13549 variant of the XXHash hash function. This hash supports a seed which defaults
13550 to zero but a different value maybe passed as the <seed> argument. This hash
13551 is known to be very good and very fast so it can be used to hash URLs and/or
13552 URL parameters for use as stick-table keys to collect statistics with a low
13553 collision rate, though care must be taken as the algorithm is not considered
13554 as cryptographically secure.
13555
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013556
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200135577.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020013558--------------------------------------------
13559
13560A first set of sample fetch methods applies to internal information which does
13561not even relate to any client information. These ones are sometimes used with
13562"monitor-fail" directives to report an internal status to external watchers.
13563The sample fetch methods described in this section are usable anywhere.
13564
13565always_false : boolean
13566 Always returns the boolean "false" value. It may be used with ACLs as a
13567 temporary replacement for another one when adjusting configurations.
13568
13569always_true : boolean
13570 Always returns the boolean "true" value. It may be used with ACLs as a
13571 temporary replacement for another one when adjusting configurations.
13572
13573avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010013574 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020013575 divided by the number of active servers. The current backend is used if no
13576 backend is specified. This is very similar to "queue" except that the size of
13577 the farm is considered, in order to give a more accurate measurement of the
13578 time it may take for a new connection to be processed. The main usage is with
13579 ACL to return a sorry page to new users when it becomes certain they will get
13580 a degraded service, or to pass to the backend servers in a header so that
13581 they decide to work in degraded mode or to disable some functions to speed up
13582 the processing a bit. Note that in the event there would not be any active
13583 server anymore, twice the number of queued connections would be considered as
13584 the measured value. This is a fair estimate, as we expect one server to get
13585 back soon anyway, but we still prefer to send new traffic to another backend
13586 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
13587 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010013588
Willy Tarreau74ca5042013-06-11 23:12:07 +020013589be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020013590 Applies to the number of currently established connections on the backend,
13591 possibly including the connection being evaluated. If no backend name is
13592 specified, the current one is used. But it is also possible to check another
13593 backend. It can be used to use a specific farm when the nominal one is full.
13594 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013595
Willy Tarreau74ca5042013-06-11 23:12:07 +020013596be_sess_rate([<backend>]) : integer
13597 Returns an integer value corresponding to the sessions creation rate on the
13598 backend, in number of new sessions per second. This is used with ACLs to
13599 switch to an alternate backend when an expensive or fragile one reaches too
Davor Ocelice9ed2812017-12-25 17:49:28 +010013600 high a session rate, or to limit abuse of service (e.g. prevent sucking of an
Willy Tarreau74ca5042013-06-11 23:12:07 +020013601 online dictionary). It can also be useful to add this element to logs using a
13602 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013603
13604 Example :
13605 # Redirect to an error page if the dictionary is requested too often
13606 backend dynamic
13607 mode http
13608 acl being_scanned be_sess_rate gt 100
13609 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010013610
Davor Ocelice9ed2812017-12-25 17:49:28 +010013611bin(<hex>) : bin
Thierry FOURNIERcc103292015-06-06 19:30:17 +020013612 Returns a binary chain. The input is the hexadecimal representation
13613 of the string.
13614
13615bool(<bool>) : bool
13616 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
13617 'false' and '0' are the same. 'true' and '1' are the same.
13618
Willy Tarreau74ca5042013-06-11 23:12:07 +020013619connslots([<backend>]) : integer
13620 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013621 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020013622 connections on all servers and the maximum queue size. This is probably only
13623 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050013624
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080013625 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020013626 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080013627 usage; see "use_backend" keyword) can be redirected to a different backend.
13628
Willy Tarreau55165fe2009-05-10 12:02:55 +020013629 'connslots' = number of available server connection slots, + number of
13630 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080013631
Willy Tarreaua36af912009-10-10 12:02:45 +020013632 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020013633 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020013634 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020013635 you want to be able to differentiate between different backends, and their
Davor Ocelice9ed2812017-12-25 17:49:28 +010013636 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020013637 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020013638 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080013639
Willy Tarreau55165fe2009-05-10 12:02:55 +020013640 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
13641 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020013642 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020013643 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080013644
Willy Tarreau6236d3a2013-07-25 14:28:25 +020013645date([<offset>]) : integer
13646 Returns the current date as the epoch (number of seconds since 01/01/1970).
13647 If an offset value is specified, then it is a number of seconds that is added
13648 to the current date before returning the value. This is particularly useful
13649 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020013650 It is useful combined with the http_date converter.
13651
13652 Example :
13653
13654 # set an expires header to now+1 hour in every response
13655 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020013656
Etienne Carrierea792a0a2018-01-17 13:43:24 +010013657date_us : integer
13658 Return the microseconds part of the date (the "second" part is returned by
13659 date sample). This sample is coherent with the date sample as it is comes
13660 from the same timeval structure.
13661
Willy Tarreaud716f9b2017-10-13 11:03:15 +020013662distcc_body(<token>[,<occ>]) : binary
13663 Parses a distcc message and returns the body associated to occurrence #<occ>
13664 of the token <token>. Occurrences start at 1, and when unspecified, any may
13665 match though in practice only the first one is checked for now. This can be
13666 used to extract file names or arguments in files built using distcc through
13667 haproxy. Please refer to distcc's protocol documentation for the complete
13668 list of supported tokens.
13669
13670distcc_param(<token>[,<occ>]) : integer
13671 Parses a distcc message and returns the parameter associated to occurrence
13672 #<occ> of the token <token>. Occurrences start at 1, and when unspecified,
13673 any may match though in practice only the first one is checked for now. This
13674 can be used to extract certain information such as the protocol version, the
13675 file size or the argument in files built using distcc through haproxy.
13676 Another use case consists in waiting for the start of the preprocessed file
13677 contents before connecting to the server to avoid keeping idle connections.
13678 Please refer to distcc's protocol documentation for the complete list of
13679 supported tokens.
13680
13681 Example :
13682 # wait up to 20s for the pre-processed file to be uploaded
13683 tcp-request inspect-delay 20s
13684 tcp-request content accept if { distcc_param(DOTI) -m found }
13685 # send large files to the big farm
13686 use_backend big_farm if { distcc_param(DOTI) gt 1000000 }
13687
Willy Tarreau595ec542013-06-12 21:34:28 +020013688env(<name>) : string
13689 Returns a string containing the value of environment variable <name>. As a
13690 reminder, environment variables are per-process and are sampled when the
13691 process starts. This can be useful to pass some information to a next hop
13692 server, or with ACLs to take specific action when the process is started a
13693 certain way.
13694
13695 Examples :
13696 # Pass the Via header to next hop with the local hostname in it
13697 http-request add-header Via 1.1\ %[env(HOSTNAME)]
13698
13699 # reject cookie-less requests when the STOP environment variable is set
13700 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
13701
Willy Tarreau74ca5042013-06-11 23:12:07 +020013702fe_conn([<frontend>]) : integer
13703 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010013704 possibly including the connection being evaluated. If no frontend name is
13705 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020013706 frontend. It can be used to return a sorry page before hard-blocking, or to
13707 use a specific backend to drain new requests when the farm is considered
Davor Ocelice9ed2812017-12-25 17:49:28 +010013708 full. This is mostly used with ACLs but can also be used to pass some
Willy Tarreau74ca5042013-06-11 23:12:07 +020013709 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
13710 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020013711
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020013712fe_req_rate([<frontend>]) : integer
13713 Returns an integer value corresponding to the number of HTTP requests per
13714 second sent to a frontend. This number can differ from "fe_sess_rate" in
13715 situations where client-side keep-alive is enabled.
13716
Willy Tarreau74ca5042013-06-11 23:12:07 +020013717fe_sess_rate([<frontend>]) : integer
13718 Returns an integer value corresponding to the sessions creation rate on the
13719 frontend, in number of new sessions per second. This is used with ACLs to
13720 limit the incoming session rate to an acceptable range in order to prevent
13721 abuse of service at the earliest moment, for example when combined with other
13722 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
13723 down below the limit. It can also be useful to add this element to logs using
13724 a log-format directive. See also the "rate-limit sessions" directive for use
13725 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010013726
13727 Example :
13728 # This frontend limits incoming mails to 10/s with a max of 100
13729 # concurrent connections. We accept any connection below 10/s, and
13730 # force excess clients to wait for 100 ms. Since clients are limited to
13731 # 100 max, there cannot be more than 10 incoming mails per second.
13732 frontend mail
13733 bind :25
13734 mode tcp
13735 maxconn 100
13736 acl too_fast fe_sess_rate ge 10
13737 tcp-request inspect-delay 100ms
13738 tcp-request content accept if ! too_fast
13739 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010013740
Nenad Merdanovic807a6e72017-03-12 22:00:00 +010013741hostname : string
13742 Returns the system hostname.
13743
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013744int(<integer>) : signed integer
13745 Returns a signed integer.
13746
Thierry FOURNIERcc103292015-06-06 19:30:17 +020013747ipv4(<ipv4>) : ipv4
13748 Returns an ipv4.
13749
13750ipv6(<ipv6>) : ipv6
13751 Returns an ipv6.
13752
13753meth(<method>) : method
13754 Returns a method.
13755
Willy Tarreau0f30d262014-11-24 16:02:05 +010013756nbproc : integer
13757 Returns an integer value corresponding to the number of processes that were
13758 started (it equals the global "nbproc" setting). This is useful for logging
13759 and debugging purposes.
13760
Willy Tarreau74ca5042013-06-11 23:12:07 +020013761nbsrv([<backend>]) : integer
13762 Returns an integer value corresponding to the number of usable servers of
13763 either the current backend or the named backend. This is mostly used with
13764 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010013765 switch to an alternate backend when the number of servers is too low to
13766 to handle some load. It is useful to report a failure when combined with
13767 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010013768
Willy Tarreau0f30d262014-11-24 16:02:05 +010013769proc : integer
13770 Returns an integer value corresponding to the position of the process calling
13771 the function, between 1 and global.nbproc. This is useful for logging and
13772 debugging purposes.
13773
Willy Tarreau74ca5042013-06-11 23:12:07 +020013774queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010013775 Returns the total number of queued connections of the designated backend,
13776 including all the connections in server queues. If no backend name is
13777 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020013778 one. This is useful with ACLs or to pass statistics to backend servers. This
13779 can be used to take actions when queuing goes above a known level, generally
13780 indicating a surge of traffic or a massive slowdown on the servers. One
13781 possible action could be to reject new users but still accept old ones. See
13782 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
13783
Willy Tarreau84310e22014-02-14 11:59:04 +010013784rand([<range>]) : integer
13785 Returns a random integer value within a range of <range> possible values,
13786 starting at zero. If the range is not specified, it defaults to 2^32, which
13787 gives numbers between 0 and 4294967295. It can be useful to pass some values
13788 needed to take some routing decisions for example, or just for debugging
13789 purposes. This random must not be used for security purposes.
13790
Willy Tarreau74ca5042013-06-11 23:12:07 +020013791srv_conn([<backend>/]<server>) : integer
13792 Returns an integer value corresponding to the number of currently established
13793 connections on the designated server, possibly including the connection being
13794 evaluated. If <backend> is omitted, then the server is looked up in the
13795 current backend. It can be used to use a specific farm when one server is
13796 full, or to inform the server about our view of the number of active
13797 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
13798 methods.
13799
13800srv_is_up([<backend>/]<server>) : boolean
13801 Returns true when the designated server is UP, and false when it is either
13802 DOWN or in maintenance mode. If <backend> is omitted, then the server is
13803 looked up in the current backend. It is mainly used to take action based on
Davor Ocelice9ed2812017-12-25 17:49:28 +010013804 an external status reported via a health check (e.g. a geographical site's
Willy Tarreau74ca5042013-06-11 23:12:07 +020013805 availability). Another possible use which is more of a hack consists in
13806 using dummy servers as boolean variables that can be enabled or disabled from
13807 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
13808
Willy Tarreauff2b7af2017-10-13 11:46:26 +020013809srv_queue([<backend>/]<server>) : integer
13810 Returns an integer value corresponding to the number of connections currently
13811 pending in the designated server's queue. If <backend> is omitted, then the
13812 server is looked up in the current backend. It can sometimes be used together
13813 with the "use-server" directive to force to use a known faster server when it
13814 is not much loaded. See also the "srv_conn", "avg_queue" and "queue" sample
13815 fetch methods.
13816
Willy Tarreau74ca5042013-06-11 23:12:07 +020013817srv_sess_rate([<backend>/]<server>) : integer
13818 Returns an integer corresponding to the sessions creation rate on the
13819 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013820 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020013821 used with ACLs but can make sense with logs too. This is used to switch to an
13822 alternate backend when an expensive or fragile one reaches too high a session
Davor Ocelice9ed2812017-12-25 17:49:28 +010013823 rate, or to limit abuse of service (e.g. prevent latent requests from
Willy Tarreau74ca5042013-06-11 23:12:07 +020013824 overloading servers).
13825
13826 Example :
13827 # Redirect to a separate back
13828 acl srv1_full srv_sess_rate(be1/srv1) gt 50
13829 acl srv2_full srv_sess_rate(be1/srv2) gt 50
13830 use_backend be2 if srv1_full or srv2_full
13831
Willy Tarreau0f30d262014-11-24 16:02:05 +010013832stopping : boolean
13833 Returns TRUE if the process calling the function is currently stopping. This
13834 can be useful for logging, or for relaxing certain checks or helping close
13835 certain connections upon graceful shutdown.
13836
Thierry FOURNIERcc103292015-06-06 19:30:17 +020013837str(<string>) : string
13838 Returns a string.
13839
Willy Tarreau74ca5042013-06-11 23:12:07 +020013840table_avl([<table>]) : integer
13841 Returns the total number of available entries in the current proxy's
13842 stick-table or in the designated stick-table. See also table_cnt.
13843
13844table_cnt([<table>]) : integer
13845 Returns the total number of entries currently in use in the current proxy's
13846 stick-table or in the designated stick-table. See also src_conn_cnt and
13847 table_avl for other entry counting methods.
13848
Christopher Faulet34adb2a2017-11-21 21:45:38 +010013849thread : integer
13850 Returns an integer value corresponding to the position of the thread calling
13851 the function, between 0 and (global.nbthread-1). This is useful for logging
13852 and debugging purposes.
13853
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013854var(<var-name>) : undefined
13855 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010013856 sample fetch fails. The name of the variable starts with an indication
13857 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013858 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013859 "sess" : the variable is shared with the whole session
13860 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013861 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010013862 "req" : the variable is shared only during request processing,
13863 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013864 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013865 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013866
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200138677.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020013868----------------------------------
13869
13870The layer 4 usually describes just the transport layer which in haproxy is
13871closest to the connection, where no content is yet made available. The fetch
13872methods described here are usable as low as the "tcp-request connection" rule
13873sets unless they require some future information. Those generally include
13874TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013875the incoming connection. For retrieving a value from a sticky counters, the
13876counter number can be explicitly set as 0, 1, or 2 using the pre-defined
13877"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013878argument when using the "sc_" prefix. An optional table may be specified with
13879the "sc*" form, in which case the currently tracked key will be looked up into
13880this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013881
13882be_id : integer
13883 Returns an integer containing the current backend's id. It can be used in
13884 frontends with responses to check which backend processed the request.
13885
Marcin Deranekd2471c22016-12-12 14:08:05 +010013886be_name : string
13887 Returns a string containing the current backend's name. It can be used in
13888 frontends with responses to check which backend processed the request.
13889
Willy Tarreau74ca5042013-06-11 23:12:07 +020013890dst : ip
13891 This is the destination IPv4 address of the connection on the client side,
13892 which is the address the client connected to. It can be useful when running
13893 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
13894 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
13895 RFC 4291.
13896
13897dst_conn : integer
13898 Returns an integer value corresponding to the number of currently established
13899 connections on the same socket including the one being evaluated. It is
13900 normally used with ACLs but can as well be used to pass the information to
13901 servers in an HTTP header or in logs. It can be used to either return a sorry
13902 page before hard-blocking, or to use a specific backend to drain new requests
13903 when the socket is considered saturated. This offers the ability to assign
13904 different limits to different listening ports or addresses. See also the
13905 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013906
Willy Tarreau16e01562016-08-09 16:46:18 +020013907dst_is_local : boolean
13908 Returns true if the destination address of the incoming connection is local
13909 to the system, or false if the address doesn't exist on the system, meaning
13910 that it was intercepted in transparent mode. It can be useful to apply
13911 certain rules by default to forwarded traffic and other rules to the traffic
Davor Ocelice9ed2812017-12-25 17:49:28 +010013912 targeting the real address of the machine. For example the stats page could
Willy Tarreau16e01562016-08-09 16:46:18 +020013913 be delivered only on this address, or SSH access could be locally redirected.
13914 Please note that the check involves a few system calls, so it's better to do
13915 it only once per connection.
13916
Willy Tarreau74ca5042013-06-11 23:12:07 +020013917dst_port : integer
13918 Returns an integer value corresponding to the destination TCP port of the
13919 connection on the client side, which is the port the client connected to.
13920 This might be used when running in transparent mode, when assigning dynamic
13921 ports to some clients for a whole application session, to stick all users to
13922 a same server, or to pass the destination port information to a server using
13923 an HTTP header.
13924
Willy Tarreau60ca10a2017-08-18 15:26:54 +020013925fc_http_major : integer
13926 Reports the front connection's HTTP major version encoding, which may be 1
13927 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
13928 encoding and not on the version present in the request header.
13929
Emeric Brun4f603012017-01-05 15:11:44 +010013930fc_rcvd_proxy : boolean
13931 Returns true if the client initiated the connection with a PROXY protocol
13932 header.
13933
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020013934fc_rtt(<unit>) : integer
13935 Returns the Round Trip Time (RTT) measured by the kernel for the client
13936 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
13937 can be set to "ms" for milliseconds or "us" for microseconds. If the server
13938 connection is not established, if the connection is not TCP or if the
13939 operating system does not support TCP_INFO, for example Linux kernels before
13940 2.4, the sample fetch fails.
13941
13942fc_rttvar(<unit>) : integer
13943 Returns the Round Trip Time (RTT) variance measured by the kernel for the
13944 client connection. <unit> is facultative, by default the unit is milliseconds.
13945 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
13946 server connection is not established, if the connection is not TCP or if the
13947 operating system does not support TCP_INFO, for example Linux kernels before
13948 2.4, the sample fetch fails.
13949
Joe Williams30fcd392016-08-10 07:06:44 -070013950fc_unacked(<unit>) : integer
13951 Returns the unacked counter measured by the kernel for the client connection.
13952 If the server connection is not established, if the connection is not TCP or
13953 if the operating system does not support TCP_INFO, for example Linux kernels
13954 before 2.4, the sample fetch fails.
13955
13956fc_sacked(<unit>) : integer
13957 Returns the sacked counter measured by the kernel for the client connection.
13958 If the server connection is not established, if the connection is not TCP or
13959 if the operating system does not support TCP_INFO, for example Linux kernels
13960 before 2.4, the sample fetch fails.
13961
13962fc_retrans(<unit>) : integer
13963 Returns the retransmits counter measured by the kernel for the client
13964 connection. If the server connection is not established, if the connection is
13965 not TCP or if the operating system does not support TCP_INFO, for example
13966 Linux kernels before 2.4, the sample fetch fails.
13967
13968fc_fackets(<unit>) : integer
13969 Returns the fack counter measured by the kernel for the client
13970 connection. If the server connection is not established, if the connection is
13971 not TCP or if the operating system does not support TCP_INFO, for example
13972 Linux kernels before 2.4, the sample fetch fails.
13973
13974fc_lost(<unit>) : integer
13975 Returns the lost counter measured by the kernel for the client
13976 connection. If the server connection is not established, if the connection is
13977 not TCP or if the operating system does not support TCP_INFO, for example
13978 Linux kernels before 2.4, the sample fetch fails.
13979
13980fc_reordering(<unit>) : integer
13981 Returns the reordering counter measured by the kernel for the client
13982 connection. If the server connection is not established, if the connection is
13983 not TCP or if the operating system does not support TCP_INFO, for example
13984 Linux kernels before 2.4, the sample fetch fails.
13985
Marcin Deranek9a66dfb2018-04-13 14:37:50 +020013986fe_defbe : string
13987 Returns a string containing the frontend's default backend name. It can be
13988 used in frontends to check which backend will handle requests by default.
13989
Willy Tarreau74ca5042013-06-11 23:12:07 +020013990fe_id : integer
13991 Returns an integer containing the current frontend's id. It can be used in
Marcin Deranek6e413ed2016-12-13 12:40:01 +010013992 backends to check from which frontend it was called, or to stick all users
Willy Tarreau74ca5042013-06-11 23:12:07 +020013993 coming via a same frontend to the same server.
