blob: adbad95d346d4bf46c80a27cad9b88038da4e9b8 [file] [log] [blame]
Willy Tarreau2212e6a2015-10-13 14:40:55 +02001 ------------------------
2 HAProxy Management Guide
3 ------------------------
Willy Tarreau33205c22020-07-07 16:35:28 +02004 version 2.3
Willy Tarreau2212e6a2015-10-13 14:40:55 +02005
6
7This document describes how to start, stop, manage, and troubleshoot HAProxy,
8as well as some known limitations and traps to avoid. It does not describe how
9to configure it (for this please read configuration.txt).
10
11Note to documentation contributors :
12 This document is formatted with 80 columns per line, with even number of
13 spaces for indentation and without tabs. Please follow these rules strictly
14 so that it remains easily printable everywhere. If you add sections, please
15 update the summary below for easier searching.
16
17
18Summary
19-------
20
211. Prerequisites
222. Quick reminder about HAProxy's architecture
233. Starting HAProxy
244. Stopping and restarting HAProxy
255. File-descriptor limitations
266. Memory management
277. CPU usage
288. Logging
299. Statistics and monitoring
Willy Tarreau44aed902015-10-13 14:45:29 +0200309.1. CSV format
Willy Tarreau5d8b9792016-03-11 11:09:34 +0100319.2. Typed output format
329.3. Unix Socket commands
William Lallemand142db372018-12-11 18:56:45 +0100339.4. Master CLI
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003410. Tricks for easier configuration management
3511. Well-known traps to avoid
3612. Debugging and performance issues
3713. Security considerations
38
39
401. Prerequisites
41----------------
42
43In this document it is assumed that the reader has sufficient administration
44skills on a UNIX-like operating system, uses the shell on a daily basis and is
45familiar with troubleshooting utilities such as strace and tcpdump.
46
47
482. Quick reminder about HAProxy's architecture
49----------------------------------------------
50
Willy Tarreau3f364482019-02-27 15:01:46 +010051HAProxy is a multi-threaded, event-driven, non-blocking daemon. This means is
Willy Tarreau2212e6a2015-10-13 14:40:55 +020052uses event multiplexing to schedule all of its activities instead of relying on
53the system to schedule between multiple activities. Most of the time it runs as
54a single process, so the output of "ps aux" on a system will report only one
55"haproxy" process, unless a soft reload is in progress and an older process is
56finishing its job in parallel to the new one. It is thus always easy to trace
Willy Tarreau3f364482019-02-27 15:01:46 +010057its activity using the strace utility. In order to scale with the number of
58available processors, by default haproxy will start one worker thread per
59processor it is allowed to run on. Unless explicitly configured differently,
60the incoming traffic is spread over all these threads, all running the same
61event loop. A great care is taken to limit inter-thread dependencies to the
62strict minimum, so as to try to achieve near-linear scalability. This has some
63impacts such as the fact that a given connection is served by a single thread.
64Thus in order to use all available processing capacity, it is needed to have at
65least as many connections as there are threads, which is almost always granted.
Willy Tarreau2212e6a2015-10-13 14:40:55 +020066
67HAProxy is designed to isolate itself into a chroot jail during startup, where
68it cannot perform any file-system access at all. This is also true for the
69libraries it depends on (eg: libc, libssl, etc). The immediate effect is that
70a running process will not be able to reload a configuration file to apply
71changes, instead a new process will be started using the updated configuration
72file. Some other less obvious effects are that some timezone files or resolver
73files the libc might attempt to access at run time will not be found, though
74this should generally not happen as they're not needed after startup. A nice
75consequence of this principle is that the HAProxy process is totally stateless,
76and no cleanup is needed after it's killed, so any killing method that works
77will do the right thing.
78
79HAProxy doesn't write log files, but it relies on the standard syslog protocol
80to send logs to a remote server (which is often located on the same system).
81
82HAProxy uses its internal clock to enforce timeouts, that is derived from the
83system's time but where unexpected drift is corrected. This is done by limiting
84the time spent waiting in poll() for an event, and measuring the time it really
85took. In practice it never waits more than one second. This explains why, when
86running strace over a completely idle process, periodic calls to poll() (or any
87of its variants) surrounded by two gettimeofday() calls are noticed. They are
88normal, completely harmless and so cheap that the load they imply is totally
89undetectable at the system scale, so there's nothing abnormal there. Example :
90
91 16:35:40.002320 gettimeofday({1442759740, 2605}, NULL) = 0
92 16:35:40.002942 epoll_wait(0, {}, 200, 1000) = 0
93 16:35:41.007542 gettimeofday({1442759741, 7641}, NULL) = 0
94 16:35:41.007998 gettimeofday({1442759741, 8114}, NULL) = 0
95 16:35:41.008391 epoll_wait(0, {}, 200, 1000) = 0
96 16:35:42.011313 gettimeofday({1442759742, 11411}, NULL) = 0
97
98HAProxy is a TCP proxy, not a router. It deals with established connections that
99have been validated by the kernel, and not with packets of any form nor with
100sockets in other states (eg: no SYN_RECV nor TIME_WAIT), though their existence
101may prevent it from binding a port. It relies on the system to accept incoming
102connections and to initiate outgoing connections. An immediate effect of this is
103that there is no relation between packets observed on the two sides of a
104forwarded connection, which can be of different size, numbers and even family.