13994
Marcin Deranekd2471c22016-12-12 14:08:05 +010013995fe_name : string
13996 Returns a string containing the current frontend's name. It can be used in
13997 backends to check from which frontend it was called, or to stick all users
13998 coming via a same frontend to the same server.
13999
Cyril Bonté62ba8702014-04-22 23:52:25 +020014000sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014001sc0_bytes_in_rate([<table>]) : integer
14002sc1_bytes_in_rate([<table>]) : integer
14003sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014004 Returns the average client-to-server bytes rate from the currently tracked
14005 counters, measured in amount of bytes over the period configured in the
14006 table. See also src_bytes_in_rate.
14007
Cyril Bonté62ba8702014-04-22 23:52:25 +020014008sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014009sc0_bytes_out_rate([<table>]) : integer
14010sc1_bytes_out_rate([<table>]) : integer
14011sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014012 Returns the average server-to-client bytes rate from the currently tracked
14013 counters, measured in amount of bytes over the period configured in the
14014 table. See also src_bytes_out_rate.
14015
Cyril Bonté62ba8702014-04-22 23:52:25 +020014016sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014017sc0_clr_gpc0([<table>]) : integer
14018sc1_clr_gpc0([<table>]) : integer
14019sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020014020 Clears the first General Purpose Counter associated to the currently tracked
14021 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010014022 stored value is zero, so first invocation will always return zero. This is
14023 typically used as a second ACL in an expression in order to mark a connection
14024 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020014025
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014026 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020014027 # block if 5 consecutive requests continue to come faster than 10 sess
14028 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014029 acl abuse sc0_http_req_rate gt 10
14030 acl kill sc0_inc_gpc0 gt 5
14031 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020014032 tcp-request connection accept if !abuse save
14033 tcp-request connection reject if abuse kill
14034
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014035sc_clr_gpc1(<ctr>[,<table>]) : integer
14036sc0_clr_gpc1([<table>]) : integer
14037sc1_clr_gpc1([<table>]) : integer
14038sc2_clr_gpc1([<table>]) : integer
14039 Clears the second General Purpose Counter associated to the currently tracked
14040 counters, and returns its previous value. Before the first invocation, the
14041 stored value is zero, so first invocation will always return zero. This is
14042 typically used as a second ACL in an expression in order to mark a connection
14043 when a first ACL was verified.
14044
Cyril Bonté62ba8702014-04-22 23:52:25 +020014045sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014046sc0_conn_cnt([<table>]) : integer
14047sc1_conn_cnt([<table>]) : integer
14048sc2_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014049 Returns the cumulative number of incoming connections from currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020014050 counters. See also src_conn_cnt.
14051
Cyril Bonté62ba8702014-04-22 23:52:25 +020014052sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014053sc0_conn_cur([<table>]) : integer
14054sc1_conn_cur([<table>]) : integer
14055sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014056 Returns the current amount of concurrent connections tracking the same
14057 tracked counters. This number is automatically incremented when tracking
14058 begins and decremented when tracking stops. See also src_conn_cur.
14059
Cyril Bonté62ba8702014-04-22 23:52:25 +020014060sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014061sc0_conn_rate([<table>]) : integer
14062sc1_conn_rate([<table>]) : integer
14063sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014064 Returns the average connection rate from the currently tracked counters,
14065 measured in amount of connections over the period configured in the table.
14066 See also src_conn_rate.
14067
Cyril Bonté62ba8702014-04-22 23:52:25 +020014068sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014069sc0_get_gpc0([<table>]) : integer
14070sc1_get_gpc0([<table>]) : integer
14071sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014072 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014073 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014074
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014075sc_get_gpc1(<ctr>[,<table>]) : integer
14076sc0_get_gpc1([<table>]) : integer
14077sc1_get_gpc1([<table>]) : integer
14078sc2_get_gpc1([<table>]) : integer
14079 Returns the value of the second General Purpose Counter associated to the
14080 currently tracked counters. See also src_get_gpc1 and sc/sc0/sc1/sc2_inc_gpc1.
14081
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014082sc_get_gpt0(<ctr>[,<table>]) : integer
14083sc0_get_gpt0([<table>]) : integer
14084sc1_get_gpt0([<table>]) : integer
14085sc2_get_gpt0([<table>]) : integer
14086 Returns the value of the first General Purpose Tag associated to the
14087 currently tracked counters. See also src_get_gpt0.
14088
Cyril Bonté62ba8702014-04-22 23:52:25 +020014089sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014090sc0_gpc0_rate([<table>]) : integer
14091sc1_gpc0_rate([<table>]) : integer
14092sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014093 Returns the average increment rate of the first General Purpose Counter
14094 associated to the currently tracked counters. It reports the frequency
14095 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014096 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
14097 that the "gpc0_rate" counter must be stored in the stick-table for a value to
14098 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020014099
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014100sc_gpc1_rate(<ctr>[,<table>]) : integer
14101sc0_gpc1_rate([<table>]) : integer
14102sc1_gpc1_rate([<table>]) : integer
14103sc2_gpc1_rate([<table>]) : integer
14104 Returns the average increment rate of the second General Purpose Counter
14105 associated to the currently tracked counters. It reports the frequency
14106 which the gpc1 counter was incremented over the configured period. See also
14107 src_gpcA_rate, sc/sc0/sc1/sc2_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
14108 that the "gpc1_rate" counter must be stored in the stick-table for a value to
14109 be returned, as "gpc1" only holds the event count.
14110
Cyril Bonté62ba8702014-04-22 23:52:25 +020014111sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014112sc0_http_err_cnt([<table>]) : integer
14113sc1_http_err_cnt([<table>]) : integer
14114sc2_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014115 Returns the cumulative number of HTTP errors from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020014116 counters. This includes the both request errors and 4xx error responses.
14117 See also src_http_err_cnt.
14118
Cyril Bonté62ba8702014-04-22 23:52:25 +020014119sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014120sc0_http_err_rate([<table>]) : integer
14121sc1_http_err_rate([<table>]) : integer
14122sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014123 Returns the average rate of HTTP errors from the currently tracked counters,
14124 measured in amount of errors over the period configured in the table. This
14125 includes the both request errors and 4xx error responses. See also
14126 src_http_err_rate.
14127
Cyril Bonté62ba8702014-04-22 23:52:25 +020014128sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014129sc0_http_req_cnt([<table>]) : integer
14130sc1_http_req_cnt([<table>]) : integer
14131sc2_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014132 Returns the cumulative number of HTTP requests from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020014133 counters. This includes every started request, valid or not. See also
14134 src_http_req_cnt.
14135
Cyril Bonté62ba8702014-04-22 23:52:25 +020014136sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014137sc0_http_req_rate([<table>]) : integer
14138sc1_http_req_rate([<table>]) : integer
14139sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014140 Returns the average rate of HTTP requests from the currently tracked
14141 counters, measured in amount of requests over the period configured in
14142 the table. This includes every started request, valid or not. See also
14143 src_http_req_rate.
14144
Cyril Bonté62ba8702014-04-22 23:52:25 +020014145sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014146sc0_inc_gpc0([<table>]) : integer
14147sc1_inc_gpc0([<table>]) : integer
14148sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014149 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010014150 tracked counters, and returns its new value. Before the first invocation,
14151 the stored value is zero, so first invocation will increase it to 1 and will
14152 return 1. This is typically used as a second ACL in an expression in order
14153 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020014154
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014155 Example:
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014156 acl abuse sc0_http_req_rate gt 10
14157 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020014158 tcp-request connection reject if abuse kill
14159
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014160sc_inc_gpc1(<ctr>[,<table>]) : integer
14161sc0_inc_gpc1([<table>]) : integer
14162sc1_inc_gpc1([<table>]) : integer
14163sc2_inc_gpc1([<table>]) : integer
14164 Increments the second General Purpose Counter associated to the currently
14165 tracked counters, and returns its new value. Before the first invocation,
14166 the stored value is zero, so first invocation will increase it to 1 and will
14167 return 1. This is typically used as a second ACL in an expression in order
14168 to mark a connection when a first ACL was verified.
14169
Cyril Bonté62ba8702014-04-22 23:52:25 +020014170sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014171sc0_kbytes_in([<table>]) : integer
14172sc1_kbytes_in([<table>]) : integer
14173sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020014174 Returns the total amount of client-to-server data from the currently tracked
14175 counters, measured in kilobytes. The test is currently performed on 32-bit
14176 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020014177
Cyril Bonté62ba8702014-04-22 23:52:25 +020014178sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014179sc0_kbytes_out([<table>]) : integer
14180sc1_kbytes_out([<table>]) : integer
14181sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020014182 Returns the total amount of server-to-client data from the currently tracked
14183 counters, measured in kilobytes. The test is currently performed on 32-bit
14184 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020014185
Cyril Bonté62ba8702014-04-22 23:52:25 +020014186sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014187sc0_sess_cnt([<table>]) : integer
14188sc1_sess_cnt([<table>]) : integer
14189sc2_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014190 Returns the cumulative number of incoming connections that were transformed
Willy Tarreaue9656522010-08-17 15:40:09 +020014191 into sessions, which means that they were accepted by a "tcp-request
14192 connection" rule, from the currently tracked counters. A backend may count
14193 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014194 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020014195 with the client. See also src_sess_cnt.
14196
Cyril Bonté62ba8702014-04-22 23:52:25 +020014197sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014198sc0_sess_rate([<table>]) : integer
14199sc1_sess_rate([<table>]) : integer
14200sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014201 Returns the average session rate from the currently tracked counters,
14202 measured in amount of sessions over the period configured in the table. A
14203 session is a connection that got past the early "tcp-request connection"
14204 rules. A backend may count more sessions than connections because each
14205 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014206 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020014207
Cyril Bonté62ba8702014-04-22 23:52:25 +020014208sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020014209sc0_tracked([<table>]) : boolean
14210sc1_tracked([<table>]) : boolean
14211sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020014212 Returns true if the designated session counter is currently being tracked by
14213 the current session. This can be useful when deciding whether or not we want
14214 to set some values in a header passed to the server.
14215
Cyril Bonté62ba8702014-04-22 23:52:25 +020014216sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014217sc0_trackers([<table>]) : integer
14218sc1_trackers([<table>]) : integer
14219sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010014220 Returns the current amount of concurrent connections tracking the same
14221 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014222 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010014223 that it does not rely on any stored information but on the table's reference
14224 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020014225 may sometimes be more suited for layer7 tracking. It can be used to tell a
14226 server how many concurrent connections there are from a given address for
14227 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010014228
Willy Tarreau74ca5042013-06-11 23:12:07 +020014229so_id : integer
14230 Returns an integer containing the current listening socket's id. It is useful
14231 in frontends involving many "bind" lines, or to stick all users coming via a
14232 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014233
Willy Tarreau74ca5042013-06-11 23:12:07 +020014234src : ip
Davor Ocelice9ed2812017-12-25 17:49:28 +010014235 This is the source IPv4 address of the client of the session. It is of type
Willy Tarreau74ca5042013-06-11 23:12:07 +020014236 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
14237 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
14238 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010014239 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
14240 directive is used, it can be the address of a client behind another
14241 PROXY-protocol compatible component for all rule sets except
14242 "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014243
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014244 Example:
14245 # add an HTTP header in requests with the originating address' country
14246 http-request set-header X-Country %[src,map_ip(geoip.lst)]
14247
Willy Tarreau74ca5042013-06-11 23:12:07 +020014248src_bytes_in_rate([<table>]) : integer
14249 Returns the average bytes rate from the incoming connection's source address
14250 in the current proxy's stick-table or in the designated stick-table, measured
14251 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014252 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014253
Willy Tarreau74ca5042013-06-11 23:12:07 +020014254src_bytes_out_rate([<table>]) : integer
14255 Returns the average bytes rate to the incoming connection's source address in
14256 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020014257 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014258 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014259
Willy Tarreau74ca5042013-06-11 23:12:07 +020014260src_clr_gpc0([<table>]) : integer
14261 Clears the first General Purpose Counter associated to the incoming
14262 connection's source address in the current proxy's stick-table or in the
14263 designated stick-table, and returns its previous value. If the address is not
14264 found, an entry is created and 0 is returned. This is typically used as a
14265 second ACL in an expression in order to mark a connection when a first ACL
14266 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020014267
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014268 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020014269 # block if 5 consecutive requests continue to come faster than 10 sess
14270 # per second, and reset the counter as soon as the traffic slows down.
14271 acl abuse src_http_req_rate gt 10
14272 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010014273 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020014274 tcp-request connection accept if !abuse save
14275 tcp-request connection reject if abuse kill
14276
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014277src_clr_gpc1([<table>]) : integer
14278 Clears the second General Purpose Counter associated to the incoming
14279 connection's source address in the current proxy's stick-table or in the
14280 designated stick-table, and returns its previous value. If the address is not
14281 found, an entry is created and 0 is returned. This is typically used as a
14282 second ACL in an expression in order to mark a connection when a first ACL
14283 was verified.
14284
Willy Tarreau74ca5042013-06-11 23:12:07 +020014285src_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014286 Returns the cumulative number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020014287 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020014288 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014289 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014290
Willy Tarreau74ca5042013-06-11 23:12:07 +020014291src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020014292 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020014293 current incoming connection's source address in the current proxy's
14294 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014295 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014296
Willy Tarreau74ca5042013-06-11 23:12:07 +020014297src_conn_rate([<table>]) : integer
14298 Returns the average connection rate from the incoming connection's source
14299 address in the current proxy's stick-table or in the designated stick-table,
14300 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014301 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014302
Willy Tarreau74ca5042013-06-11 23:12:07 +020014303src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020014304 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020014305 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020014306 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014307 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014308
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014309src_get_gpc1([<table>]) : integer
14310 Returns the value of the second General Purpose Counter associated to the
14311 incoming connection's source address in the current proxy's stick-table or in
14312 the designated stick-table. If the address is not found, zero is returned.
14313 See also sc/sc0/sc1/sc2_get_gpc1 and src_inc_gpc1.
14314
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014315src_get_gpt0([<table>]) : integer
14316 Returns the value of the first General Purpose Tag associated to the
14317 incoming connection's source address in the current proxy's stick-table or in
14318 the designated stick-table. If the address is not found, zero is returned.
14319 See also sc/sc0/sc1/sc2_get_gpt0.
14320
Willy Tarreau74ca5042013-06-11 23:12:07 +020014321src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014322 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020014323 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014324 stick-table or in the designated stick-table. It reports the frequency
14325 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014326 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
14327 that the "gpc0_rate" counter must be stored in the stick-table for a value to
14328 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014329
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014330src_gpc1_rate([<table>]) : integer
14331 Returns the average increment rate of the second General Purpose Counter
14332 associated to the incoming connection's source address in the current proxy's
14333 stick-table or in the designated stick-table. It reports the frequency
14334 which the gpc1 counter was incremented over the configured period. See also
14335 sc/sc0/sc1/sc2_gpc1_rate, src_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
14336 that the "gpc1_rate" counter must be stored in the stick-table for a value to
14337 be returned, as "gpc1" only holds the event count.
14338
Willy Tarreau74ca5042013-06-11 23:12:07 +020014339src_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014340 Returns the cumulative number of HTTP errors from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020014341 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020014342 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014343 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020014344 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014345
Willy Tarreau74ca5042013-06-11 23:12:07 +020014346src_http_err_rate([<table>]) : integer
14347 Returns the average rate of HTTP errors from the incoming connection's source
14348 address in the current proxy's stick-table or in the designated stick-table,
14349 measured in amount of errors over the period configured in the table. This
14350 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014351 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014352
Willy Tarreau74ca5042013-06-11 23:12:07 +020014353src_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014354 Returns the cumulative number of HTTP requests from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020014355 source address in the current proxy's stick-table or in the designated stick-
14356 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014357 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014358
Willy Tarreau74ca5042013-06-11 23:12:07 +020014359src_http_req_rate([<table>]) : integer
14360 Returns the average rate of HTTP requests from the incoming connection's
14361 source address in the current proxy's stick-table or in the designated stick-
14362 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020014363 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014364 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014365
Willy Tarreau74ca5042013-06-11 23:12:07 +020014366src_inc_gpc0([<table>]) : integer
14367 Increments the first General Purpose Counter associated to the incoming
14368 connection's source address in the current proxy's stick-table or in the
14369 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014370 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014371 This is typically used as a second ACL in an expression in order to mark a
14372 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020014373
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014374 Example:
Willy Tarreauc9705a12010-07-27 20:05:50 +020014375 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010014376 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020014377 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020014378
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014379src_inc_gpc1([<table>]) : integer
14380 Increments the second General Purpose Counter associated to the incoming
14381 connection's source address in the current proxy's stick-table or in the
14382 designated stick-table, and returns its new value. If the address is not
14383 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc1.
14384 This is typically used as a second ACL in an expression in order to mark a
14385 connection when a first ACL was verified.
14386
Willy Tarreau16e01562016-08-09 16:46:18 +020014387src_is_local : boolean
14388 Returns true if the source address of the incoming connection is local to the
14389 system, or false if the address doesn't exist on the system, meaning that it
14390 comes from a remote machine. Note that UNIX addresses are considered local.
14391 It can be useful to apply certain access restrictions based on where the
Davor Ocelice9ed2812017-12-25 17:49:28 +010014392 client comes from (e.g. require auth or https for remote machines). Please
Willy Tarreau16e01562016-08-09 16:46:18 +020014393 note that the check involves a few system calls, so it's better to do it only
14394 once per connection.
14395
Willy Tarreau74ca5042013-06-11 23:12:07 +020014396src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020014397 Returns the total amount of data received from the incoming connection's
14398 source address in the current proxy's stick-table or in the designated
14399 stick-table, measured in kilobytes. If the address is not found, zero is
14400 returned. The test is currently performed on 32-bit integers, which limits
14401 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014402
Willy Tarreau74ca5042013-06-11 23:12:07 +020014403src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020014404 Returns the total amount of data sent to the incoming connection's source
14405 address in the current proxy's stick-table or in the designated stick-table,
14406 measured in kilobytes. If the address is not found, zero is returned. The
14407 test is currently performed on 32-bit integers, which limits values to 4
14408 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020014409
Willy Tarreau74ca5042013-06-11 23:12:07 +020014410src_port : integer
14411 Returns an integer value corresponding to the TCP source port of the
14412 connection on the client side, which is the port the client connected from.
14413 Usage of this function is very limited as modern protocols do not care much
14414 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014415
Willy Tarreau74ca5042013-06-11 23:12:07 +020014416src_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014417 Returns the cumulative number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020014418 connection's source IPv4 address in the current proxy's stick-table or in the
14419 designated stick-table, that were transformed into sessions, which means that
14420 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014421 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014422
Willy Tarreau74ca5042013-06-11 23:12:07 +020014423src_sess_rate([<table>]) : integer
14424 Returns the average session rate from the incoming connection's source
14425 address in the current proxy's stick-table or in the designated stick-table,
14426 measured in amount of sessions over the period configured in the table. A
14427 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014428 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014429
Willy Tarreau74ca5042013-06-11 23:12:07 +020014430src_updt_conn_cnt([<table>]) : integer
14431 Creates or updates the entry associated to the incoming connection's source
14432 address in the current proxy's stick-table or in the designated stick-table.
14433 This table must be configured to store the "conn_cnt" data type, otherwise
14434 the match will be ignored. The current count is incremented by one, and the
14435 expiration timer refreshed. The updated count is returned, so this match
14436 can't return zero. This was used to reject service abusers based on their
14437 source address. Note: it is recommended to use the more complete "track-sc*"
14438 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020014439
14440 Example :
14441 # This frontend limits incoming SSH connections to 3 per 10 second for
14442 # each source address, and rejects excess connections until a 10 second
14443 # silence is observed. At most 20 addresses are tracked.
14444 listen ssh
14445 bind :22
14446 mode tcp
14447 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020014448 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020014449 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020014450 server local 127.0.0.1:22
14451
Willy Tarreau74ca5042013-06-11 23:12:07 +020014452srv_id : integer
14453 Returns an integer containing the server's id when processing the response.
14454 While it's almost only used with ACLs, it may be used for logging or
14455 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020014456
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200144577.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020014458----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020014459
Willy Tarreau74ca5042013-06-11 23:12:07 +020014460The layer 5 usually describes just the session layer which in haproxy is
14461closest to the session once all the connection handshakes are finished, but
14462when no content is yet made available. The fetch methods described here are
14463usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014464future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020014465
Ben Shillitof25e8e52016-12-02 14:25:37 +00001446651d.all(<prop>[,<prop>*]) : string
14467 Returns values for the properties requested as a string, where values are
14468 separated by the delimiter specified with "51degrees-property-separator".
14469 The device is identified using all the important HTTP headers from the
14470 request. The function can be passed up to five property names, and if a
14471 property name can't be found, the value "NoData" is returned.
14472
14473 Example :
14474 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request
14475 # containing the three properties requested using all relevant headers from
14476 # the request.
14477 frontend http-in
14478 bind *:8081
14479 default_backend servers
14480 http-request set-header X-51D-DeviceTypeMobileTablet \
14481 %[51d.all(DeviceType,IsMobile,IsTablet)]
14482
Emeric Brun645ae792014-04-30 14:21:06 +020014483ssl_bc : boolean
14484 Returns true when the back connection was made via an SSL/TLS transport
14485 layer and is locally deciphered. This means the outgoing connection was made
14486 other a server with the "ssl" option.
14487
14488ssl_bc_alg_keysize : integer
14489 Returns the symmetric cipher key size supported in bits when the outgoing
14490 connection was made over an SSL/TLS transport layer.
14491
14492ssl_bc_cipher : string
14493 Returns the name of the used cipher when the outgoing connection was made
14494 over an SSL/TLS transport layer.
14495
Emeric Brun74f7ffa2018-02-19 16:14:12 +010014496ssl_bc_is_resumed : boolean
14497 Returns true when the back connection was made over an SSL/TLS transport
14498 layer and the newly created SSL session was resumed using a cached
14499 session or a TLS ticket.
14500
Emeric Brun645ae792014-04-30 14:21:06 +020014501ssl_bc_protocol : string
14502 Returns the name of the used protocol when the outgoing connection was made
14503 over an SSL/TLS transport layer.
14504
Emeric Brunb73a9b02014-04-30 18:49:19 +020014505ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020014506 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020014507 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
14508 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020014509
14510ssl_bc_session_id : binary
14511 Returns the SSL ID of the back connection when the outgoing connection was
14512 made over an SSL/TLS transport layer. It is useful to log if we want to know
14513 if session was reused or not.
14514
Patrick Hemmere0275472018-04-28 19:15:51 -040014515ssl_bc_session_key : binary
14516 Returns the SSL session master key of the back connection when the outgoing
14517 connection was made over an SSL/TLS transport layer. It is useful to decrypt
14518 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
14519 BoringSSL.
14520
Emeric Brun645ae792014-04-30 14:21:06 +020014521ssl_bc_use_keysize : integer
14522 Returns the symmetric cipher key size used in bits when the outgoing
14523 connection was made over an SSL/TLS transport layer.
14524
Willy Tarreau74ca5042013-06-11 23:12:07 +020014525ssl_c_ca_err : integer
14526 When the incoming connection was made over an SSL/TLS transport layer,
14527 returns the ID of the first error detected during verification of the client
14528 certificate at depth > 0, or 0 if no error was encountered during this
14529 verification process. Please refer to your SSL library's documentation to
14530 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020014531
Willy Tarreau74ca5042013-06-11 23:12:07 +020014532ssl_c_ca_err_depth : integer
14533 When the incoming connection was made over an SSL/TLS transport layer,
14534 returns the depth in the CA chain of the first error detected during the
14535 verification of the client certificate. If no error is encountered, 0 is
14536 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010014537
Emeric Brun43e79582014-10-29 19:03:26 +010014538ssl_c_der : binary
14539 Returns the DER formatted certificate presented by the client when the
14540 incoming connection was made over an SSL/TLS transport layer. When used for
14541 an ACL, the value(s) to match against can be passed in hexadecimal form.
14542
Willy Tarreau74ca5042013-06-11 23:12:07 +020014543ssl_c_err : integer
14544 When the incoming connection was made over an SSL/TLS transport layer,
14545 returns the ID of the first error detected during verification at depth 0, or
14546 0 if no error was encountered during this verification process. Please refer
14547 to your SSL library's documentation to find the exhaustive list of error
14548 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020014549
Willy Tarreau74ca5042013-06-11 23:12:07 +020014550ssl_c_i_dn([<entry>[,<occ>]]) : string
14551 When the incoming connection was made over an SSL/TLS transport layer,
14552 returns the full distinguished name of the issuer of the certificate
14553 presented by the client when no <entry> is specified, or the value of the
14554 first given entry found from the beginning of the DN. If a positive/negative
14555 occurrence number is specified as the optional second argument, it returns
14556 the value of the nth given entry value from the beginning/end of the DN.
14557 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
14558 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020014559
Willy Tarreau74ca5042013-06-11 23:12:07 +020014560ssl_c_key_alg : string
14561 Returns the name of the algorithm used to generate the key of the certificate
14562 presented by the client when the incoming connection was made over an SSL/TLS
14563 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020014564
Willy Tarreau74ca5042013-06-11 23:12:07 +020014565ssl_c_notafter : string
14566 Returns the end date presented by the client as a formatted string
14567 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
14568 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020014569
Willy Tarreau74ca5042013-06-11 23:12:07 +020014570ssl_c_notbefore : string
14571 Returns the start date presented by the client as a formatted string
14572 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
14573 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010014574
Willy Tarreau74ca5042013-06-11 23:12:07 +020014575ssl_c_s_dn([<entry>[,<occ>]]) : string
14576 When the incoming connection was made over an SSL/TLS transport layer,
14577 returns the full distinguished name of the subject of the certificate
14578 presented by the client when no <entry> is specified, or the value of the
14579 first given entry found from the beginning of the DN. If a positive/negative
14580 occurrence number is specified as the optional second argument, it returns
14581 the value of the nth given entry value from the beginning/end of the DN.
14582 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
14583 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010014584
Willy Tarreau74ca5042013-06-11 23:12:07 +020014585ssl_c_serial : binary
14586 Returns the serial of the certificate presented by the client when the
14587 incoming connection was made over an SSL/TLS transport layer. When used for
14588 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020014589
Willy Tarreau74ca5042013-06-11 23:12:07 +020014590ssl_c_sha1 : binary
14591 Returns the SHA-1 fingerprint of the certificate presented by the client when
14592 the incoming connection was made over an SSL/TLS transport layer. This can be
14593 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020014594 Note that the output is binary, so if you want to pass that signature to the
14595 server, you need to encode it in hex or base64, such as in the example below:
14596
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014597 Example:
Willy Tarreau2d0caa32014-07-02 19:01:22 +020014598 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020014599
Willy Tarreau74ca5042013-06-11 23:12:07 +020014600ssl_c_sig_alg : string
14601 Returns the name of the algorithm used to sign the certificate presented by
14602 the client when the incoming connection was made over an SSL/TLS transport
14603 layer.
Emeric Brun87855892012-10-17 17:39:35 +020014604
Willy Tarreau74ca5042013-06-11 23:12:07 +020014605ssl_c_used : boolean
14606 Returns true if current SSL session uses a client certificate even if current
14607 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020014608
Willy Tarreau74ca5042013-06-11 23:12:07 +020014609ssl_c_verify : integer
14610 Returns the verify result error ID when the incoming connection was made over
14611 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
14612 refer to your SSL library's documentation for an exhaustive list of error
14613 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020014614
Willy Tarreau74ca5042013-06-11 23:12:07 +020014615ssl_c_version : integer
14616 Returns the version of the certificate presented by the client when the
14617 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020014618
Emeric Brun43e79582014-10-29 19:03:26 +010014619ssl_f_der : binary
14620 Returns the DER formatted certificate presented by the frontend when the
14621 incoming connection was made over an SSL/TLS transport layer. When used for
14622 an ACL, the value(s) to match against can be passed in hexadecimal form.
14623
Willy Tarreau74ca5042013-06-11 23:12:07 +020014624ssl_f_i_dn([<entry>[,<occ>]]) : string
14625 When the incoming connection was made over an SSL/TLS transport layer,
14626 returns the full distinguished name of the issuer of the certificate
14627 presented by the frontend when no <entry> is specified, or the value of the
14628 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020014629 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020014630 the value of the nth given entry value from the beginning/end of the DN.
14631 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
14632 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020014633
Willy Tarreau74ca5042013-06-11 23:12:07 +020014634ssl_f_key_alg : string
14635 Returns the name of the algorithm used to generate the key of the certificate
14636 presented by the frontend when the incoming connection was made over an
14637 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020014638
Willy Tarreau74ca5042013-06-11 23:12:07 +020014639ssl_f_notafter : string
14640 Returns the end date presented by the frontend as a formatted string
14641 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
14642 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020014643
Willy Tarreau74ca5042013-06-11 23:12:07 +020014644ssl_f_notbefore : string
14645 Returns the start date presented by the frontend as a formatted string
14646 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
14647 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020014648
Willy Tarreau74ca5042013-06-11 23:12:07 +020014649ssl_f_s_dn([<entry>[,<occ>]]) : string
14650 When the incoming connection was made over an SSL/TLS transport layer,
14651 returns the full distinguished name of the subject of the certificate
14652 presented by the frontend when no <entry> is specified, or the value of the
14653 first given entry found from the beginning of the DN. If a positive/negative
14654 occurrence number is specified as the optional second argument, it returns
14655 the value of the nth given entry value from the beginning/end of the DN.
14656 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
14657 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020014658
Willy Tarreau74ca5042013-06-11 23:12:07 +020014659ssl_f_serial : binary
14660 Returns the serial of the certificate presented by the frontend when the
14661 incoming connection was made over an SSL/TLS transport layer. When used for
14662 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020014663
Emeric Brun55f4fa82014-04-30 17:11:25 +020014664ssl_f_sha1 : binary
14665 Returns the SHA-1 fingerprint of the certificate presented by the frontend
14666 when the incoming connection was made over an SSL/TLS transport layer. This
14667 can be used to know which certificate was chosen using SNI.
14668
Willy Tarreau74ca5042013-06-11 23:12:07 +020014669ssl_f_sig_alg : string
14670 Returns the name of the algorithm used to sign the certificate presented by
14671 the frontend when the incoming connection was made over an SSL/TLS transport
14672 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020014673
Willy Tarreau74ca5042013-06-11 23:12:07 +020014674ssl_f_version : integer
14675 Returns the version of the certificate presented by the frontend when the
14676 incoming connection was made over an SSL/TLS transport layer.
14677
14678ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020014679 Returns true when the front connection was made via an SSL/TLS transport
14680 layer and is locally deciphered. This means it has matched a socket declared
14681 with a "bind" line having the "ssl" option.
14682
Willy Tarreau74ca5042013-06-11 23:12:07 +020014683 Example :
14684 # This passes "X-Proto: https" to servers when client connects over SSL
14685 listen http-https
14686 bind :80
14687 bind :443 ssl crt /etc/haproxy.pem
14688 http-request add-header X-Proto https if { ssl_fc }
14689
14690ssl_fc_alg_keysize : integer
14691 Returns the symmetric cipher key size supported in bits when the incoming
14692 connection was made over an SSL/TLS transport layer.
14693
14694ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014695 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020014696 incoming connection made via a TLS transport layer and locally deciphered by
14697 haproxy. The result is a string containing the protocol name advertised by
14698 the client. The SSL library must have been built with support for TLS
14699 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
14700 not advertised unless the "alpn" keyword on the "bind" line specifies a
14701 protocol list. Also, nothing forces the client to pick a protocol from this
14702 list, any other one may be requested. The TLS ALPN extension is meant to
14703 replace the TLS NPN extension. See also "ssl_fc_npn".
14704
Willy Tarreau74ca5042013-06-11 23:12:07 +020014705ssl_fc_cipher : string
14706 Returns the name of the used cipher when the incoming connection was made
14707 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020014708
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010014709ssl_fc_cipherlist_bin : binary
14710 Returns the binary form of the client hello cipher list. The maximum returned
14711 value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010014712 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010014713
14714ssl_fc_cipherlist_hex : string
14715 Returns the binary form of the client hello cipher list encoded as
14716 hexadecimal. The maximum returned value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010014717 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010014718
14719ssl_fc_cipherlist_str : string
14720 Returns the decoded text form of the client hello cipher list. The maximum
14721 number of ciphers returned is according with the value of
14722 "tune.ssl.capture-cipherlist-size". Note that this sample-fetch is only
Davor Ocelice9ed2812017-12-25 17:49:28 +010014723 available with OpenSSL >= 1.0.2. If the function is not enabled, this
Emmanuel Hocdetddcde192017-09-01 17:32:08 +020014724 sample-fetch returns the hash like "ssl_fc_cipherlist_xxh".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010014725
14726ssl_fc_cipherlist_xxh : integer
14727 Returns a xxh64 of the cipher list. This hash can be return only is the value
14728 "tune.ssl.capture-cipherlist-size" is set greater than 0, however the hash
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010014729 take in account all the data of the cipher list.
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010014730
Willy Tarreau74ca5042013-06-11 23:12:07 +020014731ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020014732 Returns true if a client certificate is present in an incoming connection over
14733 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010014734 Note: on SSL session resumption with Session ID or TLS ticket, client
14735 certificate is not present in the current connection but may be retrieved
14736 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
14737 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020014738
Olivier Houchardccaa7de2017-10-02 11:51:03 +020014739ssl_fc_has_early : boolean
14740 Returns true if early data were sent, and the handshake didn't happen yet. As
14741 it has security implications, it is useful to be able to refuse those, or
14742 wait until the handshake happened.
14743
Willy Tarreau74ca5042013-06-11 23:12:07 +020014744ssl_fc_has_sni : boolean
14745 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020014746 in an incoming connection was made over an SSL/TLS transport layer. Returns
14747 true when the incoming connection presents a TLS SNI field. This requires
14748 that the SSL library is build with support for TLS extensions enabled (check
14749 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020014750
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020014751ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020014752 Returns true if the SSL/TLS session has been resumed through the use of
Jérôme Magnin4a326cb2018-01-15 14:01:17 +010014753 SSL session cache or TLS tickets on an incoming connection over an SSL/TLS
14754 transport layer.
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020014755
Willy Tarreau74ca5042013-06-11 23:12:07 +020014756ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014757 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020014758 made via a TLS transport layer and locally deciphered by haproxy. The result
14759 is a string containing the protocol name advertised by the client. The SSL
14760 library must have been built with support for TLS extensions enabled (check
14761 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
14762 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
14763 forces the client to pick a protocol from this list, any other one may be
14764 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020014765
Willy Tarreau74ca5042013-06-11 23:12:07 +020014766ssl_fc_protocol : string
14767 Returns the name of the used protocol when the incoming connection was made
14768 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020014769
Emeric Brunb73a9b02014-04-30 18:49:19 +020014770ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040014771 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020014772 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
14773 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040014774
Willy Tarreau74ca5042013-06-11 23:12:07 +020014775ssl_fc_session_id : binary
14776 Returns the SSL ID of the front connection when the incoming connection was
14777 made over an SSL/TLS transport layer. It is useful to stick a given client to
14778 a server. It is important to note that some browsers refresh their session ID
14779 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020014780
Patrick Hemmere0275472018-04-28 19:15:51 -040014781ssl_fc_session_key : binary
14782 Returns the SSL session master key of the front connection when the incoming
14783 connection was made over an SSL/TLS transport layer. It is useful to decrypt
14784 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
14785 BoringSSL.
14786
14787
Willy Tarreau74ca5042013-06-11 23:12:07 +020014788ssl_fc_sni : string
14789 This extracts the Server Name Indication TLS extension (SNI) field from an
14790 incoming connection made via an SSL/TLS transport layer and locally
14791 deciphered by haproxy. The result (when present) typically is a string
14792 matching the HTTPS host name (253 chars or less). The SSL library must have
14793 been built with support for TLS extensions enabled (check haproxy -vv).
14794
14795 This fetch is different from "req_ssl_sni" above in that it applies to the
14796 connection being deciphered by haproxy and not to SSL contents being blindly
14797 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020014798 requires that the SSL library is build with support for TLS extensions
14799 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020014800
Willy Tarreau74ca5042013-06-11 23:12:07 +020014801 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020014802 ssl_fc_sni_end : suffix match
14803 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020014804
Willy Tarreau74ca5042013-06-11 23:12:07 +020014805ssl_fc_use_keysize : integer
14806 Returns the symmetric cipher key size used in bits when the incoming
14807 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020014808
Willy Tarreaub6fb4202008-07-20 11:18:28 +020014809
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200148107.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020014811------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020014812
Willy Tarreau74ca5042013-06-11 23:12:07 +020014813Fetching samples from buffer contents is a bit different from the previous
14814sample fetches above because the sampled data are ephemeral. These data can
14815only be used when they're available and will be lost when they're forwarded.
14816For this reason, samples fetched from buffer contents during a request cannot
14817be used in a response for example. Even while the data are being fetched, they
14818can change. Sometimes it is necessary to set some delays or combine multiple
14819sample fetch methods to ensure that the expected data are complete and usable,
14820for example through TCP request content inspection. Please see the "tcp-request
14821content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020014822
Willy Tarreau74ca5042013-06-11 23:12:07 +020014823payload(<offset>,<length>) : binary (deprecated)
Davor Ocelice9ed2812017-12-25 17:49:28 +010014824 This is an alias for "req.payload" when used in the context of a request (e.g.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014825 "stick on", "stick match"), and for "res.payload" when used in the context of
14826 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010014827
Willy Tarreau74ca5042013-06-11 23:12:07 +020014828payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
14829 This is an alias for "req.payload_lv" when used in the context of a request
Davor Ocelice9ed2812017-12-25 17:49:28 +010014830 (e.g. "stick on", "stick match"), and for "res.payload_lv" when used in the
Willy Tarreau74ca5042013-06-11 23:12:07 +020014831 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010014832
Thierry FOURNIERd7d88812017-04-19 15:15:14 +020014833req.hdrs : string
14834 Returns the current request headers as string including the last empty line
14835 separating headers from the request body. The last empty line can be used to
14836 detect a truncated header block. This sample fetch is useful for some SPOE
14837 headers analyzers and for advanced logging.
14838
Thierry FOURNIER5617dce2017-04-09 05:38:19 +020014839req.hdrs_bin : binary
14840 Returns the current request headers contained in preparsed binary form. This
14841 is useful for offloading some processing with SPOE. Each string is described
14842 by a length followed by the number of bytes indicated in the length. The
14843 length is represented using the variable integer encoding detailed in the
14844 SPOE documentation. The end of the list is marked by a couple of empty header
14845 names and values (length of 0 for both).