105Since a connection may only be accepted from a socket in LISTEN state, all the
106sockets it is listening to are necessarily visible using the "netstat" utility
107to show listening sockets. Example :
108
109 # netstat -ltnp
110 Active Internet connections (only servers)
111 Proto Recv-Q Send-Q Local Address Foreign Address State PID/Program name
112 tcp 0 0 0.0.0.0:22 0.0.0.0:* LISTEN 1629/sshd
113 tcp 0 0 0.0.0.0:80 0.0.0.0:* LISTEN 2847/haproxy
114 tcp 0 0 0.0.0.0:443 0.0.0.0:* LISTEN 2847/haproxy
115
116
1173. Starting HAProxy
118-------------------
119
120HAProxy is started by invoking the "haproxy" program with a number of arguments
121passed on the command line. The actual syntax is :
122
123 $ haproxy [<options>]*
124
125where [<options>]* is any number of options. An option always starts with '-'
126followed by one of more letters, and possibly followed by one or multiple extra
127arguments. Without any option, HAProxy displays the help page with a reminder
128about supported options. Available options may vary slightly based on the
129operating system. A fair number of these options overlap with an equivalent one
130if the "global" section. In this case, the command line always has precedence
131over the configuration file, so that the command line can be used to quickly
132enforce some settings without touching the configuration files. The current
133list of options is :
134
135 -- <cfgfile>* : all the arguments following "--" are paths to configuration
Maxime de Roucy379d9c72016-05-13 23:52:56 +0200136 file/directory to be loaded and processed in the declaration order. It is
137 mostly useful when relying on the shell to load many files that are
138 numerically ordered. See also "-f". The difference between "--" and "-f" is
139 that one "-f" must be placed before each file name, while a single "--" is
140 needed before all file names. Both options can be used together, the
141 command line ordering still applies. When more than one file is specified,
142 each file must start on a section boundary, so the first keyword of each
143 file must be one of "global", "defaults", "peers", "listen", "frontend",
144 "backend", and so on. A file cannot contain just a server list for example.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200145
Maxime de Roucy379d9c72016-05-13 23:52:56 +0200146 -f <cfgfile|cfgdir> : adds <cfgfile> to the list of configuration files to be
147 loaded. If <cfgdir> is a directory, all the files (and only files) it
Dan Lloyd8e48b872016-07-01 21:01:18 -0400148 contains are added in lexical order (using LC_COLLATE=C) to the list of
Maxime de Roucy379d9c72016-05-13 23:52:56 +0200149 configuration files to be loaded ; only files with ".cfg" extension are
150 added, only non hidden files (not prefixed with ".") are added.
151 Configuration files are loaded and processed in their declaration order.
152 This option may be specified multiple times to load multiple files. See
153 also "--". The difference between "--" and "-f" is that one "-f" must be
154 placed before each file name, while a single "--" is needed before all file
155 names. Both options can be used together, the command line ordering still
156 applies. When more than one file is specified, each file must start on a
157 section boundary, so the first keyword of each file must be one of
158 "global", "defaults", "peers", "listen", "frontend", "backend", and so on.
159 A file cannot contain just a server list for example.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200160
161 -C <dir> : changes to directory <dir> before loading configuration
162 files. This is useful when using relative paths. Warning when using
163 wildcards after "--" which are in fact replaced by the shell before
164 starting haproxy.
165
166 -D : start as a daemon. The process detaches from the current terminal after
167 forking, and errors are not reported anymore in the terminal. It is
168 equivalent to the "daemon" keyword in the "global" section of the
169 configuration. It is recommended to always force it in any init script so
170 that a faulty configuration doesn't prevent the system from booting.
171
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200172 -L <name> : change the local peer name to <name>, which defaults to the local
William Lallemanddaf4cd22018-04-17 16:46:13 +0200173 hostname. This is used only with peers replication. You can use the
174 variable $HAPROXY_LOCALPEER in the configuration file to reference the
175 peer name.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200176
177 -N <limit> : sets the default per-proxy maxconn to <limit> instead of the
178 builtin default value (usually 2000). Only useful for debugging.
179
180 -V : enable verbose mode (disables quiet mode). Reverts the effect of "-q" or
181 "quiet".
182
William Lallemande202b1e2017-06-01 17:38:56 +0200183 -W : master-worker mode. It is equivalent to the "master-worker" keyword in
184 the "global" section of the configuration. This mode will launch a "master"
185 which will monitor the "workers". Using this mode, you can reload HAProxy
186 directly by sending a SIGUSR2 signal to the master. The master-worker mode
187 is compatible either with the foreground or daemon mode. It is
188 recommended to use this mode with multiprocess and systemd.
189
Pavlos Parissisf65f2572018-02-07 21:42:16 +0100190 -Ws : master-worker mode with support of `notify` type of systemd service.
191 This option is only available when HAProxy was built with `USE_SYSTEMD`
192 build option enabled.
193
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200194 -c : only performs a check of the configuration files and exits before trying
195 to bind. The exit status is zero if everything is OK, or non-zero if an
Willy Tarreaubebd2122020-04-15 16:06:11 +0200196 error is encountered. Presence of warnings will be reported if any.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200197
198 -d : enable debug mode. This disables daemon mode, forces the process to stay
199 in foreground and to show incoming and outgoing events. It is equivalent to
200 the "global" section's "debug" keyword. It must never be used in an init
201 script.
202
203 -dG : disable use of getaddrinfo() to resolve host names into addresses. It
204 can be used when suspecting that getaddrinfo() doesn't work as expected.
205 This option was made available because many bogus implementations of
206 getaddrinfo() exist on various systems and cause anomalies that are
207 difficult to troubleshoot.
208
Dan Lloyd8e48b872016-07-01 21:01:18 -0400209 -dM[<byte>] : forces memory poisoning, which means that each and every
Willy Tarreaubafbe012017-11-24 17:34:44 +0100210 memory region allocated with malloc() or pool_alloc() will be filled with
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200211 <byte> before being passed to the caller. When <byte> is not specified, it
212 defaults to 0x50 ('P'). While this slightly slows down operations, it is
213 useful to reliably trigger issues resulting from missing initializations in
214 the code that cause random crashes. Note that -dM0 has the effect of
215 turning any malloc() into a calloc(). In any case if a bug appears or
216 disappears when using this option it means there is a bug in haproxy, so
217 please report it.