14846
14847 *(<str:header-name><str:header-value>)<empty string><empty string>
14848
14849 int: refer to the SPOE documentation for the encoding
14850 str: <int:length><bytes>
14851
Willy Tarreau74ca5042013-06-11 23:12:07 +020014852req.len : integer
14853req_len : integer (deprecated)
14854 Returns an integer value corresponding to the number of bytes present in the
14855 request buffer. This is mostly used in ACL. It is important to understand
14856 that this test does not return false as long as the buffer is changing. This
14857 means that a check with equality to zero will almost always immediately match
14858 at the beginning of the session, while a test for more data will wait for
14859 that data to come in and return false only when haproxy is certain that no
14860 more data will come in. This test was designed to be used with TCP request
14861 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014862
Willy Tarreau74ca5042013-06-11 23:12:07 +020014863req.payload(<offset>,<length>) : binary
14864 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020014865 in the request buffer. As a special case, if the <length> argument is zero,
14866 the the whole buffer from <offset> to the end is extracted. This can be used
14867 with ACLs in order to check for the presence of some content in a buffer at
14868 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014869
Willy Tarreau74ca5042013-06-11 23:12:07 +020014870 ACL alternatives :
14871 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014872
Willy Tarreau74ca5042013-06-11 23:12:07 +020014873req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
14874 This extracts a binary block whose size is specified at <offset1> for <length>
14875 bytes, and which starts at <offset2> if specified or just after the length in
14876 the request buffer. The <offset2> parameter also supports relative offsets if
14877 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014878
Willy Tarreau74ca5042013-06-11 23:12:07 +020014879 ACL alternatives :
14880 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014881
Willy Tarreau74ca5042013-06-11 23:12:07 +020014882 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014883
Willy Tarreau74ca5042013-06-11 23:12:07 +020014884req.proto_http : boolean
14885req_proto_http : boolean (deprecated)
14886 Returns true when data in the request buffer look like HTTP and correctly
14887 parses as such. It is the same parser as the common HTTP request parser which
14888 is used so there should be no surprises. The test does not match until the
14889 request is complete, failed or timed out. This test may be used to report the
14890 protocol in TCP logs, but the biggest use is to block TCP request analysis
14891 until a complete HTTP request is present in the buffer, for example to track
14892 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014893
Willy Tarreau74ca5042013-06-11 23:12:07 +020014894 Example:
14895 # track request counts per "base" (concatenation of Host+URL)
14896 tcp-request inspect-delay 10s
14897 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014898 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020014899
Willy Tarreau74ca5042013-06-11 23:12:07 +020014900req.rdp_cookie([<name>]) : string
14901rdp_cookie([<name>]) : string (deprecated)
14902 When the request buffer looks like the RDP protocol, extracts the RDP cookie
14903 <name>, or any cookie if unspecified. The parser only checks for the first
14904 cookie, as illustrated in the RDP protocol specification. The cookie name is
14905 case insensitive. Generally the "MSTS" cookie name will be used, as it can
14906 contain the user name of the client connecting to the server if properly
14907 configured on the client. The "MSTSHASH" cookie is often used as well for
14908 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014909
Willy Tarreau74ca5042013-06-11 23:12:07 +020014910 This differs from "balance rdp-cookie" in that any balancing algorithm may be
14911 used and thus the distribution of clients to backend servers is not linked to
14912 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
14913 such as "balance roundrobin" or "balance leastconn" will lead to a more even
14914 distribution of clients to backend servers than the hash used by "balance
14915 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014916
Willy Tarreau74ca5042013-06-11 23:12:07 +020014917 ACL derivatives :
14918 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014919
Willy Tarreau74ca5042013-06-11 23:12:07 +020014920 Example :
14921 listen tse-farm
14922 bind 0.0.0.0:3389
14923 # wait up to 5s for an RDP cookie in the request
14924 tcp-request inspect-delay 5s
14925 tcp-request content accept if RDP_COOKIE
14926 # apply RDP cookie persistence
14927 persist rdp-cookie
14928 # Persist based on the mstshash cookie
14929 # This is only useful makes sense if
14930 # balance rdp-cookie is not used
14931 stick-table type string size 204800
14932 stick on req.rdp_cookie(mstshash)
14933 server srv1 1.1.1.1:3389
14934 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014935
Willy Tarreau74ca5042013-06-11 23:12:07 +020014936 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
14937 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014938
Willy Tarreau74ca5042013-06-11 23:12:07 +020014939req.rdp_cookie_cnt([name]) : integer
14940rdp_cookie_cnt([name]) : integer (deprecated)
14941 Tries to parse the request buffer as RDP protocol, then returns an integer
14942 corresponding to the number of RDP cookies found. If an optional cookie name
14943 is passed, only cookies matching this name are considered. This is mostly
14944 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014945
Willy Tarreau74ca5042013-06-11 23:12:07 +020014946 ACL derivatives :
14947 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014948
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020014949req.ssl_ec_ext : boolean
14950 Returns a boolean identifying if client sent the Supported Elliptic Curves
14951 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020014952 message. This can be used to present ECC compatible clients with EC
14953 certificate and to use RSA for all others, on the same IP address. Note that
14954 this only applies to raw contents found in the request buffer and not to
14955 contents deciphered via an SSL data layer, so this will not work with "bind"
14956 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020014957
Willy Tarreau74ca5042013-06-11 23:12:07 +020014958req.ssl_hello_type : integer
14959req_ssl_hello_type : integer (deprecated)
14960 Returns an integer value containing the type of the SSL hello message found
14961 in the request buffer if the buffer contains data that parse as a complete
14962 SSL (v3 or superior) client hello message. Note that this only applies to raw
14963 contents found in the request buffer and not to contents deciphered via an
14964 SSL data layer, so this will not work with "bind" lines having the "ssl"
14965 option. This is mostly used in ACL to detect presence of an SSL hello message
14966 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014967
Willy Tarreau74ca5042013-06-11 23:12:07 +020014968req.ssl_sni : string
14969req_ssl_sni : string (deprecated)
14970 Returns a string containing the value of the Server Name TLS extension sent
14971 by a client in a TLS stream passing through the request buffer if the buffer
14972 contains data that parse as a complete SSL (v3 or superior) client hello
14973 message. Note that this only applies to raw contents found in the request
14974 buffer and not to contents deciphered via an SSL data layer, so this will not
14975 work with "bind" lines having the "ssl" option. SNI normally contains the
14976 name of the host the client tries to connect to (for recent browsers). SNI is
14977 useful for allowing or denying access to certain hosts when SSL/TLS is used
14978 by the client. This test was designed to be used with TCP request content
14979 inspection. If content switching is needed, it is recommended to first wait
14980 for a complete client hello (type 1), like in the example below. See also
14981 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014982
Willy Tarreau74ca5042013-06-11 23:12:07 +020014983 ACL derivatives :
14984 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014985
Willy Tarreau74ca5042013-06-11 23:12:07 +020014986 Examples :
14987 # Wait for a client hello for at most 5 seconds
14988 tcp-request inspect-delay 5s
14989 tcp-request content accept if { req_ssl_hello_type 1 }
14990 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
14991 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014992
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053014993req.ssl_st_ext : integer
14994 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
14995 Returns 1 if the client sent SessionTicket TLS Extension
14996 Returns 2 if the client also sent non-zero length TLS SessionTicket
14997 Note that this only applies to raw contents found in the request buffer and
14998 not to contents deciphered via an SSL data layer, so this will not work with
14999 "bind" lines having the "ssl" option. This can for example be used to detect
15000 whether the client sent a SessionTicket or not and stick it accordingly, if
15001 no SessionTicket then stick on SessionID or don't stick as there's no server
15002 side state is there when SessionTickets are in use.
15003
Willy Tarreau74ca5042013-06-11 23:12:07 +020015004req.ssl_ver : integer
15005req_ssl_ver : integer (deprecated)
15006 Returns an integer value containing the version of the SSL/TLS protocol of a
15007 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
15008 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
15009 composed of the major version multiplied by 65536, added to the minor
15010 version. Note that this only applies to raw contents found in the request
15011 buffer and not to contents deciphered via an SSL data layer, so this will not
15012 work with "bind" lines having the "ssl" option. The ACL version of the test
Davor Ocelice9ed2812017-12-25 17:49:28 +010015013 matches against a decimal notation in the form MAJOR.MINOR (e.g. 3.1). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020015014 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015015
Willy Tarreau74ca5042013-06-11 23:12:07 +020015016 ACL derivatives :
15017 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010015018
Willy Tarreau47e8eba2013-09-11 23:28:46 +020015019res.len : integer
15020 Returns an integer value corresponding to the number of bytes present in the
15021 response buffer. This is mostly used in ACL. It is important to understand
15022 that this test does not return false as long as the buffer is changing. This
15023 means that a check with equality to zero will almost always immediately match
15024 at the beginning of the session, while a test for more data will wait for
15025 that data to come in and return false only when haproxy is certain that no
15026 more data will come in. This test was designed to be used with TCP response
15027 content inspection.
15028
Willy Tarreau74ca5042013-06-11 23:12:07 +020015029res.payload(<offset>,<length>) : binary
15030 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020015031 in the response buffer. As a special case, if the <length> argument is zero,
15032 the the whole buffer from <offset> to the end is extracted. This can be used
15033 with ACLs in order to check for the presence of some content in a buffer at
15034 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015035
Willy Tarreau74ca5042013-06-11 23:12:07 +020015036res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
15037 This extracts a binary block whose size is specified at <offset1> for <length>
15038 bytes, and which starts at <offset2> if specified or just after the length in
15039 the response buffer. The <offset2> parameter also supports relative offsets
15040 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015041
Willy Tarreau74ca5042013-06-11 23:12:07 +020015042 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015043
Willy Tarreau971f7b62015-09-29 14:06:59 +020015044res.ssl_hello_type : integer
15045rep_ssl_hello_type : integer (deprecated)
15046 Returns an integer value containing the type of the SSL hello message found
15047 in the response buffer if the buffer contains data that parses as a complete
15048 SSL (v3 or superior) hello message. Note that this only applies to raw
15049 contents found in the response buffer and not to contents deciphered via an
15050 SSL data layer, so this will not work with "server" lines having the "ssl"
15051 option. This is mostly used in ACL to detect presence of an SSL hello message
15052 that is supposed to contain an SSL session ID usable for stickiness.
15053
Willy Tarreau74ca5042013-06-11 23:12:07 +020015054wait_end : boolean
15055 This fetch either returns true when the inspection period is over, or does
15056 not fetch. It is only used in ACLs, in conjunction with content analysis to
Davor Ocelice9ed2812017-12-25 17:49:28 +010015057 avoid returning a wrong verdict early. It may also be used to delay some
Willy Tarreau74ca5042013-06-11 23:12:07 +020015058 actions, such as a delayed reject for some special addresses. Since it either
15059 stops the rules evaluation or immediately returns true, it is recommended to
Davor Ocelice9ed2812017-12-25 17:49:28 +010015060 use this acl as the last one in a rule. Please note that the default ACL
Willy Tarreau74ca5042013-06-11 23:12:07 +020015061 "WAIT_END" is always usable without prior declaration. This test was designed
15062 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015063
Willy Tarreau74ca5042013-06-11 23:12:07 +020015064 Examples :
15065 # delay every incoming request by 2 seconds
15066 tcp-request inspect-delay 2s
15067 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010015068
Willy Tarreau74ca5042013-06-11 23:12:07 +020015069 # don't immediately tell bad guys they are rejected
15070 tcp-request inspect-delay 10s
15071 acl goodguys src 10.0.0.0/24
15072 acl badguys src 10.0.1.0/24
15073 tcp-request content accept if goodguys
15074 tcp-request content reject if badguys WAIT_END
15075 tcp-request content reject
15076
15077
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200150787.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020015079--------------------------------------
15080
15081It is possible to fetch samples from HTTP contents, requests and responses.
15082This application layer is also called layer 7. It is only possible to fetch the
15083data in this section when a full HTTP request or response has been parsed from
15084its respective request or response buffer. This is always the case with all
15085HTTP specific rules and for sections running with "mode http". When using TCP
15086content inspection, it may be necessary to support an inspection delay in order
15087to let the request or response come in first. These fetches may require a bit
15088more CPU resources than the layer 4 ones, but not much since the request and
15089response are indexed.
15090
15091base : string
15092 This returns the concatenation of the first Host header and the path part of
15093 the request, which starts at the first slash and ends before the question
15094 mark. It can be useful in virtual hosted environments to detect URL abuses as
15095 well as to improve shared caches efficiency. Using this with a limited size
15096 stick table also allows one to collect statistics about most commonly
15097 requested objects by host/path. With ACLs it can allow simple content
15098 switching rules involving the host and the path at the same time, such as
15099 "www.example.com/favicon.ico". See also "path" and "uri".
15100
15101 ACL derivatives :
15102 base : exact string match
15103 base_beg : prefix match
15104 base_dir : subdir match
15105 base_dom : domain match
15106 base_end : suffix match
15107 base_len : length match
15108 base_reg : regex match
15109 base_sub : substring match
15110
15111base32 : integer
15112 This returns a 32-bit hash of the value returned by the "base" fetch method
15113 above. This is useful to track per-URL activity on high traffic sites without
15114 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020015115 memory. The output type is an unsigned integer. The hash function used is
15116 SDBM with full avalanche on the output. Technically, base32 is exactly equal
15117 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020015118
15119base32+src : binary
15120 This returns the concatenation of the base32 fetch above and the src fetch
15121 below. The resulting type is of type binary, with a size of 8 or 20 bytes
15122 depending on the source address family. This can be used to track per-IP,
15123 per-URL counters.
15124
William Lallemand65ad6e12014-01-31 15:08:02 +010015125capture.req.hdr(<idx>) : string
15126 This extracts the content of the header captured by the "capture request
15127 header", idx is the position of the capture keyword in the configuration.
15128 The first entry is an index of 0. See also: "capture request header".
15129
15130capture.req.method : string
15131 This extracts the METHOD of an HTTP request. It can be used in both request
15132 and response. Unlike "method", it can be used in both request and response
15133 because it's allocated.
15134
15135capture.req.uri : string
15136 This extracts the request's URI, which starts at the first slash and ends
15137 before the first space in the request (without the host part). Unlike "path"
15138 and "url", it can be used in both request and response because it's
15139 allocated.
15140
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020015141capture.req.ver : string
15142 This extracts the request's HTTP version and returns either "HTTP/1.0" or
15143 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
15144 logs because it relies on a persistent flag.
15145
William Lallemand65ad6e12014-01-31 15:08:02 +010015146capture.res.hdr(<idx>) : string
15147 This extracts the content of the header captured by the "capture response
15148 header", idx is the position of the capture keyword in the configuration.
15149 The first entry is an index of 0.
15150 See also: "capture response header"
15151
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020015152capture.res.ver : string
15153 This extracts the response's HTTP version and returns either "HTTP/1.0" or
15154 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
15155 persistent flag.
15156
Willy Tarreaua5910cc2015-05-02 00:46:08 +020015157req.body : binary
15158 This returns the HTTP request's available body as a block of data. It
15159 requires that the request body has been buffered made available using
15160 "option http-buffer-request". In case of chunked-encoded body, currently only
15161 the first chunk is analyzed.
15162
Thierry FOURNIER9826c772015-05-20 15:50:54 +020015163req.body_param([<name>) : string
15164 This fetch assumes that the body of the POST request is url-encoded. The user
15165 can check if the "content-type" contains the value
15166 "application/x-www-form-urlencoded". This extracts the first occurrence of the
15167 parameter <name> in the body, which ends before '&'. The parameter name is
15168 case-sensitive. If no name is given, any parameter will match, and the first
15169 one will be returned. The result is a string corresponding to the value of the
15170 parameter <name> as presented in the request body (no URL decoding is
15171 performed). Note that the ACL version of this fetch iterates over multiple
15172 parameters and will iteratively report all parameters values if no name is
15173 given.
15174
Willy Tarreaua5910cc2015-05-02 00:46:08 +020015175req.body_len : integer
15176 This returns the length of the HTTP request's available body in bytes. It may
15177 be lower than the advertised length if the body is larger than the buffer. It
15178 requires that the request body has been buffered made available using
15179 "option http-buffer-request".
15180
15181req.body_size : integer
15182 This returns the advertised length of the HTTP request's body in bytes. It
15183 will represent the advertised Content-Length header, or the size of the first
15184 chunk in case of chunked encoding. In order to parse the chunks, it requires
15185 that the request body has been buffered made available using
15186 "option http-buffer-request".
15187
Willy Tarreau74ca5042013-06-11 23:12:07 +020015188req.cook([<name>]) : string
15189cook([<name>]) : string (deprecated)
15190 This extracts the last occurrence of the cookie name <name> on a "Cookie"
15191 header line from the request, and returns its value as string. If no name is
15192 specified, the first cookie value is returned. When used with ACLs, all
15193 matching cookies are evaluated. Spaces around the name and the value are
15194 ignored as requested by the Cookie header specification (RFC6265). The cookie
15195 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
15196 well return an empty value if it is present. Use the "found" match to detect
15197 presence. Use the res.cook() variant for response cookies sent by the server.
15198
15199 ACL derivatives :
15200 cook([<name>]) : exact string match
15201 cook_beg([<name>]) : prefix match
15202 cook_dir([<name>]) : subdir match
15203 cook_dom([<name>]) : domain match
15204 cook_end([<name>]) : suffix match
15205 cook_len([<name>]) : length match
15206 cook_reg([<name>]) : regex match
15207 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010015208
Willy Tarreau74ca5042013-06-11 23:12:07 +020015209req.cook_cnt([<name>]) : integer
15210cook_cnt([<name>]) : integer (deprecated)
15211 Returns an integer value representing the number of occurrences of the cookie
15212 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015213
Willy Tarreau74ca5042013-06-11 23:12:07 +020015214req.cook_val([<name>]) : integer
15215cook_val([<name>]) : integer (deprecated)
15216 This extracts the last occurrence of the cookie name <name> on a "Cookie"
15217 header line from the request, and converts its value to an integer which is
15218 returned. If no name is specified, the first cookie value is returned. When
15219 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020015220
Willy Tarreau74ca5042013-06-11 23:12:07 +020015221cookie([<name>]) : string (deprecated)
15222 This extracts the last occurrence of the cookie name <name> on a "Cookie"
15223 header line from the request, or a "Set-Cookie" header from the response, and
15224 returns its value as a string. A typical use is to get multiple clients
15225 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020015226 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020015227 multi-peer synchronization and state keeping across restarts. If no name is
15228 specified, the first cookie value is returned. This fetch should not be used
15229 anymore and should be replaced by req.cook() or res.cook() instead as it
15230 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015231
Willy Tarreau74ca5042013-06-11 23:12:07 +020015232hdr([<name>[,<occ>]]) : string
15233 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
15234 used on responses. Please refer to these respective fetches for more details.
15235 In case of doubt about the fetch direction, please use the explicit ones.
15236 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015237 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015238
Willy Tarreau74ca5042013-06-11 23:12:07 +020015239req.fhdr(<name>[,<occ>]) : string
15240 This extracts the last occurrence of header <name> in an HTTP request. When
15241 used from an ACL, all occurrences are iterated over until a match is found.
15242 Optionally, a specific occurrence might be specified as a position number.
15243 Positive values indicate a position from the first occurrence, with 1 being
15244 the first one. Negative values indicate positions relative to the last one,
15245 with -1 being the last one. It differs from req.hdr() in that any commas
15246 present in the value are returned and are not used as delimiters. This is
15247 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015248
Willy Tarreau74ca5042013-06-11 23:12:07 +020015249req.fhdr_cnt([<name>]) : integer
15250 Returns an integer value representing the number of occurrences of request
15251 header field name <name>, or the total number of header fields if <name> is
15252 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
15253 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015254
Willy Tarreau74ca5042013-06-11 23:12:07 +020015255req.hdr([<name>[,<occ>]]) : string
15256 This extracts the last occurrence of header <name> in an HTTP request. When
15257 used from an ACL, all occurrences are iterated over until a match is found.
15258 Optionally, a specific occurrence might be specified as a position number.
15259 Positive values indicate a position from the first occurrence, with 1 being
15260 the first one. Negative values indicate positions relative to the last one,
15261 with -1 being the last one. A typical use is with the X-Forwarded-For header
15262 once converted to IP, associated with an IP stick-table. The function
15263 considers any comma as a delimiter for distinct values. If full-line headers
Lukas Tribus23953682017-04-28 13:24:30 +000015264 are desired instead, use req.fhdr(). Please carefully check RFC7231 to know
Willy Tarreau74ca5042013-06-11 23:12:07 +020015265 how certain headers are supposed to be parsed. Also, some of them are case
Davor Ocelice9ed2812017-12-25 17:49:28 +010015266 insensitive (e.g. Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010015267
Willy Tarreau74ca5042013-06-11 23:12:07 +020015268 ACL derivatives :
15269 hdr([<name>[,<occ>]]) : exact string match
15270 hdr_beg([<name>[,<occ>]]) : prefix match
15271 hdr_dir([<name>[,<occ>]]) : subdir match
15272 hdr_dom([<name>[,<occ>]]) : domain match
15273 hdr_end([<name>[,<occ>]]) : suffix match
15274 hdr_len([<name>[,<occ>]]) : length match
15275 hdr_reg([<name>[,<occ>]]) : regex match
15276 hdr_sub([<name>[,<occ>]]) : substring match
15277
15278req.hdr_cnt([<name>]) : integer
15279hdr_cnt([<header>]) : integer (deprecated)
15280 Returns an integer value representing the number of occurrences of request
15281 header field name <name>, or the total number of header field values if
15282 <name> is not specified. It is important to remember that one header line may
15283 count as several headers if it has several values. The function considers any
15284 comma as a delimiter for distinct values. If full-line headers are desired
15285 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
15286 detect presence, absence or abuse of a specific header, as well as to block
15287 request smuggling attacks by rejecting requests which contain more than one
15288 of certain headers. See "req.hdr" for more information on header matching.
15289
15290req.hdr_ip([<name>[,<occ>]]) : ip
15291hdr_ip([<name>[,<occ>]]) : ip (deprecated)
15292 This extracts the last occurrence of header <name> in an HTTP request,
15293 converts it to an IPv4 or IPv6 address and returns this address. When used
15294 with ACLs, all occurrences are checked, and if <name> is omitted, every value
15295 of every header is checked. Optionally, a specific occurrence might be
15296 specified as a position number. Positive values indicate a position from the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015297 first occurrence, with 1 being the first one. Negative values indicate
Willy Tarreau74ca5042013-06-11 23:12:07 +020015298 positions relative to the last one, with -1 being the last one. A typical use
15299 is with the X-Forwarded-For and X-Client-IP headers.
15300
15301req.hdr_val([<name>[,<occ>]]) : integer
15302hdr_val([<name>[,<occ>]]) : integer (deprecated)
15303 This extracts the last occurrence of header <name> in an HTTP request, and
15304 converts it to an integer value. When used with ACLs, all occurrences are
15305 checked, and if <name> is omitted, every value of every header is checked.
15306 Optionally, a specific occurrence might be specified as a position number.
15307 Positive values indicate a position from the first occurrence, with 1 being
15308 the first one. Negative values indicate positions relative to the last one,
15309 with -1 being the last one. A typical use is with the X-Forwarded-For header.
15310
15311http_auth(<userlist>) : boolean
15312 Returns a boolean indicating whether the authentication data received from
15313 the client match a username & password stored in the specified userlist. This
15314 fetch function is not really useful outside of ACLs. Currently only http
15315 basic auth is supported.
15316
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010015317http_auth_group(<userlist>) : string
15318 Returns a string corresponding to the user name found in the authentication
15319 data received from the client if both the user name and password are valid
15320 according to the specified userlist. The main purpose is to use it in ACLs
15321 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015322 This fetch function is not really useful outside of ACLs. Currently only http
15323 basic auth is supported.
15324
15325 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010015326 http_auth_group(<userlist>) : group ...
15327 Returns true when the user extracted from the request and whose password is
15328 valid according to the specified userlist belongs to at least one of the
15329 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015330
15331http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020015332 Returns true when the request being processed is the first one of the
15333 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020015334 from some requests when a request is not the first one, or to help grouping
15335 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020015336
Willy Tarreau74ca5042013-06-11 23:12:07 +020015337method : integer + string
15338 Returns an integer value corresponding to the method in the HTTP request. For
15339 example, "GET" equals 1 (check sources to establish the matching). Value 9
15340 means "other method" and may be converted to a string extracted from the
15341 stream. This should not be used directly as a sample, this is only meant to
15342 be used from ACLs, which transparently convert methods from patterns to these
15343 integer + string values. Some predefined ACL already check for most common
15344 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015345
Willy Tarreau74ca5042013-06-11 23:12:07 +020015346 ACL derivatives :
15347 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020015348
Willy Tarreau74ca5042013-06-11 23:12:07 +020015349 Example :
15350 # only accept GET and HEAD requests
15351 acl valid_method method GET HEAD
15352 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020015353
Willy Tarreau74ca5042013-06-11 23:12:07 +020015354path : string
15355 This extracts the request's URL path, which starts at the first slash and
15356 ends before the question mark (without the host part). A typical use is with
15357 prefetch-capable caches, and with portals which need to aggregate multiple
15358 information from databases and keep them in caches. Note that with outgoing
15359 caches, it would be wiser to use "url" instead. With ACLs, it's typically
Davor Ocelice9ed2812017-12-25 17:49:28 +010015360 used to match exact file names (e.g. "/login.php"), or directory parts using
Willy Tarreau74ca5042013-06-11 23:12:07 +020015361 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015362
Willy Tarreau74ca5042013-06-11 23:12:07 +020015363 ACL derivatives :
15364 path : exact string match
15365 path_beg : prefix match
15366 path_dir : subdir match
15367 path_dom : domain match
15368 path_end : suffix match
15369 path_len : length match
15370 path_reg : regex match
15371 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020015372
Willy Tarreau49ad95c2015-01-19 15:06:26 +010015373query : string
15374 This extracts the request's query string, which starts after the first
15375 question mark. If no question mark is present, this fetch returns nothing. If
15376 a question mark is present but nothing follows, it returns an empty string.
15377 This means it's possible to easily know whether a query string is present
Tim Düsterhus4896c442016-11-29 02:15:19 +010015378 using the "found" matching method. This fetch is the complement of "path"
Willy Tarreau49ad95c2015-01-19 15:06:26 +010015379 which stops before the question mark.
15380
Willy Tarreaueb27ec72015-02-20 13:55:29 +010015381req.hdr_names([<delim>]) : string
15382 This builds a string made from the concatenation of all header names as they
15383 appear in the request when the rule is evaluated. The default delimiter is
15384 the comma (',') but it may be overridden as an optional argument <delim>. In
15385 this case, only the first character of <delim> is considered.
15386
Willy Tarreau74ca5042013-06-11 23:12:07 +020015387req.ver : string
15388req_ver : string (deprecated)
15389 Returns the version string from the HTTP request, for example "1.1". This can
15390 be useful for logs, but is mostly there for ACL. Some predefined ACL already
15391 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015392
Willy Tarreau74ca5042013-06-11 23:12:07 +020015393 ACL derivatives :
15394 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020015395
Willy Tarreau74ca5042013-06-11 23:12:07 +020015396res.comp : boolean
15397 Returns the boolean "true" value if the response has been compressed by
15398 HAProxy, otherwise returns boolean "false". This may be used to add
15399 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015400
Willy Tarreau74ca5042013-06-11 23:12:07 +020015401res.comp_algo : string
15402 Returns a string containing the name of the algorithm used if the response
15403 was compressed by HAProxy, for example : "deflate". This may be used to add
15404 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015405
Willy Tarreau74ca5042013-06-11 23:12:07 +020015406res.cook([<name>]) : string
15407scook([<name>]) : string (deprecated)
15408 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
15409 header line from the response, and returns its value as string. If no name is
15410 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020015411
Willy Tarreau74ca5042013-06-11 23:12:07 +020015412 ACL derivatives :
15413 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020015414
Willy Tarreau74ca5042013-06-11 23:12:07 +020015415res.cook_cnt([<name>]) : integer
15416scook_cnt([<name>]) : integer (deprecated)
15417 Returns an integer value representing the number of occurrences of the cookie
15418 <name> in the response, or all cookies if <name> is not specified. This is
15419 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015420
Willy Tarreau74ca5042013-06-11 23:12:07 +020015421res.cook_val([<name>]) : integer
15422scook_val([<name>]) : integer (deprecated)
15423 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
15424 header line from the response, and converts its value to an integer which is
15425 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015426
Willy Tarreau74ca5042013-06-11 23:12:07 +020015427res.fhdr([<name>[,<occ>]]) : string
15428 This extracts the last occurrence of header <name> in an HTTP response, or of
15429 the last header if no <name> is specified. Optionally, a specific occurrence
15430 might be specified as a position number. Positive values indicate a position
15431 from the first occurrence, with 1 being the first one. Negative values
15432 indicate positions relative to the last one, with -1 being the last one. It
15433 differs from res.hdr() in that any commas present in the value are returned
15434 and are not used as delimiters. If this is not desired, the res.hdr() fetch
15435 should be used instead. This is sometimes useful with headers such as Date or
15436 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015437
Willy Tarreau74ca5042013-06-11 23:12:07 +020015438res.fhdr_cnt([<name>]) : integer
15439 Returns an integer value representing the number of occurrences of response
15440 header field name <name>, or the total number of header fields if <name> is
15441 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
15442 the number of full line headers and does not stop on commas. If this is not
15443 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015444
Willy Tarreau74ca5042013-06-11 23:12:07 +020015445res.hdr([<name>[,<occ>]]) : string
15446shdr([<name>[,<occ>]]) : string (deprecated)
15447 This extracts the last occurrence of header <name> in an HTTP response, or of
15448 the last header if no <name> is specified. Optionally, a specific occurrence
15449 might be specified as a position number. Positive values indicate a position
15450 from the first occurrence, with 1 being the first one. Negative values
15451 indicate positions relative to the last one, with -1 being the last one. This
15452 can be useful to learn some data into a stick-table. The function considers
15453 any comma as a delimiter for distinct values. If this is not desired, the
15454 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015455
Willy Tarreau74ca5042013-06-11 23:12:07 +020015456 ACL derivatives :
15457 shdr([<name>[,<occ>]]) : exact string match
15458 shdr_beg([<name>[,<occ>]]) : prefix match
15459 shdr_dir([<name>[,<occ>]]) : subdir match
15460 shdr_dom([<name>[,<occ>]]) : domain match
15461 shdr_end([<name>[,<occ>]]) : suffix match
15462 shdr_len([<name>[,<occ>]]) : length match
15463 shdr_reg([<name>[,<occ>]]) : regex match
15464 shdr_sub([<name>[,<occ>]]) : substring match
15465
15466res.hdr_cnt([<name>]) : integer
15467shdr_cnt([<name>]) : integer (deprecated)
15468 Returns an integer value representing the number of occurrences of response
15469 header field name <name>, or the total number of header fields if <name> is
15470 not specified. The function considers any comma as a delimiter for distinct
15471 values. If this is not desired, the res.fhdr_cnt() fetch should be used
15472 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015473
Willy Tarreau74ca5042013-06-11 23:12:07 +020015474res.hdr_ip([<name>[,<occ>]]) : ip
15475shdr_ip([<name>[,<occ>]]) : ip (deprecated)
15476 This extracts the last occurrence of header <name> in an HTTP response,
15477 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
15478 specific occurrence might be specified as a position number. Positive values
15479 indicate a position from the first occurrence, with 1 being the first one.
15480 Negative values indicate positions relative to the last one, with -1 being
15481 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015482
Willy Tarreaueb27ec72015-02-20 13:55:29 +010015483res.hdr_names([<delim>]) : string
15484 This builds a string made from the concatenation of all header names as they
15485 appear in the response when the rule is evaluated. The default delimiter is
15486 the comma (',') but it may be overridden as an optional argument <delim>. In
15487 this case, only the first character of <delim> is considered.
15488
Willy Tarreau74ca5042013-06-11 23:12:07 +020015489res.hdr_val([<name>[,<occ>]]) : integer
15490shdr_val([<name>[,<occ>]]) : integer (deprecated)
15491 This extracts the last occurrence of header <name> in an HTTP response, and
15492 converts it to an integer value. Optionally, a specific occurrence might be
15493 specified as a position number. Positive values indicate a position from the
15494 first occurrence, with 1 being the first one. Negative values indicate
15495 positions relative to the last one, with -1 being the last one. This can be
15496 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010015497
Willy Tarreau74ca5042013-06-11 23:12:07 +020015498res.ver : string
15499resp_ver : string (deprecated)
15500 Returns the version string from the HTTP response, for example "1.1". This
15501 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020015502
Willy Tarreau74ca5042013-06-11 23:12:07 +020015503 ACL derivatives :
15504 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010015505
Willy Tarreau74ca5042013-06-11 23:12:07 +020015506set-cookie([<name>]) : string (deprecated)
15507 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
15508 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020015509 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020015510 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010015511
Willy Tarreau74ca5042013-06-11 23:12:07 +020015512 This fetch function is deprecated and has been superseded by the "res.cook"
15513 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010015514
Willy Tarreau74ca5042013-06-11 23:12:07 +020015515status : integer
15516 Returns an integer containing the HTTP status code in the HTTP response, for
15517 example, 302. It is mostly used within ACLs and integer ranges, for example,
15518 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015519
Thierry Fournier0e00dca2016-04-07 15:47:40 +020015520unique-id : string
15521 Returns the unique-id attached to the request. The directive
15522 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
15523 fails. Note that the unique-id is usually used with HTTP requests, however this
15524 sample fetch can be used with other protocols. Obviously, if it is used with
15525 other protocols than HTTP, the unique-id-format directive must not contain
15526 HTTP parts. See: unique-id-format and unique-id-header
15527
Willy Tarreau74ca5042013-06-11 23:12:07 +020015528url : string
15529 This extracts the request's URL as presented in the request. A typical use is
15530 with prefetch-capable caches, and with portals which need to aggregate
15531 multiple information from databases and keep them in caches. With ACLs, using
15532 "path" is preferred over using "url", because clients may send a full URL as
15533 is normally done with proxies. The only real use is to match "*" which does
15534 not match in "path", and for which there is already a predefined ACL. See
15535 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015536
Willy Tarreau74ca5042013-06-11 23:12:07 +020015537 ACL derivatives :
15538 url : exact string match
15539 url_beg : prefix match
15540 url_dir : subdir match
15541 url_dom : domain match
15542 url_end : suffix match
15543 url_len : length match
15544 url_reg : regex match
15545 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015546
Willy Tarreau74ca5042013-06-11 23:12:07 +020015547url_ip : ip
15548 This extracts the IP address from the request's URL when the host part is
15549 presented as an IP address. Its use is very limited. For instance, a
15550 monitoring system might use this field as an alternative for the source IP in
15551 order to test what path a given source address would follow, or to force an
15552 entry in a table for a given source address. With ACLs it can be used to
15553 restrict access to certain systems through a proxy, for example when combined
15554 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015555
Willy Tarreau74ca5042013-06-11 23:12:07 +020015556url_port : integer
15557 This extracts the port part from the request's URL. Note that if the port is
15558 not specified in the request, port 80 is assumed. With ACLs it can be used to
15559 restrict access to certain systems through a proxy, for example when combined
15560 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015561
Willy Tarreau1ede1da2015-05-07 16:06:18 +020015562urlp([<name>[,<delim>]]) : string
15563url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020015564 This extracts the first occurrence of the parameter <name> in the query
15565 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020015566 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
15567 any parameter will match, and the first one will be returned. The result is
15568 a string corresponding to the value of the parameter <name> as presented in
15569 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020015570 stickiness based on a client ID, to extract an application cookie passed as a
15571 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020015572 this fetch iterates over multiple parameters and will iteratively report all
15573 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015574
Willy Tarreau74ca5042013-06-11 23:12:07 +020015575 ACL derivatives :
15576 urlp(<name>[,<delim>]) : exact string match
15577 urlp_beg(<name>[,<delim>]) : prefix match
15578 urlp_dir(<name>[,<delim>]) : subdir match
15579 urlp_dom(<name>[,<delim>]) : domain match
15580 urlp_end(<name>[,<delim>]) : suffix match
15581 urlp_len(<name>[,<delim>]) : length match
15582 urlp_reg(<name>[,<delim>]) : regex match
15583 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015584
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015585
Willy Tarreau74ca5042013-06-11 23:12:07 +020015586 Example :
15587 # match http://example.com/foo?PHPSESSIONID=some_id
15588 stick on urlp(PHPSESSIONID)
15589 # match http://example.com/foo;JSESSIONID=some_id
15590 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020015591
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015592urlp_val([<name>[,<delim>]]) : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020015593 See "urlp" above. This one extracts the URL parameter <name> in the request
15594 and converts it to an integer value. This can be used for session stickiness
15595 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020015596
Dragan Dosen0070cd52016-06-16 12:19:49 +020015597url32 : integer
15598 This returns a 32-bit hash of the value obtained by concatenating the first
15599 Host header and the whole URL including parameters (not only the path part of
15600 the request, as in the "base32" fetch above). This is useful to track per-URL
15601 activity. A shorter hash is stored, saving a lot of memory. The output type
15602 is an unsigned integer.
15603
15604url32+src : binary
15605 This returns the concatenation of the "url32" fetch and the "src" fetch. The
15606 resulting type is of type binary, with a size of 8 or 20 bytes depending on
15607 the source address family. This can be used to track per-IP, per-URL counters.
15608
Willy Tarreau198a7442008-01-17 12:05:32 +010015609
Willy Tarreau74ca5042013-06-11 23:12:07 +0200156107.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015611---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010015612
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015613Some predefined ACLs are hard-coded so that they do not have to be declared in
15614every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020015615order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010015616
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015617ACL name Equivalent to Usage
15618---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015619FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020015620HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015621HTTP_1.0 req_ver 1.0 match HTTP version 1.0
15622HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010015623HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
15624HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
15625HTTP_URL_SLASH url_beg / match URL beginning with "/"
15626HTTP_URL_STAR url * match URL equal to "*"
15627LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015628METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020015629METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015630METH_GET method GET HEAD match HTTP GET or HEAD method
15631METH_HEAD method HEAD match HTTP HEAD method
15632METH_OPTIONS method OPTIONS match HTTP OPTIONS method
15633METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020015634METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015635METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020015636RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015637REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010015638TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015639WAIT_END wait_end wait for end of content analysis
15640---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010015641
Willy Tarreaub937b7e2010-01-12 15:27:54 +010015642
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200156438. Logging
15644----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010015645
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015646One of HAProxy's strong points certainly lies is its precise logs. It probably
15647provides the finest level of information available for such a product, which is
15648very important for troubleshooting complex environments. Standard information
15649provided in logs include client ports, TCP/HTTP state timers, precise session
15650state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015651to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015652headers.
15653
15654In order to improve administrators reactivity, it offers a great transparency
15655about encountered problems, both internal and external, and it is possible to
15656send logs to different sources at the same time with different level filters :
15657
15658 - global process-level logs (system errors, start/stop, etc..)
15659 - per-instance system and internal errors (lack of resource, bugs, ...)
15660 - per-instance external troubles (servers up/down, max connections)
15661 - per-instance activity (client connections), either at the establishment or
15662 at the termination.
Davor Ocelice9ed2812017-12-25 17:49:28 +010015663 - per-request control of log-level, e.g.
Jim Freeman9e8714b2015-05-26 09:16:34 -060015664 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015665
15666The ability to distribute different levels of logs to different log servers
15667allow several production teams to interact and to fix their problems as soon
15668as possible. For example, the system team might monitor system-wide errors,
15669while the application team might be monitoring the up/down for their servers in
15670real time, and the security team might analyze the activity logs with one hour
15671delay.
15672
15673
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200156748.1. Log levels
15675---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015676
Simon Hormandf791f52011-05-29 15:01:10 +090015677TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015678source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090015679HTTP request, HTTP return code, number of bytes transmitted, conditions
15680in which the session ended, and even exchanged cookies values. For example
15681track a particular user's problems. All messages may be sent to up to two
15682syslog servers. Check the "log" keyword in section 4.2 for more information
15683about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015684
15685
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200156868.2. Log formats
15687----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015688
William Lallemand48940402012-01-30 16:47:22 +010015689HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090015690and will be detailed in the following sections. A few of them may vary
15691slightly with the configuration, due to indicators specific to certain
15692options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015693
15694 - the default format, which is very basic and very rarely used. It only
15695 provides very basic information about the incoming connection at the moment
15696 it is accepted : source IP:port, destination IP:port, and frontend-name.
15697 This mode will eventually disappear so it will not be described to great
15698 extents.
15699
15700 - the TCP format, which is more advanced. This format is enabled when "option
15701 tcplog" is set on the frontend. HAProxy will then usually wait for the
15702 connection to terminate before logging. This format provides much richer
15703 information, such as timers, connection counts, queue size, etc... This
15704 format is recommended for pure TCP proxies.
15705
15706 - the HTTP format, which is the most advanced for HTTP proxying. This format
15707 is enabled when "option httplog" is set on the frontend. It provides the
15708 same information as the TCP format with some HTTP-specific fields such as
15709 the request, the status code, and captures of headers and cookies. This
15710 format is recommended for HTTP proxies.
15711
Emeric Brun3a058f32009-06-30 18:26:00 +020015712 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
15713 fields arranged in the same order as the CLF format. In this mode, all
15714 timers, captures, flags, etc... appear one per field after the end of the
15715 common fields, in the same order they appear in the standard HTTP format.
15716
William Lallemand48940402012-01-30 16:47:22 +010015717 - the custom log format, allows you to make your own log line.
15718
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015719Next sections will go deeper into details for each of these formats. Format
15720specification will be performed on a "field" basis. Unless stated otherwise, a
15721field is a portion of text delimited by any number of spaces. Since syslog
15722servers are susceptible of inserting fields at the beginning of a line, it is
15723always assumed that the first field is the one containing the process name and
15724identifier.
15725
15726Note : Since log lines may be quite long, the log examples in sections below
15727 might be broken into multiple lines. The example log lines will be
15728 prefixed with 3 closing angle brackets ('>>>') and each time a log is
15729 broken into multiple lines, each non-final line will end with a
15730 backslash ('\') and the next line will start indented by two characters.
15731
15732
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200157338.2.1. Default log format
15734-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015735
15736This format is used when no specific option is set. The log is emitted as soon
15737as the connection is accepted. One should note that this currently is the only
15738format which logs the request's destination IP and ports.
15739
15740 Example :
15741 listen www
15742 mode http
15743 log global
15744 server srv1 127.0.0.1:8000
15745
15746 >>> Feb 6 12:12:09 localhost \
15747 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
15748 (www/HTTP)
15749
15750 Field Format Extract from the example above
15751 1 process_name '[' pid ']:' haproxy[14385]:
15752 2 'Connect from' Connect from
15753 3 source_ip ':' source_port 10.0.1.2:33312
15754 4 'to' to
15755 5 destination_ip ':' destination_port 10.0.3.31:8012
15756 6 '(' frontend_name '/' mode ')' (www/HTTP)
15757
15758Detailed fields description :
15759 - "source_ip" is the IP address of the client which initiated the connection.
15760 - "source_port" is the TCP port of the client which initiated the connection.
15761 - "destination_ip" is the IP address the client connected to.
15762 - "destination_port" is the TCP port the client connected to.
15763 - "frontend_name" is the name of the frontend (or listener) which received
15764 and processed the connection.
15765 - "mode is the mode the frontend is operating (TCP or HTTP).
15766
Willy Tarreauceb24bc2010-11-09 12:46:41 +010015767In case of a UNIX socket, the source and destination addresses are marked as
15768"unix:" and the ports reflect the internal ID of the socket which accepted the
15769connection (the same ID as reported in the stats).