218
219 -dS : disable use of the splice() system call. It is equivalent to the
220 "global" section's "nosplice" keyword. This may be used when splice() is
221 suspected to behave improperly or to cause performance issues, or when
222 using strace to see the forwarded data (which do not appear when using
223 splice()).
224
225 -dV : disable SSL verify on the server side. It is equivalent to having
226 "ssl-server-verify none" in the "global" section. This is useful when
227 trying to reproduce production issues out of the production
228 environment. Never use this in an init script as it degrades SSL security
229 to the servers.
230
Willy Tarreau3eb10b82020-04-15 16:42:39 +0200231 -dW : if set, haproxy will refuse to start if any warning was emitted while
232 processing the configuration. This helps detect subtle mistakes and keep the
233 configuration clean and portable across versions. It is recommended to set
234 this option in service scripts when configurations are managed by humans,
235 but it is recommended not to use it with generated configurations, which
236 tend to emit more warnings. It may be combined with "-c" to cause warnings
237 in checked configurations to fail. This is equivalent to global option
238 "zero-warning".
239
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200240 -db : disable background mode and multi-process mode. The process remains in
241 foreground. It is mainly used during development or during small tests, as
242 Ctrl-C is enough to stop the process. Never use it in an init script.
243
244 -de : disable the use of the "epoll" poller. It is equivalent to the "global"
245 section's keyword "noepoll". It is mostly useful when suspecting a bug
246 related to this poller. On systems supporting epoll, the fallback will
247 generally be the "poll" poller.
248
249 -dk : disable the use of the "kqueue" poller. It is equivalent to the
250 "global" section's keyword "nokqueue". It is mostly useful when suspecting
251 a bug related to this poller. On systems supporting kqueue, the fallback
252 will generally be the "poll" poller.
253
254 -dp : disable the use of the "poll" poller. It is equivalent to the "global"
255 section's keyword "nopoll". It is mostly useful when suspecting a bug
256 related to this poller. On systems supporting poll, the fallback will
257 generally be the "select" poller, which cannot be disabled and is limited
258 to 1024 file descriptors.
259
Willy Tarreau3eed10e2016-11-07 21:03:16 +0100260 -dr : ignore server address resolution failures. It is very common when
261 validating a configuration out of production not to have access to the same
262 resolvers and to fail on server address resolution, making it difficult to
263 test a configuration. This option simply appends the "none" method to the
264 list of address resolution methods for all servers, ensuring that even if
265 the libc fails to resolve an address, the startup sequence is not
266 interrupted.
267
Willy Tarreau70060452015-12-14 12:46:07 +0100268 -m <limit> : limit the total allocatable memory to <limit> megabytes across
269 all processes. This may cause some connection refusals or some slowdowns
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200270 depending on the amount of memory needed for normal operations. This is
Willy Tarreau70060452015-12-14 12:46:07 +0100271 mostly used to force the processes to work in a constrained resource usage
272 scenario. It is important to note that the memory is not shared between
273 processes, so in a multi-process scenario, this value is first divided by
274 global.nbproc before forking.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200275
276 -n <limit> : limits the per-process connection limit to <limit>. This is
277 equivalent to the global section's keyword "maxconn". It has precedence
278 over this keyword. This may be used to quickly force lower limits to avoid
279 a service outage on systems where resource limits are too low.
280
281 -p <file> : write all processes' pids into <file> during startup. This is
282 equivalent to the "global" section's keyword "pidfile". The file is opened
283 before entering the chroot jail, and after doing the chdir() implied by
284 "-C". Each pid appears on its own line.
285
286 -q : set "quiet" mode. This disables some messages during the configuration
287 parsing and during startup. It can be used in combination with "-c" to
288 just check if a configuration file is valid or not.
289
William Lallemand142db372018-12-11 18:56:45 +0100290 -S <bind>[,bind_options...]: in master-worker mode, bind a master CLI, which
291 allows the access to every processes, running or leaving ones.
292 For security reasons, it is recommended to bind the master CLI to a local
293 UNIX socket. The bind options are the same as the keyword "bind" in
294 the configuration file with words separated by commas instead of spaces.
295
296 Note that this socket can't be used to retrieve the listening sockets from
297 an old process during a seamless reload.
298
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200299 -sf <pid>* : send the "finish" signal (SIGUSR1) to older processes after boot
300 completion to ask them to finish what they are doing and to leave. <pid>
301 is a list of pids to signal (one per argument). The list ends on any
302 option starting with a "-". It is not a problem if the list of pids is
303 empty, so that it can be built on the fly based on the result of a command
304 like "pidof" or "pgrep".
305
306 -st <pid>* : send the "terminate" signal (SIGTERM) to older processes after
307 boot completion to terminate them immediately without finishing what they
308 were doing. <pid> is a list of pids to signal (one per argument). The list
309 is ends on any option starting with a "-". It is not a problem if the list
310 of pids is empty, so that it can be built on the fly based on the result of
311 a command like "pidof" or "pgrep".
312
313 -v : report the version and build date.
314
315 -vv : display the version, build options, libraries versions and usable
316 pollers. This output is systematically requested when filing a bug report.
317
Olivier Houchardd33fc3a2017-04-05 22:50:59 +0200318 -x <unix_socket> : connect to the specified socket and try to retrieve any
319 listening sockets from the old process, and use them instead of trying to
320 bind new ones. This is useful to avoid missing any new connection when
William Lallemandf6975e92017-05-26 17:42:10 +0200321 reloading the configuration on Linux. The capability must be enable on the
322 stats socket using "expose-fd listeners" in your configuration.
Olivier Houchardd33fc3a2017-04-05 22:50:59 +0200323
Dan Lloyd8e48b872016-07-01 21:01:18 -0400324A safe way to start HAProxy from an init file consists in forcing the daemon
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200325mode, storing existing pids to a pid file and using this pid file to notify
326older processes to finish before leaving :
327
328 haproxy -f /etc/haproxy.cfg \
329 -D -p /var/run/haproxy.pid -sf $(cat /var/run/haproxy.pid)
330
331When the configuration is split into a few specific files (eg: tcp vs http),
332it is recommended to use the "-f" option :
333
334 haproxy -f /etc/haproxy/global.cfg -f /etc/haproxy/stats.cfg \
335 -f /etc/haproxy/default-tcp.cfg -f /etc/haproxy/tcp.cfg \
336 -f /etc/haproxy/default-http.cfg -f /etc/haproxy/http.cfg \
337 -D -p /var/run/haproxy.pid -sf $(cat /var/run/haproxy.pid)
338
339When an unknown number of files is expected, such as customer-specific files,
340it is recommended to assign them a name starting with a fixed-size sequence
341number and to use "--" to load them, possibly after loading some defaults :
342
343 haproxy -f /etc/haproxy/global.cfg -f /etc/haproxy/stats.cfg \
344 -f /etc/haproxy/default-tcp.cfg -f /etc/haproxy/tcp.cfg \
345 -f /etc/haproxy/default-http.cfg -f /etc/haproxy/http.cfg \
346 -D -p /var/run/haproxy.pid -sf $(cat /var/run/haproxy.pid) \
347 -f /etc/haproxy/default-customers.cfg -- /etc/haproxy/customers/*
348
349Sometimes a failure to start may happen for whatever reason. Then it is
350important to verify if the version of HAProxy you are invoking is the expected
351version and if it supports the features you are expecting (eg: SSL, PCRE,
352compression, Lua, etc). This can be verified using "haproxy -vv". Some
353important information such as certain build options, the target system and
354the versions of the libraries being used are reported there. It is also what
355you will systematically be asked for when posting a bug report :
356
357 $ haproxy -vv
358 HA-Proxy version 1.6-dev7-a088d3-4 2015/10/08
359 Copyright 2000-2015 Willy Tarreau <willy@haproxy.org>
360
361 Build options :
362 TARGET = linux2628
363 CPU = generic
364 CC = gcc
365 CFLAGS = -pg -O0 -g -fno-strict-aliasing -Wdeclaration-after-statement \
366 -DBUFSIZE=8030 -DMAXREWRITE=1030 -DSO_MARK=36 -DTCP_REPAIR=19
367 OPTIONS = USE_ZLIB=1 USE_DLMALLOC=1 USE_OPENSSL=1 USE_LUA=1 USE_PCRE=1
368
369 Default settings :
370 maxconn = 2000, bufsize = 8030, maxrewrite = 1030, maxpollevents = 200
371
372 Encrypted password support via crypt(3): yes
373 Built with zlib version : 1.2.6
374 Compression algorithms supported : identity("identity"), deflate("deflate"), \
375 raw-deflate("deflate"), gzip("gzip")
376 Built with OpenSSL version : OpenSSL 1.0.1o 12 Jun 2015
377 Running on OpenSSL version : OpenSSL 1.0.1o 12 Jun 2015
378 OpenSSL library supports TLS extensions : yes
379 OpenSSL library supports SNI : yes
380 OpenSSL library supports prefer-server-ciphers : yes
381 Built with PCRE version : 8.12 2011-01-15
382 PCRE library supports JIT : no (USE_PCRE_JIT not set)
383 Built with Lua version : Lua 5.3.1
384 Built with transparent proxy support using: IP_TRANSPARENT IP_FREEBIND
385
386 Available polling systems :
387 epoll : pref=300, test result OK
388 poll : pref=200, test result OK
389 select : pref=150, test result OK
390 Total: 3 (3 usable), will use epoll.
391
392The relevant information that many non-developer users can verify here are :
393 - the version : 1.6-dev7-a088d3-4 above means the code is currently at commit
394 ID "a088d3" which is the 4th one after after official version "1.6-dev7".
395 Version 1.6-dev7 would show as "1.6-dev7-8c1ad7". What matters here is in
396 fact "1.6-dev7". This is the 7th development version of what will become
397 version 1.6 in the future. A development version not suitable for use in
398 production (unless you know exactly what you are doing). A stable version
399 will show as a 3-numbers version, such as "1.5.14-16f863", indicating the
400 14th level of fix on top of version 1.5. This is a production-ready version.
401
402 - the release date : 2015/10/08. It is represented in the universal
403 year/month/day format. Here this means August 8th, 2015. Given that stable
404 releases are issued every few months (1-2 months at the beginning, sometimes
405 6 months once the product becomes very stable), if you're seeing an old date
406 here, it means you're probably affected by a number of bugs or security
407 issues that have since been fixed and that it might be worth checking on the
408 official site.
409
410 - build options : they are relevant to people who build their packages
411 themselves, they can explain why things are not behaving as expected. For
412 example the development version above was built for Linux 2.6.28 or later,
Dan Lloyd8e48b872016-07-01 21:01:18 -0400413 targeting a generic CPU (no CPU-specific optimizations), and lacks any
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200414 code optimization (-O0) so it will perform poorly in terms of performance.
415
416 - libraries versions : zlib version is reported as found in the library
417 itself. In general zlib is considered a very stable product and upgrades
418 are almost never needed. OpenSSL reports two versions, the version used at
419 build time and the one being used, as found on the system. These ones may
420 differ by the last letter but never by the numbers. The build date is also
421 reported because most OpenSSL bugs are security issues and need to be taken
422 seriously, so this library absolutely needs to be kept up to date. Seeing a
423 4-months old version here is highly suspicious and indeed an update was
424 missed. PCRE provides very fast regular expressions and is highly
425 recommended. Certain of its extensions such as JIT are not present in all
426 versions and still young so some people prefer not to build with them,
Dan Lloyd8e48b872016-07-01 21:01:18 -0400427 which is why the build status is reported as well. Regarding the Lua
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200428 scripting language, HAProxy expects version 5.3 which is very young since
429 it was released a little time before HAProxy 1.6. It is important to check
430 on the Lua web site if some fixes are proposed for this branch.