15770
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015771It is advised not to use this deprecated format for newer installations as it
15772will eventually disappear.
15773
15774
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200157758.2.2. TCP log format
15776---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015777
15778The TCP format is used when "option tcplog" is specified in the frontend, and
15779is the recommended format for pure TCP proxies. It provides a lot of precious
15780information for troubleshooting. Since this format includes timers and byte
15781counts, the log is normally emitted at the end of the session. It can be
15782emitted earlier if "option logasap" is specified, which makes sense in most
15783environments with long sessions such as remote terminals. Sessions which match
15784the "monitor" rules are never logged. It is also possible not to emit logs for
15785sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020015786specifying "option dontlognull" in the frontend. Successful connections will
15787not be logged if "option dontlog-normal" is specified in the frontend. A few
15788fields may slightly vary depending on some configuration options, those are
15789marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015790
15791 Example :
15792 frontend fnt
15793 mode tcp
15794 option tcplog
15795 log global
15796 default_backend bck
15797
15798 backend bck
15799 server srv1 127.0.0.1:8000
15800
15801 >>> Feb 6 12:12:56 localhost \
15802 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
15803 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
15804
15805 Field Format Extract from the example above
15806 1 process_name '[' pid ']:' haproxy[14387]:
15807 2 client_ip ':' client_port 10.0.1.2:33313
15808 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
15809 4 frontend_name fnt
15810 5 backend_name '/' server_name bck/srv1
15811 6 Tw '/' Tc '/' Tt* 0/0/5007
15812 7 bytes_read* 212
15813 8 termination_state --
15814 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
15815 10 srv_queue '/' backend_queue 0/0
15816
15817Detailed fields description :
15818 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010015819 connection to haproxy. If the connection was accepted on a UNIX socket
15820 instead, the IP address would be replaced with the word "unix". Note that
15821 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010015822 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010015823 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010015824 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015825
15826 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010015827 If the connection was accepted on a UNIX socket instead, the port would be
15828 replaced with the ID of the accepting socket, which is also reported in the
15829 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015830
15831 - "accept_date" is the exact date when the connection was received by haproxy
15832 (which might be very slightly different from the date observed on the
15833 network if there was some queuing in the system's backlog). This is usually
15834 the same date which may appear in any upstream firewall's log.
15835
15836 - "frontend_name" is the name of the frontend (or listener) which received
15837 and processed the connection.
15838
15839 - "backend_name" is the name of the backend (or listener) which was selected
15840 to manage the connection to the server. This will be the same as the
15841 frontend if no switching rule has been applied, which is common for TCP
15842 applications.
15843
15844 - "server_name" is the name of the last server to which the connection was
15845 sent, which might differ from the first one if there were connection errors
15846 and a redispatch occurred. Note that this server belongs to the backend
15847 which processed the request. If the connection was aborted before reaching
15848 a server, "<NOSRV>" is indicated instead of a server name.
15849
15850 - "Tw" is the total time in milliseconds spent waiting in the various queues.
15851 It can be "-1" if the connection was aborted before reaching the queue.
15852 See "Timers" below for more details.
15853
15854 - "Tc" is the total time in milliseconds spent waiting for the connection to
15855 establish to the final server, including retries. It can be "-1" if the
15856 connection was aborted before a connection could be established. See
15857 "Timers" below for more details.
15858
15859 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015860 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015861 "option logasap" was specified, then the time counting stops at the moment
15862 the log is emitted. In this case, a '+' sign is prepended before the value,
15863 indicating that the final one will be larger. See "Timers" below for more
15864 details.
15865
15866 - "bytes_read" is the total number of bytes transmitted from the server to
15867 the client when the log is emitted. If "option logasap" is specified, the
15868 this value will be prefixed with a '+' sign indicating that the final one
15869 may be larger. Please note that this value is a 64-bit counter, so log
15870 analysis tools must be able to handle it without overflowing.
15871
15872 - "termination_state" is the condition the session was in when the session
15873 ended. This indicates the session state, which side caused the end of
15874 session to happen, and for what reason (timeout, error, ...). The normal
15875 flags should be "--", indicating the session was closed by either end with
15876 no data remaining in buffers. See below "Session state at disconnection"
15877 for more details.
15878
15879 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015880 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015881 limits have been reached. For instance, if actconn is close to 512 when
15882 multiple connection errors occur, chances are high that the system limits
15883 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015884 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015885
15886 - "feconn" is the total number of concurrent connections on the frontend when
15887 the session was logged. It is useful to estimate the amount of resource
15888 required to sustain high loads, and to detect when the frontend's "maxconn"
15889 has been reached. Most often when this value increases by huge jumps, it is
15890 because there is congestion on the backend servers, but sometimes it can be
15891 caused by a denial of service attack.
15892
15893 - "beconn" is the total number of concurrent connections handled by the
15894 backend when the session was logged. It includes the total number of
15895 concurrent connections active on servers as well as the number of
15896 connections pending in queues. It is useful to estimate the amount of
15897 additional servers needed to support high loads for a given application.
15898 Most often when this value increases by huge jumps, it is because there is
15899 congestion on the backend servers, but sometimes it can be caused by a
15900 denial of service attack.
15901
15902 - "srv_conn" is the total number of concurrent connections still active on
15903 the server when the session was logged. It can never exceed the server's
15904 configured "maxconn" parameter. If this value is very often close or equal
15905 to the server's "maxconn", it means that traffic regulation is involved a
15906 lot, meaning that either the server's maxconn value is too low, or that
15907 there aren't enough servers to process the load with an optimal response
15908 time. When only one of the server's "srv_conn" is high, it usually means
15909 that this server has some trouble causing the connections to take longer to
15910 be processed than on other servers.
15911
15912 - "retries" is the number of connection retries experienced by this session
15913 when trying to connect to the server. It must normally be zero, unless a
15914 server is being stopped at the same moment the connection was attempted.
15915 Frequent retries generally indicate either a network problem between
15916 haproxy and the server, or a misconfigured system backlog on the server
15917 preventing new connections from being queued. This field may optionally be
15918 prefixed with a '+' sign, indicating that the session has experienced a
15919 redispatch after the maximal retry count has been reached on the initial
15920 server. In this case, the server name appearing in the log is the one the
15921 connection was redispatched to, and not the first one, though both may
15922 sometimes be the same in case of hashing for instance. So as a general rule
15923 of thumb, when a '+' is present in front of the retry count, this count
15924 should not be attributed to the logged server.
15925
15926 - "srv_queue" is the total number of requests which were processed before
15927 this one in the server queue. It is zero when the request has not gone
15928 through the server queue. It makes it possible to estimate the approximate
15929 server's response time by dividing the time spent in queue by the number of
15930 requests in the queue. It is worth noting that if a session experiences a
15931 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010015932 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015933 backend queue unless a redispatch occurs.
15934
15935 - "backend_queue" is the total number of requests which were processed before
15936 this one in the backend's global queue. It is zero when the request has not
15937 gone through the global queue. It makes it possible to estimate the average
15938 queue length, which easily translates into a number of missing servers when
15939 divided by a server's "maxconn" parameter. It is worth noting that if a
15940 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010015941 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015942 through both the server queue and the backend queue unless a redispatch
15943 occurs.
15944
15945
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200159468.2.3. HTTP log format
15947----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015948
15949The HTTP format is the most complete and the best suited for HTTP proxies. It
15950is enabled by when "option httplog" is specified in the frontend. It provides
15951the same level of information as the TCP format with additional features which
15952are specific to the HTTP protocol. Just like the TCP format, the log is usually
15953emitted at the end of the session, unless "option logasap" is specified, which
15954generally only makes sense for download sites. A session which matches the
15955"monitor" rules will never logged. It is also possible not to log sessions for
15956which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020015957frontend. Successful connections will not be logged if "option dontlog-normal"
15958is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015959
15960Most fields are shared with the TCP log, some being different. A few fields may
15961slightly vary depending on some configuration options. Those ones are marked
15962with a star ('*') after the field name below.
15963
15964 Example :
15965 frontend http-in
15966 mode http
15967 option httplog
15968 log global
15969 default_backend bck
15970
15971 backend static
15972 server srv1 127.0.0.1:8000
15973
15974 >>> Feb 6 12:14:14 localhost \
15975 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
15976 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 +010015977 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015978
15979 Field Format Extract from the example above
15980 1 process_name '[' pid ']:' haproxy[14389]:
15981 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015982 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015983 4 frontend_name http-in
15984 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015985 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015986 7 status_code 200
15987 8 bytes_read* 2750
15988 9 captured_request_cookie -
15989 10 captured_response_cookie -
15990 11 termination_state ----
15991 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
15992 13 srv_queue '/' backend_queue 0/0
15993 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
15994 15 '{' captured_response_headers* '}' {}
15995 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010015996
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015997Detailed fields description :
15998 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010015999 connection to haproxy. If the connection was accepted on a UNIX socket
16000 instead, the IP address would be replaced with the word "unix". Note that
16001 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016002 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010016003 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016004 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016005
16006 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016007 If the connection was accepted on a UNIX socket instead, the port would be
16008 replaced with the ID of the accepting socket, which is also reported in the
16009 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016010
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016011 - "request_date" is the exact date when the first byte of the HTTP request
16012 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016013
16014 - "frontend_name" is the name of the frontend (or listener) which received
16015 and processed the connection.
16016
16017 - "backend_name" is the name of the backend (or listener) which was selected
16018 to manage the connection to the server. This will be the same as the
16019 frontend if no switching rule has been applied.
16020
16021 - "server_name" is the name of the last server to which the connection was
16022 sent, which might differ from the first one if there were connection errors
16023 and a redispatch occurred. Note that this server belongs to the backend
16024 which processed the request. If the request was aborted before reaching a
16025 server, "<NOSRV>" is indicated instead of a server name. If the request was
16026 intercepted by the stats subsystem, "<STATS>" is indicated instead.
16027
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016028 - "TR" is the total time in milliseconds spent waiting for a full HTTP
16029 request from the client (not counting body) after the first byte was
16030 received. It can be "-1" if the connection was aborted before a complete
16031 request could be received or the a bad request was received. It should
16032 always be very small because a request generally fits in one single packet.
16033 Large times here generally indicate network issues between the client and
16034 haproxy or requests being typed by hand. See "Timers" below for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016035
16036 - "Tw" is the total time in milliseconds spent waiting in the various queues.
16037 It can be "-1" if the connection was aborted before reaching the queue.
16038 See "Timers" below for more details.
16039
16040 - "Tc" is the total time in milliseconds spent waiting for the connection to
16041 establish to the final server, including retries. It can be "-1" if the
16042 request was aborted before a connection could be established. See "Timers"
16043 below for more details.
16044
16045 - "Tr" is the total time in milliseconds spent waiting for the server to send
16046 a full HTTP response, not counting data. It can be "-1" if the request was
16047 aborted before a complete response could be received. It generally matches
16048 the server's processing time for the request, though it may be altered by
16049 the amount of data sent by the client to the server. Large times here on
16050 "GET" requests generally indicate an overloaded server. See "Timers" below
16051 for more details.
16052
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016053 - "Ta" is the time the request remained active in haproxy, which is the total
16054 time in milliseconds elapsed between the first byte of the request was
16055 received and the last byte of response was sent. It covers all possible
16056 processing except the handshake (see Th) and idle time (see Ti). There is
16057 one exception, if "option logasap" was specified, then the time counting
16058 stops at the moment the log is emitted. In this case, a '+' sign is
16059 prepended before the value, indicating that the final one will be larger.
16060 See "Timers" below for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016061
16062 - "status_code" is the HTTP status code returned to the client. This status
16063 is generally set by the server, but it might also be set by haproxy when
16064 the server cannot be reached or when its response is blocked by haproxy.
16065
16066 - "bytes_read" is the total number of bytes transmitted to the client when
16067 the log is emitted. This does include HTTP headers. If "option logasap" is
16068 specified, the this value will be prefixed with a '+' sign indicating that
16069 the final one may be larger. Please note that this value is a 64-bit
16070 counter, so log analysis tools must be able to handle it without
16071 overflowing.
16072
16073 - "captured_request_cookie" is an optional "name=value" entry indicating that
16074 the client had this cookie in the request. The cookie name and its maximum
16075 length are defined by the "capture cookie" statement in the frontend
16076 configuration. The field is a single dash ('-') when the option is not
16077 set. Only one cookie may be captured, it is generally used to track session
16078 ID exchanges between a client and a server to detect session crossing
16079 between clients due to application bugs. For more details, please consult
16080 the section "Capturing HTTP headers and cookies" below.
16081
16082 - "captured_response_cookie" is an optional "name=value" entry indicating
16083 that the server has returned a cookie with its response. The cookie name
16084 and its maximum length are defined by the "capture cookie" statement in the
16085 frontend configuration. The field is a single dash ('-') when the option is
16086 not set. Only one cookie may be captured, it is generally used to track
16087 session ID exchanges between a client and a server to detect session
16088 crossing between clients due to application bugs. For more details, please
16089 consult the section "Capturing HTTP headers and cookies" below.
16090
16091 - "termination_state" is the condition the session was in when the session
16092 ended. This indicates the session state, which side caused the end of
16093 session to happen, for what reason (timeout, error, ...), just like in TCP
16094 logs, and information about persistence operations on cookies in the last
16095 two characters. The normal flags should begin with "--", indicating the
16096 session was closed by either end with no data remaining in buffers. See
16097 below "Session state at disconnection" for more details.
16098
16099 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016100 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016101 limits have been reached. For instance, if actconn is close to 512 or 1024
16102 when multiple connection errors occur, chances are high that the system
16103 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016104 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016105 system.
16106
16107 - "feconn" is the total number of concurrent connections on the frontend when
16108 the session was logged. It is useful to estimate the amount of resource
16109 required to sustain high loads, and to detect when the frontend's "maxconn"
16110 has been reached. Most often when this value increases by huge jumps, it is
16111 because there is congestion on the backend servers, but sometimes it can be
16112 caused by a denial of service attack.
16113
16114 - "beconn" is the total number of concurrent connections handled by the
16115 backend when the session was logged. It includes the total number of
16116 concurrent connections active on servers as well as the number of
16117 connections pending in queues. It is useful to estimate the amount of
16118 additional servers needed to support high loads for a given application.
16119 Most often when this value increases by huge jumps, it is because there is
16120 congestion on the backend servers, but sometimes it can be caused by a
16121 denial of service attack.
16122
16123 - "srv_conn" is the total number of concurrent connections still active on
16124 the server when the session was logged. It can never exceed the server's
16125 configured "maxconn" parameter. If this value is very often close or equal
16126 to the server's "maxconn", it means that traffic regulation is involved a
16127 lot, meaning that either the server's maxconn value is too low, or that
16128 there aren't enough servers to process the load with an optimal response
16129 time. When only one of the server's "srv_conn" is high, it usually means
16130 that this server has some trouble causing the requests to take longer to be
16131 processed than on other servers.
16132
16133 - "retries" is the number of connection retries experienced by this session
16134 when trying to connect to the server. It must normally be zero, unless a
16135 server is being stopped at the same moment the connection was attempted.
16136 Frequent retries generally indicate either a network problem between
16137 haproxy and the server, or a misconfigured system backlog on the server
16138 preventing new connections from being queued. This field may optionally be
16139 prefixed with a '+' sign, indicating that the session has experienced a
16140 redispatch after the maximal retry count has been reached on the initial
16141 server. In this case, the server name appearing in the log is the one the
16142 connection was redispatched to, and not the first one, though both may
16143 sometimes be the same in case of hashing for instance. So as a general rule
16144 of thumb, when a '+' is present in front of the retry count, this count
16145 should not be attributed to the logged server.
16146
16147 - "srv_queue" is the total number of requests which were processed before
16148 this one in the server queue. It is zero when the request has not gone
16149 through the server queue. It makes it possible to estimate the approximate
16150 server's response time by dividing the time spent in queue by the number of
16151 requests in the queue. It is worth noting that if a session experiences a
16152 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010016153 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016154 backend queue unless a redispatch occurs.
16155
16156 - "backend_queue" is the total number of requests which were processed before
16157 this one in the backend's global queue. It is zero when the request has not
16158 gone through the global queue. It makes it possible to estimate the average
16159 queue length, which easily translates into a number of missing servers when
16160 divided by a server's "maxconn" parameter. It is worth noting that if a
16161 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010016162 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016163 through both the server queue and the backend queue unless a redispatch
16164 occurs.
16165
16166 - "captured_request_headers" is a list of headers captured in the request due
16167 to the presence of the "capture request header" statement in the frontend.
16168 Multiple headers can be captured, they will be delimited by a vertical bar
16169 ('|'). When no capture is enabled, the braces do not appear, causing a
16170 shift of remaining fields. It is important to note that this field may
16171 contain spaces, and that using it requires a smarter log parser than when
16172 it's not used. Please consult the section "Capturing HTTP headers and
16173 cookies" below for more details.
16174
16175 - "captured_response_headers" is a list of headers captured in the response
16176 due to the presence of the "capture response header" statement in the
16177 frontend. Multiple headers can be captured, they will be delimited by a
16178 vertical bar ('|'). When no capture is enabled, the braces do not appear,
16179 causing a shift of remaining fields. It is important to note that this
16180 field may contain spaces, and that using it requires a smarter log parser
16181 than when it's not used. Please consult the section "Capturing HTTP headers
16182 and cookies" below for more details.
16183
16184 - "http_request" is the complete HTTP request line, including the method,
16185 request and HTTP version string. Non-printable characters are encoded (see
16186 below the section "Non-printable characters"). This is always the last
16187 field, and it is always delimited by quotes and is the only one which can
16188 contain quotes. If new fields are added to the log format, they will be
16189 added before this field. This field might be truncated if the request is
16190 huge and does not fit in the standard syslog buffer (1024 characters). This
16191 is the reason why this field must always remain the last one.
16192
16193
Cyril Bontédc4d9032012-04-08 21:57:39 +0200161948.2.4. Custom log format
16195------------------------
William Lallemand48940402012-01-30 16:47:22 +010016196
Willy Tarreau2beef582012-12-20 17:22:52 +010016197The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010016198mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010016199
Davor Ocelice9ed2812017-12-25 17:49:28 +010016200HAProxy understands some log format variables. % precedes log format variables.
William Lallemand48940402012-01-30 16:47:22 +010016201Variables can take arguments using braces ('{}'), and multiple arguments are
16202separated by commas within the braces. Flags may be added or removed by
16203prefixing them with a '+' or '-' sign.
16204
16205Special variable "%o" may be used to propagate its flags to all other
16206variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010016207("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010016208
Willy Tarreauc8368452012-12-21 00:09:23 +010016209If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020016210as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010016211less common information such as the client's SSL certificate's DN, or to log
16212the key that would be used to store an entry into a stick table.
16213
William Lallemand48940402012-01-30 16:47:22 +010016214Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016215In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010016216in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010016217
Dragan Dosen835b9212016-02-12 13:23:03 +010016218Note: when using the RFC5424 syslog message format, the characters '"',
16219'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
16220https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
16221such cases, the use of the flag "E" should be considered.
16222
William Lallemand48940402012-01-30 16:47:22 +010016223Flags are :
16224 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016225 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010016226 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
16227 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010016228
16229 Example:
16230
16231 log-format %T\ %t\ Some\ Text
16232 log-format %{+Q}o\ %t\ %s\ %{-Q}r
16233
Dragan Dosen835b9212016-02-12 13:23:03 +010016234 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
16235
William Lallemand48940402012-01-30 16:47:22 +010016236At the moment, the default HTTP format is defined this way :
16237
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016238 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
16239 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010016240
William Lallemandbddd4fd2012-02-27 11:23:10 +010016241the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010016242
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016243 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
16244 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
16245 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010016246
William Lallemandbddd4fd2012-02-27 11:23:10 +010016247and the default TCP format is defined this way :
16248
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016249 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
16250 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010016251
William Lallemand48940402012-01-30 16:47:22 +010016252Please refer to the table below for currently defined variables :
16253
William Lallemandbddd4fd2012-02-27 11:23:10 +010016254 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020016255 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016256 +---+------+-----------------------------------------------+-------------+
16257 | | %o | special variable, apply flags on all next var | |
16258 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010016259 | | %B | bytes_read (from server to client) | numeric |
16260 | H | %CC | captured_request_cookie | string |
16261 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020016262 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000016263 | H | %HM | HTTP method (ex: POST) | string |
16264 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000016265 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000016266 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
16267 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010016268 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020016269 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020016270 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016271 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016272 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020016273 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080016274 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016275 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
16276 | H | %Ti | idle time before the HTTP request | numeric |
16277 | H | %Tq | Th + Ti + TR | numeric |
16278 | H | %TR | time to receive the full request from 1st byte| numeric |
16279 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020016280 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016281 | | %Tt | Tt | numeric |
16282 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010016283 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016284 | | %ac | actconn | numeric |
16285 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010016286 | | %bc | beconn (backend concurrent connections) | numeric |
16287 | | %bi | backend_source_ip (connecting address) | IP |
16288 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016289 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010016290 | | %ci | client_ip (accepted address) | IP |
16291 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016292 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010016293 | | %fc | feconn (frontend concurrent connections) | numeric |
16294 | | %fi | frontend_ip (accepting address) | IP |
16295 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020016296 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020016297 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020016298 | | %hr | captured_request_headers default style | string |
16299 | | %hrl | captured_request_headers CLF style | string list |
16300 | | %hs | captured_response_headers default style | string |
16301 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020016302 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020016303 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020016304 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016305 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010016306 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016307 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010016308 | | %sc | srv_conn (server concurrent connections) | numeric |
16309 | | %si | server_IP (target address) | IP |
16310 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016311 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020016312 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
16313 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010016314 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016315 | H | %tr | date_time of HTTP request | date |
16316 | H | %trg | gmt_date_time of start of HTTP request | date |
16317 | H | %trl | locla_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016318 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020016319 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016320 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010016321
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020016322 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010016323
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010016324
163258.2.5. Error log format
16326-----------------------
16327
16328When an incoming connection fails due to an SSL handshake or an invalid PROXY
16329protocol header, haproxy will log the event using a shorter, fixed line format.