431
432 - Available polling systems will affect the process's scalability when
433 dealing with more than about one thousand of concurrent connections. These
434 ones are only available when the correct system was indicated in the TARGET
435 variable during the build. The "epoll" mechanism is highly recommended on
436 Linux, and the kqueue mechanism is highly recommended on BSD. Lacking them
437 will result in poll() or even select() being used, causing a high CPU usage
438 when dealing with a lot of connections.
439
440
4414. Stopping and restarting HAProxy
442----------------------------------
443
444HAProxy supports a graceful and a hard stop. The hard stop is simple, when the
445SIGTERM signal is sent to the haproxy process, it immediately quits and all
446established connections are closed. The graceful stop is triggered when the
447SIGUSR1 signal is sent to the haproxy process. It consists in only unbinding
448from listening ports, but continue to process existing connections until they
449close. Once the last connection is closed, the process leaves.
450
451The hard stop method is used for the "stop" or "restart" actions of the service
452management script. The graceful stop is used for the "reload" action which
453tries to seamlessly reload a new configuration in a new process.
454
455Both of these signals may be sent by the new haproxy process itself during a
456reload or restart, so that they are sent at the latest possible moment and only
457if absolutely required. This is what is performed by the "-st" (hard) and "-sf"
458(graceful) options respectively.
459
William Lallemande202b1e2017-06-01 17:38:56 +0200460In master-worker mode, it is not needed to start a new haproxy process in
461order to reload the configuration. The master process reacts to the SIGUSR2
462signal by reexecuting itself with the -sf parameter followed by the PIDs of
463the workers. The master will then parse the configuration file and fork new
464workers.
465
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200466To understand better how these signals are used, it is important to understand
467the whole restart mechanism.
468
469First, an existing haproxy process is running. The administrator uses a system
Jackie Tapia749f74c2020-07-22 18:59:40 -0500470specific command such as "/etc/init.d/haproxy reload" to indicate they want to
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200471take the new configuration file into effect. What happens then is the following.
472First, the service script (/etc/init.d/haproxy or equivalent) will verify that
473the configuration file parses correctly using "haproxy -c". After that it will
474try to start haproxy with this configuration file, using "-st" or "-sf".
475
476Then HAProxy tries to bind to all listening ports. If some fatal errors happen
477(eg: address not present on the system, permission denied), the process quits
478with an error. If a socket binding fails because a port is already in use, then
479the process will first send a SIGTTOU signal to all the pids specified in the
480"-st" or "-sf" pid list. This is what is called the "pause" signal. It instructs
481all existing haproxy processes to temporarily stop listening to their ports so
482that the new process can try to bind again. During this time, the old process
483continues to process existing connections. If the binding still fails (because
484for example a port is shared with another daemon), then the new process sends a
485SIGTTIN signal to the old processes to instruct them to resume operations just
486as if nothing happened. The old processes will then restart listening to the
487ports and continue to accept connections. Not that this mechanism is system
Dan Lloyd8e48b872016-07-01 21:01:18 -0400488dependent and some operating systems may not support it in multi-process mode.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200489
490If the new process manages to bind correctly to all ports, then it sends either
491the SIGTERM (hard stop in case of "-st") or the SIGUSR1 (graceful stop in case
492of "-sf") to all processes to notify them that it is now in charge of operations
493and that the old processes will have to leave, either immediately or once they
494have finished their job.
495
496It is important to note that during this timeframe, there are two small windows
497of a few milliseconds each where it is possible that a few connection failures
498will be noticed during high loads. Typically observed failure rates are around
4991 failure during a reload operation every 10000 new connections per second,
500which means that a heavily loaded site running at 30000 new connections per
501second may see about 3 failed connection upon every reload. The two situations
502where this happens are :
503
504 - if the new process fails to bind due to the presence of the old process,
505 it will first have to go through the SIGTTOU+SIGTTIN sequence, which
506 typically lasts about one millisecond for a few tens of frontends, and
507 during which some ports will not be bound to the old process and not yet
508 bound to the new one. HAProxy works around this on systems that support the
509 SO_REUSEPORT socket options, as it allows the new process to bind without
510 first asking the old one to unbind. Most BSD systems have been supporting
511 this almost forever. Linux has been supporting this in version 2.0 and
512 dropped it around 2.2, but some patches were floating around by then. It
513 was reintroduced in kernel 3.9, so if you are observing a connection
Dan Lloyd8e48b872016-07-01 21:01:18 -0400514 failure rate above the one mentioned above, please ensure that your kernel
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200515 is 3.9 or newer, or that relevant patches were backported to your kernel
516 (less likely).
517
518 - when the old processes close the listening ports, the kernel may not always
519 redistribute any pending connection that was remaining in the socket's
520 backlog. Under high loads, a SYN packet may happen just before the socket
521 is closed, and will lead to an RST packet being sent to the client. In some
522 critical environments where even one drop is not acceptable, these ones are
523 sometimes dealt with using firewall rules to block SYN packets during the
524 reload, forcing the client to retransmit. This is totally system-dependent,
525 as some systems might be able to visit other listening queues and avoid
526 this RST. A second case concerns the ACK from the client on a local socket
527 that was in SYN_RECV state just before the close. This ACK will lead to an
528 RST packet while the haproxy process is still not aware of it. This one is
Dan Lloyd8e48b872016-07-01 21:01:18 -0400529 harder to get rid of, though the firewall filtering rules mentioned above
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200530 will work well if applied one second or so before restarting the process.
531
532For the vast majority of users, such drops will never ever happen since they
533don't have enough load to trigger the race conditions. And for most high traffic
534users, the failure rate is still fairly within the noise margin provided that at
535least SO_REUSEPORT is properly supported on their systems.
536
537
5385. File-descriptor limitations
539------------------------------
540
541In order to ensure that all incoming connections will successfully be served,
542HAProxy computes at load time the total number of file descriptors that will be
543needed during the process's life. A regular Unix process is generally granted
5441024 file descriptors by default, and a privileged process can raise this limit
545itself. This is one reason for starting HAProxy as root and letting it adjust
546the limit. The default limit of 1024 file descriptors roughly allow about 500
547concurrent connections to be processed. The computation is based on the global
548maxconn parameter which limits the total number of connections per process, the
549number of listeners, the number of servers which have a health check enabled,
550the agent checks, the peers, the loggers and possibly a few other technical
551requirements. A simple rough estimate of this number consists in simply
552doubling the maxconn value and adding a few tens to get the approximate number
553of file descriptors needed.
554
555Originally HAProxy did not know how to compute this value, and it was necessary
556to pass the value using the "ulimit-n" setting in the global section. This
557explains why even today a lot of configurations are seen with this setting
558present. Unfortunately it was often miscalculated resulting in connection
559failures when approaching maxconn instead of throttling incoming connection
560while waiting for the needed resources. For this reason it is important to
Dan Lloyd8e48b872016-07-01 21:01:18 -0400561remove any vestigial "ulimit-n" setting that can remain from very old versions.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200562
563Raising the number of file descriptors to accept even moderate loads is
564mandatory but comes with some OS-specific adjustments. First, the select()
565polling system is limited to 1024 file descriptors. In fact on Linux it used
566to be capable of handling more but since certain OS ship with excessively
567restrictive SELinux policies forbidding the use of select() with more than
5681024 file descriptors, HAProxy now refuses to start in this case in order to
569avoid any issue at run time. On all supported operating systems, poll() is
570available and will not suffer from this limitation. It is automatically picked
Dan Lloyd8e48b872016-07-01 21:01:18 -0400571so there is nothing to do to get a working configuration. But poll's becomes
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200572very slow when the number of file descriptors increases. While HAProxy does its
573best to limit this performance impact (eg: via the use of the internal file
574descriptor cache and batched processing), a good rule of thumb is that using
575poll() with more than a thousand concurrent connections will use a lot of CPU.
576
577For Linux systems base on kernels 2.6 and above, the epoll() system call will
578be used. It's a much more scalable mechanism relying on callbacks in the kernel
579that guarantee a constant wake up time regardless of the number of registered
580monitored file descriptors. It is automatically used where detected, provided
581that HAProxy had been built for one of the Linux flavors. Its presence and
582support can be verified using "haproxy -vv".
583
584For BSD systems which support it, kqueue() is available as an alternative. It
585is much faster than poll() and even slightly faster than epoll() thanks to its
586batched handling of changes. At least FreeBSD and OpenBSD support it. Just like
587with Linux's epoll(), its support and availability are reported in the output
588of "haproxy -vv".
589
590Having a good poller is one thing, but it is mandatory that the process can
591reach the limits. When HAProxy starts, it immediately sets the new process's
592file descriptor limits and verifies if it succeeds. In case of failure, it
593reports it before forking so that the administrator can see the problem. As
594long as the process is started by as root, there should be no reason for this
595setting to fail. However, it can fail if the process is started by an
596unprivileged user. If there is a compelling reason for *not* starting haproxy
597as root (eg: started by end users, or by a per-application account), then the
598file descriptor limit can be raised by the system administrator for this
599specific user. The effectiveness of the setting can be verified by issuing
600"ulimit -n" from the user's command line. It should reflect the new limit.
601
602Warning: when an unprivileged user's limits are changed in this user's account,
603it is fairly common that these values are only considered when the user logs in
604and not at all in some scripts run at system boot time nor in crontabs. This is
605totally dependent on the operating system, keep in mind to check "ulimit -n"
606before starting haproxy when running this way. The general advice is never to
607start haproxy as an unprivileged user for production purposes. Another good
608reason is that it prevents haproxy from enabling some security protections.
609
610Once it is certain that the system will allow the haproxy process to use the
611requested number of file descriptors, two new system-specific limits may be
612encountered. The first one is the system-wide file descriptor limit, which is
613the total number of file descriptors opened on the system, covering all
614processes. When this limit is reached, accept() or socket() will typically
615return ENFILE. The second one is the per-process hard limit on the number of
616file descriptors, it prevents setrlimit() from being set higher. Both are very
617dependent on the operating system. On Linux, the system limit is set at boot
618based on the amount of memory. It can be changed with the "fs.file-max" sysctl.
619And the per-process hard limit is set to 1048576 by default, but it can be
620changed using the "fs.nr_open" sysctl.
621
622File descriptor limitations may be observed on a running process when they are
623set too low. The strace utility will report that accept() and socket() return
624"-1 EMFILE" when the process's limits have been reached. In this case, simply
625raising the "ulimit-n" value (or removing it) will solve the problem. If these
626system calls return "-1 ENFILE" then it means that the kernel's limits have
627been reached and that something must be done on a system-wide parameter. These
628trouble must absolutely be addressed, as they result in high CPU usage (when
629accept() fails) and failed connections that are generally visible to the user.
630One solution also consists in lowering the global maxconn value to enforce
631serialization, and possibly to disable HTTP keep-alive to force connections
632to be released and reused faster.