16330By default, logs are emitted at the LOG_INFO level, unless the option
16331"log-separate-errors" is set in the backend, in which case the LOG_ERR level
Davor Ocelice9ed2812017-12-25 17:49:28 +010016332will be used. Connections on which no data are exchanged (e.g. probes) are not
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010016333logged if the "dontlognull" option is set.
16334
16335The format looks like this :
16336
16337 >>> Dec 3 18:27:14 localhost \
16338 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
16339 Connection error during SSL handshake
16340
16341 Field Format Extract from the example above
16342 1 process_name '[' pid ']:' haproxy[6103]:
16343 2 client_ip ':' client_port 127.0.0.1:56059
16344 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
16345 4 frontend_name "/" bind_name ":" frt/f1:
16346 5 message Connection error during SSL handshake
16347
16348These fields just provide minimal information to help debugging connection
16349failures.
16350
16351
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200163528.3. Advanced logging options
16353-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016354
16355Some advanced logging options are often looked for but are not easy to find out
16356just by looking at the various options. Here is an entry point for the few
16357options which can enable better logging. Please refer to the keywords reference
16358for more information about their usage.
16359
16360
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200163618.3.1. Disabling logging of external tests
16362------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016363
16364It is quite common to have some monitoring tools perform health checks on
16365haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
16366commercial load-balancer, and sometimes it will simply be a more complete
16367monitoring system such as Nagios. When the tests are very frequent, users often
16368ask how to disable logging for those checks. There are three possibilities :
16369
16370 - if connections come from everywhere and are just TCP probes, it is often
16371 desired to simply disable logging of connections without data exchange, by
16372 setting "option dontlognull" in the frontend. It also disables logging of
16373 port scans, which may or may not be desired.
16374
16375 - if the connection come from a known source network, use "monitor-net" to
16376 declare this network as monitoring only. Any host in this network will then
16377 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016378 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016379 such as other load-balancers.
16380
16381 - if the tests are performed on a known URI, use "monitor-uri" to declare
16382 this URI as dedicated to monitoring. Any host sending this request will
16383 only get the result of a health-check, and the request will not be logged.
16384
16385
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200163868.3.2. Logging before waiting for the session to terminate
16387----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016388
16389The problem with logging at end of connection is that you have no clue about
16390what is happening during very long sessions, such as remote terminal sessions
16391or large file downloads. This problem can be worked around by specifying
Davor Ocelice9ed2812017-12-25 17:49:28 +010016392"option logasap" in the frontend. HAProxy will then log as soon as possible,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016393just before data transfer begins. This means that in case of TCP, it will still
16394log the connection status to the server, and in case of HTTP, it will log just
16395after processing the server headers. In this case, the number of bytes reported
16396is the number of header bytes sent to the client. In order to avoid confusion
16397with normal logs, the total time field and the number of bytes are prefixed
16398with a '+' sign which means that real numbers are certainly larger.
16399
16400
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200164018.3.3. Raising log level upon errors
16402------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020016403
16404Sometimes it is more convenient to separate normal traffic from errors logs,
16405for instance in order to ease error monitoring from log files. When the option
16406"log-separate-errors" is used, connections which experience errors, timeouts,
16407retries, redispatches or HTTP status codes 5xx will see their syslog level
16408raised from "info" to "err". This will help a syslog daemon store the log in
16409a separate file. It is very important to keep the errors in the normal traffic
16410file too, so that log ordering is not altered. You should also be careful if
16411you already have configured your syslog daemon to store all logs higher than
16412"notice" in an "admin" file, because the "err" level is higher than "notice".
16413
16414
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200164158.3.4. Disabling logging of successful connections
16416--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020016417
16418Although this may sound strange at first, some large sites have to deal with
16419multiple thousands of logs per second and are experiencing difficulties keeping
16420them intact for a long time or detecting errors within them. If the option
16421"dontlog-normal" is set on the frontend, all normal connections will not be
16422logged. In this regard, a normal connection is defined as one without any
16423error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
16424and a response with a status 5xx is not considered normal and will be logged
16425too. Of course, doing is is really discouraged as it will remove most of the
16426useful information from the logs. Do this only if you have no other
16427alternative.
16428
16429
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200164308.4. Timing events
16431------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016432
16433Timers provide a great help in troubleshooting network problems. All values are
16434reported in milliseconds (ms). These timers should be used in conjunction with
16435the session termination flags. In TCP mode with "option tcplog" set on the
16436frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016437mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
16438addition, three other measures are provided, "Th", "Ti", and "Tq".
16439
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010016440Timings events in HTTP mode:
16441
16442 first request 2nd request
16443 |<-------------------------------->|<-------------- ...
16444 t tr t tr ...
16445 ---|----|----|----|----|----|----|----|----|--
16446 : Th Ti TR Tw Tc Tr Td : Ti ...
16447 :<---- Tq ---->: :
16448 :<-------------- Tt -------------->:
16449 :<--------- Ta --------->:
16450
16451Timings events in TCP mode:
16452
16453 TCP session
16454 |<----------------->|
16455 t t
16456 ---|----|----|----|----|---
16457 | Th Tw Tc Td |
16458 |<------ Tt ------->|
16459
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016460 - Th: total time to accept tcp connection and execute handshakes for low level
Davor Ocelice9ed2812017-12-25 17:49:28 +010016461 protocols. Currently, these protocols are proxy-protocol and SSL. This may
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016462 only happen once during the whole connection's lifetime. A large time here
16463 may indicate that the client only pre-established the connection without
16464 speaking, that it is experiencing network issues preventing it from
Davor Ocelice9ed2812017-12-25 17:49:28 +010016465 completing a handshake in a reasonable time (e.g. MTU issues), or that an
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016466 SSL handshake was very expensive to compute.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016467
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016468 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
16469 counts between the end of the handshakes and the first byte of the HTTP
16470 request. When dealing with a second request in keep-alive mode, it starts
16471 to count after the end of the transmission the previous response. Some
16472 browsers pre-establish connections to a server in order to reduce the
16473 latency of a future request, and keep them pending until they need it. This
16474 delay will be reported as the idle time. A value of -1 indicates that
16475 nothing was received on the connection.
16476
16477 - TR: total time to get the client request (HTTP mode only). It's the time
16478 elapsed between the first bytes received and the moment the proxy received
16479 the empty line marking the end of the HTTP headers. The value "-1"
16480 indicates that the end of headers has never been seen. This happens when
16481 the client closes prematurely or times out. This time is usually very short
16482 since most requests fit in a single packet. A large time may indicate a
16483 request typed by hand during a test.
16484
16485 - Tq: total time to get the client request from the accept date or since the
16486 emission of the last byte of the previous response (HTTP mode only). It's
Davor Ocelice9ed2812017-12-25 17:49:28 +010016487 exactly equal to Th + Ti + TR unless any of them is -1, in which case it
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016488 returns -1 as well. This timer used to be very useful before the arrival of
16489 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
16490 it in favor of TR nowadays, as the idle time adds a lot of noise to the
16491 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016492
16493 - Tw: total time spent in the queues waiting for a connection slot. It
16494 accounts for backend queue as well as the server queues, and depends on the
16495 queue size, and the time needed for the server to complete previous
16496 requests. The value "-1" means that the request was killed before reaching
16497 the queue, which is generally what happens with invalid or denied requests.
16498
16499 - Tc: total time to establish the TCP connection to the server. It's the time
16500 elapsed between the moment the proxy sent the connection request, and the
16501 moment it was acknowledged by the server, or between the TCP SYN packet and
16502 the matching SYN/ACK packet in return. The value "-1" means that the
16503 connection never established.
16504
16505 - Tr: server response time (HTTP mode only). It's the time elapsed between
16506 the moment the TCP connection was established to the server and the moment
16507 the server sent its complete response headers. It purely shows its request
16508 processing time, without the network overhead due to the data transmission.
16509 It is worth noting that when the client has data to send to the server, for
16510 instance during a POST request, the time already runs, and this can distort
16511 apparent response time. For this reason, it's generally wise not to trust
16512 too much this field for POST requests initiated from clients behind an
16513 untrusted network. A value of "-1" here means that the last the response
16514 header (empty line) was never seen, most likely because the server timeout
16515 stroke before the server managed to process the request.
16516
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016517 - Ta: total active time for the HTTP request, between the moment the proxy
16518 received the first byte of the request header and the emission of the last
16519 byte of the response body. The exception is when the "logasap" option is
16520 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
16521 a '+' sign. From this field, we can deduce "Td", the data transmission time,
16522 by subtracting other timers when valid :
16523
16524 Td = Ta - (TR + Tw + Tc + Tr)
16525
16526 Timers with "-1" values have to be excluded from this equation. Note that
16527 "Ta" can never be negative.
16528
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016529 - Tt: total session duration time, between the moment the proxy accepted it
16530 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016531 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
16532 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016533 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016534
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016535 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016536
16537 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016538 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
16539 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016540
16541These timers provide precious indications on trouble causes. Since the TCP
16542protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
16543that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016544due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
16545"Tt" is close to a timeout value specified in the configuration, it often means
16546that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016547
16548Most common cases :
16549
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016550 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
16551 the client and the proxy. This is very rare on local networks but might
16552 happen when clients are on far remote networks and send large requests. It
16553 may happen that values larger than usual appear here without any network
16554 cause. Sometimes, during an attack or just after a resource starvation has
16555 ended, haproxy may accept thousands of connections in a few milliseconds.
16556 The time spent accepting these connections will inevitably slightly delay
16557 processing of other connections, and it can happen that request times in the
16558 order of a few tens of milliseconds are measured after a few thousands of
16559 new connections have been accepted at once. Using one of the keep-alive
16560 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020016561 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016562
16563 - If "Tc" is close to 3000, a packet has probably been lost between the
16564 server and the proxy during the server connection phase. This value should
16565 always be very low, such as 1 ms on local networks and less than a few tens
16566 of ms on remote networks.
16567
Willy Tarreau55165fe2009-05-10 12:02:55 +020016568 - If "Tr" is nearly always lower than 3000 except some rare values which seem
16569 to be the average majored by 3000, there are probably some packets lost
16570 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016571
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016572 - If "Ta" is large even for small byte counts, it generally is because
16573 neither the client nor the server decides to close the connection while
16574 haproxy is running in tunnel mode and both have agreed on a keep-alive
16575 connection mode. In order to solve this issue, it will be needed to specify
16576 one of the HTTP options to manipulate keep-alive or close options on either
16577 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
16578 important when connection regulation is used with the "maxconn" option on
16579 the servers, since no new connection will be sent to the server until
16580 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016581
16582Other noticeable HTTP log cases ('xx' means any value to be ignored) :
16583
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016584 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016585 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016586 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016587
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016588 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016589 or it aborted too early. Check the session termination flags
16590 then "timeout http-request" and "timeout client" settings.
16591
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016592 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016593 servers were out of order, because the request was invalid
16594 or forbidden by ACL rules. Check the session termination
16595 flags.
16596
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016597 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
16598 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016599 Check the session termination flags, then check the
16600 "timeout connect" setting. Note that the tarpit action might
16601 return similar-looking patterns, with "Tw" equal to the time
16602 the client connection was maintained open.
16603
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016604 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016605 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016606 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016607 termination flags, then check the "timeout server" setting.
16608
16609
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200166108.5. Session state at disconnection
16611-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016612
16613TCP and HTTP logs provide a session termination indicator in the
16614"termination_state" field, just before the number of active connections. It is
166152-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
16616each of which has a special meaning :
16617
16618 - On the first character, a code reporting the first event which caused the
16619 session to terminate :
16620
16621 C : the TCP session was unexpectedly aborted by the client.
16622
16623 S : the TCP session was unexpectedly aborted by the server, or the
16624 server explicitly refused it.
16625
16626 P : the session was prematurely aborted by the proxy, because of a
16627 connection limit enforcement, because a DENY filter was matched,
16628 because of a security check which detected and blocked a dangerous
16629 error in server response which might have caused information leak
Davor Ocelice9ed2812017-12-25 17:49:28 +010016630 (e.g. cacheable cookie).
Willy Tarreau570f2212013-06-10 16:42:09 +020016631
16632 L : the session was locally processed by haproxy and was not passed to
16633 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016634
16635 R : a resource on the proxy has been exhausted (memory, sockets, source
16636 ports, ...). Usually, this appears during the connection phase, and
16637 system logs should contain a copy of the precise error. If this
16638 happens, it must be considered as a very serious anomaly which
16639 should be fixed as soon as possible by any means.
16640
16641 I : an internal error was identified by the proxy during a self-check.
16642 This should NEVER happen, and you are encouraged to report any log
16643 containing this, because this would almost certainly be a bug. It
16644 would be wise to preventively restart the process after such an
16645 event too, in case it would be caused by memory corruption.
16646
Simon Horman752dc4a2011-06-21 14:34:59 +090016647 D : the session was killed by haproxy because the server was detected
16648 as down and was configured to kill all connections when going down.
16649
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070016650 U : the session was killed by haproxy on this backup server because an
16651 active server was detected as up and was configured to kill all
16652 backup connections when going up.
16653
Willy Tarreaua2a64e92011-09-07 23:01:56 +020016654 K : the session was actively killed by an admin operating on haproxy.
16655
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016656 c : the client-side timeout expired while waiting for the client to
16657 send or receive data.
16658
16659 s : the server-side timeout expired while waiting for the server to
16660 send or receive data.
16661
16662 - : normal session completion, both the client and the server closed
16663 with nothing left in the buffers.
16664
16665 - on the second character, the TCP or HTTP session state when it was closed :
16666
Willy Tarreauf7b30a92010-12-06 22:59:17 +010016667 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016668 (HTTP mode only). Nothing was sent to any server.
16669
16670 Q : the proxy was waiting in the QUEUE for a connection slot. This can
16671 only happen when servers have a 'maxconn' parameter set. It can
16672 also happen in the global queue after a redispatch consecutive to
16673 a failed attempt to connect to a dying server. If no redispatch is
16674 reported, then no connection attempt was made to any server.
16675
16676 C : the proxy was waiting for the CONNECTION to establish on the
16677 server. The server might at most have noticed a connection attempt.
16678
16679 H : the proxy was waiting for complete, valid response HEADERS from the
16680 server (HTTP only).
16681
16682 D : the session was in the DATA phase.
16683
16684 L : the proxy was still transmitting LAST data to the client while the
16685 server had already finished. This one is very rare as it can only
16686 happen when the client dies while receiving the last packets.
16687
16688 T : the request was tarpitted. It has been held open with the client
16689 during the whole "timeout tarpit" duration or until the client
16690 closed, both of which will be reported in the "Tw" timer.
16691
16692 - : normal session completion after end of data transfer.
16693
16694 - the third character tells whether the persistence cookie was provided by
16695 the client (only in HTTP mode) :
16696
16697 N : the client provided NO cookie. This is usually the case for new
16698 visitors, so counting the number of occurrences of this flag in the
16699 logs generally indicate a valid trend for the site frequentation.
16700
16701 I : the client provided an INVALID cookie matching no known server.
16702 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020016703 cookies between HTTP/HTTPS sites, persistence conditionally
16704 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016705
16706 D : the client provided a cookie designating a server which was DOWN,
16707 so either "option persist" was used and the client was sent to
16708 this server, or it was not set and the client was redispatched to
16709 another server.
16710
Willy Tarreau996a92c2010-10-13 19:30:47 +020016711 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016712 server.
16713
Willy Tarreau996a92c2010-10-13 19:30:47 +020016714 E : the client provided a valid cookie, but with a last date which was
16715 older than what is allowed by the "maxidle" cookie parameter, so
16716 the cookie is consider EXPIRED and is ignored. The request will be
16717 redispatched just as if there was no cookie.
16718
16719 O : the client provided a valid cookie, but with a first date which was
16720 older than what is allowed by the "maxlife" cookie parameter, so
16721 the cookie is consider too OLD and is ignored. The request will be
16722 redispatched just as if there was no cookie.
16723
Willy Tarreauc89ccb62012-04-05 21:18:22 +020016724 U : a cookie was present but was not used to select the server because
16725 some other server selection mechanism was used instead (typically a
16726 "use-server" rule).
16727
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016728 - : does not apply (no cookie set in configuration).
16729
16730 - the last character reports what operations were performed on the persistence
16731 cookie returned by the server (only in HTTP mode) :
16732
16733 N : NO cookie was provided by the server, and none was inserted either.
16734
16735 I : no cookie was provided by the server, and the proxy INSERTED one.
16736 Note that in "cookie insert" mode, if the server provides a cookie,
16737 it will still be overwritten and reported as "I" here.
16738
Willy Tarreau996a92c2010-10-13 19:30:47 +020016739 U : the proxy UPDATED the last date in the cookie that was presented by
16740 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016741 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020016742 date indicated in the cookie. If any other change happens, such as
16743 a redispatch, then the cookie will be marked as inserted instead.
16744
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016745 P : a cookie was PROVIDED by the server and transmitted as-is.
16746
16747 R : the cookie provided by the server was REWRITTEN by the proxy, which
16748 happens in "cookie rewrite" or "cookie prefix" modes.
16749
16750 D : the cookie provided by the server was DELETED by the proxy.
16751
16752 - : does not apply (no cookie set in configuration).
16753
Willy Tarreau996a92c2010-10-13 19:30:47 +020016754The combination of the two first flags gives a lot of information about what
16755was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016756helpful to detect server saturation, network troubles, local system resource
16757starvation, attacks, etc...
16758
16759The most common termination flags combinations are indicated below. They are
16760alphabetically sorted, with the lowercase set just after the upper case for
16761easier finding and understanding.
16762
16763 Flags Reason
16764
16765 -- Normal termination.
16766
16767 CC The client aborted before the connection could be established to the
16768 server. This can happen when haproxy tries to connect to a recently
16769 dead (or unchecked) server, and the client aborts while haproxy is
16770 waiting for the server to respond or for "timeout connect" to expire.
16771
16772 CD The client unexpectedly aborted during data transfer. This can be
16773 caused by a browser crash, by an intermediate equipment between the
16774 client and haproxy which decided to actively break the connection,
16775 by network routing issues between the client and haproxy, or by a
16776 keep-alive session between the server and the client terminated first
16777 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010016778
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016779 cD The client did not send nor acknowledge any data for as long as the
16780 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020016781 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016782
16783 CH The client aborted while waiting for the server to start responding.
16784 It might be the server taking too long to respond or the client
16785 clicking the 'Stop' button too fast.
16786
16787 cH The "timeout client" stroke while waiting for client data during a
16788 POST request. This is sometimes caused by too large TCP MSS values
16789 for PPPoE networks which cannot transport full-sized packets. It can
16790 also happen when client timeout is smaller than server timeout and
16791 the server takes too long to respond.
16792
16793 CQ The client aborted while its session was queued, waiting for a server
16794 with enough empty slots to accept it. It might be that either all the
16795 servers were saturated or that the assigned server was taking too
16796 long a time to respond.