633
634
6356. Memory management
636--------------------
637
638HAProxy uses a simple and fast pool-based memory management. Since it relies on
639a small number of different object types, it's much more efficient to pick new
640objects from a pool which already contains objects of the appropriate size than
641to call malloc() for each different size. The pools are organized as a stack or
642LIFO, so that newly allocated objects are taken from recently released objects
643still hot in the CPU caches. Pools of similar sizes are merged together, in
644order to limit memory fragmentation.
645
646By default, since the focus is set on performance, each released object is put
647back into the pool it came from, and allocated objects are never freed since
648they are expected to be reused very soon.
649
650On the CLI, it is possible to check how memory is being used in pools thanks to
651the "show pools" command :
652
653 > show pools
654 Dumping pools usage. Use SIGQUIT to flush them.
Willy Tarreau0a93b642018-10-16 07:58:39 +0200655 - Pool cache_st (16 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 1 users, @0x9ccc40=03 [SHARED]
656 - Pool pipe (32 bytes) : 5 allocated (160 bytes), 5 used, 0 failures, 2 users, @0x9ccac0=00 [SHARED]
657 - Pool comp_state (48 bytes) : 3 allocated (144 bytes), 3 used, 0 failures, 5 users, @0x9cccc0=04 [SHARED]
658 - Pool filter (64 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 3 users, @0x9ccbc0=02 [SHARED]
659 - Pool vars (80 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 2 users, @0x9ccb40=01 [SHARED]
660 - Pool uniqueid (128 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 2 users, @0x9cd240=15 [SHARED]
661 - Pool task (144 bytes) : 55 allocated (7920 bytes), 55 used, 0 failures, 1 users, @0x9cd040=11 [SHARED]
662 - Pool session (160 bytes) : 1 allocated (160 bytes), 1 used, 0 failures, 1 users, @0x9cd140=13 [SHARED]
663 - Pool h2s (208 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 2 users, @0x9ccec0=08 [SHARED]
664 - Pool h2c (288 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 1 users, @0x9cce40=07 [SHARED]
665 - Pool spoe_ctx (304 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 2 users, @0x9ccf40=09 [SHARED]
666 - Pool connection (400 bytes) : 2 allocated (800 bytes), 2 used, 0 failures, 1 users, @0x9cd1c0=14 [SHARED]
667 - Pool hdr_idx (416 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 1 users, @0x9cd340=17 [SHARED]
668 - Pool dns_resolut (480 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 1 users, @0x9ccdc0=06 [SHARED]
669 - Pool dns_answer_ (576 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 1 users, @0x9ccd40=05 [SHARED]
670 - Pool stream (960 bytes) : 1 allocated (960 bytes), 1 used, 0 failures, 1 users, @0x9cd0c0=12 [SHARED]
671 - Pool requri (1024 bytes) : 0 allocated (0 bytes), 0 used, 0 failures, 1 users, @0x9cd2c0=16 [SHARED]
672 - Pool buffer (8030 bytes) : 3 allocated (24090 bytes), 2 used, 0 failures, 1 users, @0x9cd3c0=18 [SHARED]
673 - Pool trash (8062 bytes) : 1 allocated (8062 bytes), 1 used, 0 failures, 1 users, @0x9cd440=19
674 Total: 19 pools, 42296 bytes allocated, 34266 used.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200675
676The pool name is only indicative, it's the name of the first object type using
677this pool. The size in parenthesis is the object size for objects in this pool.
678Object sizes are always rounded up to the closest multiple of 16 bytes. The
679number of objects currently allocated and the equivalent number of bytes is
680reported so that it is easy to know which pool is responsible for the highest
681memory usage. The number of objects currently in use is reported as well in the
682"used" field. The difference between "allocated" and "used" corresponds to the
Willy Tarreau0a93b642018-10-16 07:58:39 +0200683objects that have been freed and are available for immediate use. The address
684at the end of the line is the pool's address, and the following number is the
685pool index when it exists, or is reported as -1 if no index was assigned.
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200686
687It is possible to limit the amount of memory allocated per process using the
688"-m" command line option, followed by a number of megabytes. It covers all of
689the process's addressable space, so that includes memory used by some libraries
690as well as the stack, but it is a reliable limit when building a resource
691constrained system. It works the same way as "ulimit -v" on systems which have
692it, or "ulimit -d" for the other ones.
693
694If a memory allocation fails due to the memory limit being reached or because
695the system doesn't have any enough memory, then haproxy will first start to
696free all available objects from all pools before attempting to allocate memory
697again. This mechanism of releasing unused memory can be triggered by sending
698the signal SIGQUIT to the haproxy process. When doing so, the pools state prior
699to the flush will also be reported to stderr when the process runs in
700foreground.
701
702During a reload operation, the process switched to the graceful stop state also
703automatically performs some flushes after releasing any connection so that all
704possible memory is released to save it for the new process.
705
706
7077. CPU usage
708------------
709
710HAProxy normally spends most of its time in the system and a smaller part in
711userland. A finely tuned 3.5 GHz CPU can sustain a rate about 80000 end-to-end
712connection setups and closes per second at 100% CPU on a single core. When one
713core is saturated, typical figures are :
714 - 95% system, 5% user for long TCP connections or large HTTP objects
715 - 85% system and 15% user for short TCP connections or small HTTP objects in
716 close mode
717 - 70% system and 30% user for small HTTP objects in keep-alive mode
718
719The amount of rules processing and regular expressions will increase the user
720land part. The presence of firewall rules, connection tracking, complex routing
721tables in the system will instead increase the system part.
722
723On most systems, the CPU time observed during network transfers can be cut in 4
724parts :
725 - the interrupt part, which concerns all the processing performed upon I/O
726 receipt, before the target process is even known. Typically Rx packets are
727 accounted for in interrupt. On some systems such as Linux where interrupt
728 processing may be deferred to a dedicated thread, it can appear as softirq,
729 and the thread is called ksoftirqd/0 (for CPU 0). The CPU taking care of
730 this load is generally defined by the hardware settings, though in the case
731 of softirq it is often possible to remap the processing to another CPU.