16797
16798 CR The client aborted before sending a full HTTP request. Most likely
16799 the request was typed by hand using a telnet client, and aborted
16800 too early. The HTTP status code is likely a 400 here. Sometimes this
16801 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020016802 and the client. "option http-ignore-probes" can be used to ignore
16803 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016804
16805 cR The "timeout http-request" stroke before the client sent a full HTTP
16806 request. This is sometimes caused by too large TCP MSS values on the
16807 client side for PPPoE networks which cannot transport full-sized
16808 packets, or by clients sending requests by hand and not typing fast
16809 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020016810 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020016811 some browsers started to implement a "pre-connect" feature consisting
16812 in speculatively connecting to some recently visited web sites just
16813 in case the user would like to visit them. This results in many
16814 connections being established to web sites, which end up in 408
16815 Request Timeout if the timeout strikes first, or 400 Bad Request when
16816 the browser decides to close them first. These ones pollute the log
16817 and feed the error counters. Some versions of some browsers have even
16818 been reported to display the error code. It is possible to work
Davor Ocelice9ed2812017-12-25 17:49:28 +010016819 around the undesirable effects of this behavior by adding "option
Willy Tarreau0f228a02015-05-01 15:37:53 +020016820 http-ignore-probes" in the frontend, resulting in connections with
16821 zero data transfer to be totally ignored. This will definitely hide
16822 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016823
16824 CT The client aborted while its session was tarpitted. It is important to
16825 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020016826 wrong tarpit rules have been written. If a lot of them happen, it
16827 might make sense to lower the "timeout tarpit" value to something
16828 closer to the average reported "Tw" timer, in order not to consume
16829 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016830
Willy Tarreau570f2212013-06-10 16:42:09 +020016831 LR The request was intercepted and locally handled by haproxy. Generally
16832 it means that this was a redirect or a stats request.
16833
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016834 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016835 the TCP connection (the proxy received a TCP RST or an ICMP message
16836 in return). Under some circumstances, it can also be the network
Davor Ocelice9ed2812017-12-25 17:49:28 +010016837 stack telling the proxy that the server is unreachable (e.g. no route,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016838 or no ARP response on local network). When this happens in HTTP mode,
16839 the status code is likely a 502 or 503 here.
16840
16841 sC The "timeout connect" stroke before a connection to the server could
16842 complete. When this happens in HTTP mode, the status code is likely a
16843 503 or 504 here.
16844
16845 SD The connection to the server died with an error during the data
16846 transfer. This usually means that haproxy has received an RST from
16847 the server or an ICMP message from an intermediate equipment while
16848 exchanging data with the server. This can be caused by a server crash
16849 or by a network issue on an intermediate equipment.
16850
16851 sD The server did not send nor acknowledge any data for as long as the
16852 "timeout server" setting during the data phase. This is often caused
Davor Ocelice9ed2812017-12-25 17:49:28 +010016853 by too short timeouts on L4 equipment before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016854 load-balancers, ...), as well as keep-alive sessions maintained
16855 between the client and the server expiring first on haproxy.
16856
16857 SH The server aborted before sending its full HTTP response headers, or
16858 it crashed while processing the request. Since a server aborting at
16859 this moment is very rare, it would be wise to inspect its logs to
16860 control whether it crashed and why. The logged request may indicate a
16861 small set of faulty requests, demonstrating bugs in the application.
16862 Sometimes this might also be caused by an IDS killing the connection
16863 between haproxy and the server.
16864
16865 sH The "timeout server" stroke before the server could return its
16866 response headers. This is the most common anomaly, indicating too
16867 long transactions, probably caused by server or database saturation.
16868 The immediate workaround consists in increasing the "timeout server"
16869 setting, but it is important to keep in mind that the user experience
16870 will suffer from these long response times. The only long term
16871 solution is to fix the application.
16872
16873 sQ The session spent too much time in queue and has been expired. See
16874 the "timeout queue" and "timeout connect" settings to find out how to
16875 fix this if it happens too often. If it often happens massively in
16876 short periods, it may indicate general problems on the affected
16877 servers due to I/O or database congestion, or saturation caused by
16878 external attacks.
16879
16880 PC The proxy refused to establish a connection to the server because the
16881 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020016882 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016883 so that it does not happen anymore. This status is very rare and
16884 might happen when the global "ulimit-n" parameter is forced by hand.
16885
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010016886 PD The proxy blocked an incorrectly formatted chunked encoded message in
16887 a request or a response, after the server has emitted its headers. In
16888 most cases, this will indicate an invalid message from the server to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016889 the client. HAProxy supports chunk sizes of up to 2GB - 1 (2147483647
Willy Tarreauf3a3e132013-08-31 08:16:26 +020016890 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010016891
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016892 PH The proxy blocked the server's response, because it was invalid,
16893 incomplete, dangerous (cache control), or matched a security filter.
16894 In any case, an HTTP 502 error is sent to the client. One possible
16895 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010016896 containing unauthorized characters. It is also possible but quite
16897 rare, that the proxy blocked a chunked-encoding request from the
16898 client due to an invalid syntax, before the server responded. In this
16899 case, an HTTP 400 error is sent to the client and reported in the
16900 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016901
16902 PR The proxy blocked the client's HTTP request, either because of an
16903 invalid HTTP syntax, in which case it returned an HTTP 400 error to
16904 the client, or because a deny filter matched, in which case it
16905 returned an HTTP 403 error.
16906
16907 PT The proxy blocked the client's request and has tarpitted its
16908 connection before returning it a 500 server error. Nothing was sent
16909 to the server. The connection was maintained open for as long as
16910 reported by the "Tw" timer field.
16911
16912 RC A local resource has been exhausted (memory, sockets, source ports)
16913 preventing the connection to the server from establishing. The error
16914 logs will tell precisely what was missing. This is very rare and can
16915 only be solved by proper system tuning.
16916
Willy Tarreau996a92c2010-10-13 19:30:47 +020016917The combination of the two last flags gives a lot of information about how
16918persistence was handled by the client, the server and by haproxy. This is very
16919important to troubleshoot disconnections, when users complain they have to
16920re-authenticate. The commonly encountered flags are :
16921
16922 -- Persistence cookie is not enabled.
16923
16924 NN No cookie was provided by the client, none was inserted in the
16925 response. For instance, this can be in insert mode with "postonly"
16926 set on a GET request.
16927
16928 II A cookie designating an invalid server was provided by the client,
16929 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016930 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020016931 value can be presented by a client when no other server knows it.
16932
16933 NI No cookie was provided by the client, one was inserted in the
16934 response. This typically happens for first requests from every user
16935 in "insert" mode, which makes it an easy way to count real users.
16936
16937 VN A cookie was provided by the client, none was inserted in the
16938 response. This happens for most responses for which the client has
16939 already got a cookie.
16940
16941 VU A cookie was provided by the client, with a last visit date which is
16942 not completely up-to-date, so an updated cookie was provided in
16943 response. This can also happen if there was no date at all, or if
16944 there was a date but the "maxidle" parameter was not set, so that the
16945 cookie can be switched to unlimited time.
16946
16947 EI A cookie was provided by the client, with a last visit date which is
16948 too old for the "maxidle" parameter, so the cookie was ignored and a
16949 new cookie was inserted in the response.
16950
16951 OI A cookie was provided by the client, with a first visit date which is
16952 too old for the "maxlife" parameter, so the cookie was ignored and a
16953 new cookie was inserted in the response.
16954
16955 DI The server designated by the cookie was down, a new server was
16956 selected and a new cookie was emitted in the response.
16957
16958 VI The server designated by the cookie was not marked dead but could not
16959 be reached. A redispatch happened and selected another one, which was
16960 then advertised in the response.
16961
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016962
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169638.6. Non-printable characters
16964-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016965
16966In order not to cause trouble to log analysis tools or terminals during log
16967consulting, non-printable characters are not sent as-is into log files, but are
16968converted to the two-digits hexadecimal representation of their ASCII code,
16969prefixed by the character '#'. The only characters that can be logged without
16970being escaped are comprised between 32 and 126 (inclusive). Obviously, the
16971escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
16972is the same for the character '"' which becomes "#22", as well as '{', '|' and
16973'}' when logging headers.
16974
16975Note that the space character (' ') is not encoded in headers, which can cause
16976issues for tools relying on space count to locate fields. A typical header
16977containing spaces is "User-Agent".
16978
16979Last, it has been observed that some syslog daemons such as syslog-ng escape
16980the quote ('"') with a backslash ('\'). The reverse operation can safely be
16981performed since no quote may appear anywhere else in the logs.
16982
16983
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169848.7. Capturing HTTP cookies
16985---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016986
16987Cookie capture simplifies the tracking a complete user session. This can be
16988achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016989section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016990cookie will simultaneously be checked in the request ("Cookie:" header) and in
16991the response ("Set-Cookie:" header). The respective values will be reported in
16992the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016993locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016994not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
16995user switches to a new session for example, because the server will reassign it
16996a new cookie. It is also possible to detect if a server unexpectedly sets a
16997wrong cookie to a client, leading to session crossing.
16998
16999 Examples :
17000 # capture the first cookie whose name starts with "ASPSESSION"
17001 capture cookie ASPSESSION len 32
17002
17003 # capture the first cookie whose name is exactly "vgnvisitor"
17004 capture cookie vgnvisitor= len 32
17005
17006
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170078.8. Capturing HTTP headers
17008---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017009
17010Header captures are useful to track unique request identifiers set by an upper
17011proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
17012the response, one can search for information about the response length, how the
17013server asked the cache to behave, or an object location during a redirection.
17014
17015Header captures are performed using the "capture request header" and "capture
17016response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017017section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017018
17019It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017020time. Non-existent headers are logged as empty strings, and if one header
17021appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017022are grouped within braces '{' and '}' in the same order as they were declared,
17023and delimited with a vertical bar '|' without any space. Response headers
17024follow the same representation, but are displayed after a space following the
17025request headers block. These blocks are displayed just before the HTTP request
17026in the logs.
17027
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020017028As a special case, it is possible to specify an HTTP header capture in a TCP
17029frontend. The purpose is to enable logging of headers which will be parsed in
17030an HTTP backend if the request is then switched to this HTTP backend.
17031
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017032 Example :
17033 # This instance chains to the outgoing proxy
17034 listen proxy-out
17035 mode http
17036 option httplog
17037 option logasap
17038 log global
17039 server cache1 192.168.1.1:3128
17040
17041 # log the name of the virtual server
17042 capture request header Host len 20
17043
17044 # log the amount of data uploaded during a POST
17045 capture request header Content-Length len 10
17046
17047 # log the beginning of the referrer
17048 capture request header Referer len 20
17049
17050 # server name (useful for outgoing proxies only)
17051 capture response header Server len 20
17052
17053 # logging the content-length is useful with "option logasap"
17054 capture response header Content-Length len 10
17055
Davor Ocelice9ed2812017-12-25 17:49:28 +010017056 # log the expected cache behavior on the response
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017057 capture response header Cache-Control len 8
17058
17059 # the Via header will report the next proxy's name
17060 capture response header Via len 20
17061
17062 # log the URL location during a redirection
17063 capture response header Location len 20
17064
17065 >>> Aug 9 20:26:09 localhost \
17066 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
17067 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
17068 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
17069 "GET http://fr.adserver.yahoo.com/"
17070
17071 >>> Aug 9 20:30:46 localhost \
17072 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
17073 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
17074 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017075 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017076
17077 >>> Aug 9 20:30:46 localhost \
17078 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
17079 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
17080 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
17081 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017082 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017083
17084
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170858.9. Examples of logs
17086---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017087
17088These are real-world examples of logs accompanied with an explanation. Some of
17089them have been made up by hand. The syslog part has been removed for better
17090reading. Their sole purpose is to explain how to decipher them.
17091
17092 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
17093 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
17094 "HEAD / HTTP/1.0"
17095
17096 => long request (6.5s) entered by hand through 'telnet'. The server replied
17097 in 147 ms, and the session ended normally ('----')
17098
17099 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
17100 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
17101 0/9 "HEAD / HTTP/1.0"
17102
17103 => Idem, but the request was queued in the global queue behind 9 other
17104 requests, and waited there for 1230 ms.
17105
17106 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
17107 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
17108 "GET /image.iso HTTP/1.0"
17109
17110 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017111 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017112 14 ms, 243 bytes of headers were sent to the client, and total time from
17113 accept to first data byte is 30 ms.
17114
17115 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
17116 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
17117 "GET /cgi-bin/bug.cgi? HTTP/1.0"
17118
17119 => the proxy blocked a server response either because of an "rspdeny" or
17120 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020017121 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017122 risked being cached. In this case, the response is replaced with a "502
17123 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
17124 to return the 502 and not the server.
17125
17126 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017127 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 +010017128
17129 => the client never completed its request and aborted itself ("C---") after
17130 8.5s, while the proxy was waiting for the request headers ("-R--").
17131 Nothing was sent to any server.
17132
17133 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
17134 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
17135
17136 => The client never completed its request, which was aborted by the
17137 time-out ("c---") after 50s, while the proxy was waiting for the request
Davor Ocelice9ed2812017-12-25 17:49:28 +010017138 headers ("-R--"). Nothing was sent to any server, but the proxy could
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017139 send a 408 return code to the client.
17140
17141 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
17142 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
17143
17144 => This log was produced with "option tcplog". The client timed out after
17145 5 seconds ("c----").
17146
17147 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
17148 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017149 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017150
17151 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017152 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017153 (config says 'retries 3'), and no redispatch (otherwise we would have
17154 seen "/+3"). Status code 503 was returned to the client. There were 115
17155 connections on this server, 202 connections on this proxy, and 205 on
17156 the global process. It is possible that the server refused the
17157 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010017158
Willy Tarreau52b2d222011-09-07 23:48:48 +020017159
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200171609. Supported filters
17161--------------------
17162
17163Here are listed officially supported filters with the list of parameters they
17164accept. Depending on compile options, some of these filters might be
17165unavailable. The list of available filters is reported in haproxy -vv.
17166
17167See also : "filter"
17168
171699.1. Trace
17170----------
17171
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010017172filter trace [name <name>] [random-parsing] [random-forwarding] [hexdump]
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017173
17174 Arguments:
17175 <name> is an arbitrary name that will be reported in
17176 messages. If no name is provided, "TRACE" is used.
17177
17178 <random-parsing> enables the random parsing of data exchanged between
17179 the client and the server. By default, this filter
17180 parses all available data. With this parameter, it
17181 only parses a random amount of the available data.
17182
Davor Ocelice9ed2812017-12-25 17:49:28 +010017183 <random-forwarding> enables the random forwarding of parsed data. By
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017184 default, this filter forwards all previously parsed
17185 data. With this parameter, it only forwards a random
17186 amount of the parsed data.
17187
Davor Ocelice9ed2812017-12-25 17:49:28 +010017188 <hexdump> dumps all forwarded data to the server and the client.
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010017189
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017190This filter can be used as a base to develop new filters. It defines all
17191callbacks and print a message on the standard error stream (stderr) with useful
17192information for all of them. It may be useful to debug the activity of other
17193filters or, quite simply, HAProxy's activity.
17194
17195Using <random-parsing> and/or <random-forwarding> parameters is a good way to
17196tests the behavior of a filter that parses data exchanged between a client and
17197a server by adding some latencies in the processing.
17198
17199
172009.2. HTTP compression
17201---------------------
17202
17203filter compression
17204
17205The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
17206keyword must still be used to enable and configure the HTTP compression. And
17207when no other filter is used, it is enough. But it is mandatory to explicitly
17208use a filter line to enable the HTTP compression when two or more filters are
17209used for the same listener/frontend/backend. This is important to know the
17210filters evaluation order.
17211
17212See also : "compression"
17213
17214
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200172159.3. Stream Processing Offload Engine (SPOE)
17216--------------------------------------------
17217
17218filter spoe [engine <name>] config <file>
17219
17220 Arguments :
17221
17222 <name> is the engine name that will be used to find the right scope in
17223 the configuration file. If not provided, all the file will be
17224 parsed.
17225
17226 <file> is the path of the engine configuration file. This file can
17227 contain configuration of several engines. In this case, each
17228 part must be placed in its own scope.
17229
17230The Stream Processing Offload Engine (SPOE) is a filter communicating with
17231external components. It allows the offload of some specifics processing on the
Davor Ocelice9ed2812017-12-25 17:49:28 +010017232streams in tiered applications. These external components and information
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020017233exchanged with them are configured in dedicated files, for the main part. It
17234also requires dedicated backends, defined in HAProxy configuration.
17235
17236SPOE communicates with external components using an in-house binary protocol,
17237the Stream Processing Offload Protocol (SPOP).
17238
Tim Düsterhus4896c442016-11-29 02:15:19 +010017239For all information about the SPOE configuration and the SPOP specification, see
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020017240"doc/SPOE.txt".
17241
17242Important note:
17243 The SPOE filter is highly experimental for now and was not heavily
17244 tested. It is really not production ready. So use it carefully.
17245
William Lallemand86d0df02017-11-24 21:36:45 +01001724610. Cache
17247---------
17248
17249HAProxy provides a cache, which was designed to perform cache on small objects
17250(favicon, css...). This is a minimalist low-maintenance cache which runs in
17251RAM.
17252
17253The cache is based on a memory which is shared between processes and threads,
Cyril Bonté7b888f12017-11-26 22:24:31 +010017254this memory is split in blocks of 1k.
William Lallemand86d0df02017-11-24 21:36:45 +010017255
17256If an object is not used anymore, it can be deleted to store a new object
17257independently of its expiration date. The oldest objects are deleted first
17258when we try to allocate a new one.
17259
Cyril Bonté7b888f12017-11-26 22:24:31 +010017260The cache uses a hash of the host header and the URI as the key.
William Lallemand86d0df02017-11-24 21:36:45 +010017261
17262It's possible to view the status of a cache using the Unix socket command
17263"show cache" consult section 9.3 "Unix Socket commands" of Management Guide
17264for more details.
17265
17266When an object is delivered from the cache, the server name in the log is
17267replaced by "<CACHE>".
17268
Cyril Bonté7b888f12017-11-26 22:24:31 +01001726910.1. Limitation
17270----------------
William Lallemand86d0df02017-11-24 21:36:45 +010017271
17272The cache won't store and won't deliver objects in these cases:
17273
17274- If the response is not a 200
17275- If the response contains a Vary header
17276- If the response does not contain a Content-Length header or if the
17277 Content-Length + the headers size is greater than a buffer size - the
17278 reserve.
17279- If the response is not cacheable
17280
17281- If the request is not a GET
17282- If the HTTP version of the request is smaller than 1.1
William Lallemand8a16fe02018-05-22 11:04:33 +020017283- If the request contains an Authorization header
William Lallemand86d0df02017-11-24 21:36:45 +010017284
17285Caution!: Due to the current limitation of the filters, it is not recommended
17286to use the cache with other filters. Using them can cause undefined behavior
Cyril Bonté7b888f12017-11-26 22:24:31 +010017287if they modify the response (compression for example).
William Lallemand86d0df02017-11-24 21:36:45 +010017288
Cyril Bonté7b888f12017-11-26 22:24:31 +01001728910.2. Setup
17290-----------
William Lallemand86d0df02017-11-24 21:36:45 +010017291
17292To setup a cache, you must define a cache section and use it in a proxy with
17293the corresponding http-request and response actions.
17294
Cyril Bonté7b888f12017-11-26 22:24:31 +01001729510.2.1. Cache section
17296---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010017297
17298cache <name>
17299 Declare a cache section, allocate a shared cache memory named <name>, the
17300 size of cache is mandatory.
17301
17302total-max-size <megabytes>
Davor Ocelice9ed2812017-12-25 17:49:28 +010017303 Define the size in RAM of the cache in megabytes. This size is split in
17304 blocks of 1kB which are used by the cache entries.
William Lallemand86d0df02017-11-24 21:36:45 +010017305
17306max-age <seconds>
17307 Define the maximum expiration duration. The expiration is set has the lowest
17308 value between the s-maxage or max-age (in this order) directive in the
17309 Cache-Control response header and this value. The default value is 60
17310 seconds, which means that you can't cache an object more than 60 seconds by
17311 default.
17312
Cyril Bonté7b888f12017-11-26 22:24:31 +01001731310.2.2. Proxy section
17314---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010017315
17316http-request cache-use <name>
17317 Try to deliver a cached object from the cache <name>. This directive is also
17318 mandatory to store the cache as it calculates the cache hash. If you want to
17319 use a condition for both storage and delivering that's a good idea to put it
17320 after this one.
17321
17322http-response cache-store <name>
17323 Store an http-response within the cache. The storage of the response headers
17324 is done at this step, which means you can use others http-response actions
17325 to modify headers before or after the storage of the response. This action
17326 is responsible for the setup of the cache storage filter.
17327
17328
17329Example:
17330
17331 backend bck1
17332 mode http
17333
17334 http-request cache-use foobar
17335 http-response cache-store foobar
17336 server srv1 127.0.0.1:80
17337
17338 cache foobar
17339 total-max-size 4
17340 max-age 240
17341
Willy Tarreau0ba27502007-12-24 16:55:16 +010017342/*
17343 * Local variables:
17344 * fill-column: 79
17345 * End:
17346 */