732 This interrupt part will often be perceived as parasitic since it's not
733 associated with any process, but it actually is some processing being done
734 to prepare the work for the process.
735
736 - the system part, which concerns all the processing done using kernel code
737 called from userland. System calls are accounted as system for example. All
738 synchronously delivered Tx packets will be accounted for as system time. If
739 some packets have to be deferred due to queues filling up, they may then be
740 processed in interrupt context later (eg: upon receipt of an ACK opening a
741 TCP window).
742
743 - the user part, which exclusively runs application code in userland. HAProxy
744 runs exclusively in this part, though it makes heavy use of system calls.
745 Rules processing, regular expressions, compression, encryption all add to
746 the user portion of CPU consumption.
747
748 - the idle part, which is what the CPU does when there is nothing to do. For
749 example HAProxy waits for an incoming connection, or waits for some data to
750 leave, meaning the system is waiting for an ACK from the client to push
751 these data.
752
753In practice regarding HAProxy's activity, it is in general reasonably accurate
754(but totally inexact) to consider that interrupt/softirq are caused by Rx
755processing in kernel drivers, that user-land is caused by layer 7 processing
756in HAProxy, and that system time is caused by network processing on the Tx
757path.
758
759Since HAProxy runs around an event loop, it waits for new events using poll()
760(or any alternative) and processes all these events as fast as possible before
761going back to poll() waiting for new events. It measures the time spent waiting
762in poll() compared to the time spent doing processing events. The ratio of
763polling time vs total time is called the "idle" time, it's the amount of time
764spent waiting for something to happen. This ratio is reported in the stats page
765on the "idle" line, or "Idle_pct" on the CLI. When it's close to 100%, it means
766the load is extremely low. When it's close to 0%, it means that there is
767constantly some activity. While it cannot be very accurate on an overloaded
768system due to other processes possibly preempting the CPU from the haproxy
769process, it still provides a good estimate about how HAProxy considers it is
770working : if the load is low and the idle ratio is low as well, it may indicate
771that HAProxy has a lot of work to do, possibly due to very expensive rules that
772have to be processed. Conversely, if HAProxy indicates the idle is close to
773100% while things are slow, it means that it cannot do anything to speed things
774up because it is already waiting for incoming data to process. In the example
775below, haproxy is completely idle :
776
777 $ echo "show info" | socat - /var/run/haproxy.sock | grep ^Idle
778 Idle_pct: 100
779
780When the idle ratio starts to become very low, it is important to tune the
781system and place processes and interrupts correctly to save the most possible
782CPU resources for all tasks. If a firewall is present, it may be worth trying
783to disable it or to tune it to ensure it is not responsible for a large part
784of the performance limitation. It's worth noting that unloading a stateful
785firewall generally reduces both the amount of interrupt/softirq and of system
786usage since such firewalls act both on the Rx and the Tx paths. On Linux,
787unloading the nf_conntrack and ip_conntrack modules will show whether there is
788anything to gain. If so, then the module runs with default settings and you'll
789have to figure how to tune it for better performance. In general this consists
790in considerably increasing the hash table size. On FreeBSD, "pfctl -d" will
791disable the "pf" firewall and its stateful engine at the same time.
792
793If it is observed that a lot of time is spent in interrupt/softirq, it is
794important to ensure that they don't run on the same CPU. Most systems tend to
795pin the tasks on the CPU where they receive the network traffic because for
796certain workloads it improves things. But with heavily network-bound workloads
797it is the opposite as the haproxy process will have to fight against its kernel
798counterpart. Pinning haproxy to one CPU core and the interrupts to another one,
799all sharing the same L3 cache tends to sensibly increase network performance
800because in practice the amount of work for haproxy and the network stack are
801quite close, so they can almost fill an entire CPU each. On Linux this is done
802using taskset (for haproxy) or using cpu-map (from the haproxy config), and the
803interrupts are assigned under /proc/irq. Many network interfaces support
804multiple queues and multiple interrupts. In general it helps to spread them
805across a small number of CPU cores provided they all share the same L3 cache.
806Please always stop irq_balance which always does the worst possible thing on
807such workloads.
808
809For CPU-bound workloads consisting in a lot of SSL traffic or a lot of
810compression, it may be worth using multiple processes dedicated to certain
811tasks, though there is no universal rule here and experimentation will have to
812be performed.
813
814In order to increase the CPU capacity, it is possible to make HAProxy run as
815several processes, using the "nbproc" directive in the global section. There
816are some limitations though :
817 - health checks are run per process, so the target servers will get as many
818 checks as there are running processes ;
819 - maxconn values and queues are per-process so the correct value must be set
820 to avoid overloading the servers ;
821 - outgoing connections should avoid using port ranges to avoid conflicts
822 - stick-tables are per process and are not shared between processes ;
823 - each peers section may only run on a single process at a time ;
824 - the CLI operations will only act on a single process at a time.
825
826With this in mind, it appears that the easiest setup often consists in having
827one first layer running on multiple processes and in charge for the heavy
828processing, passing the traffic to a second layer running in a single process.
829This mechanism is suited to SSL and compression which are the two CPU-heavy
830features. Instances can easily be chained over UNIX sockets (which are cheaper
fengpeiyuancc123c62016-01-15 16:40:53 +0800831than TCP sockets and which do not waste ports), and the proxy protocol which is
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200832useful to pass client information to the next stage. When doing so, it is
833generally a good idea to bind all the single-process tasks to process number 1
834and extra tasks to next processes, as this will make it easier to generate
835similar configurations for different machines.
836
837On Linux versions 3.9 and above, running HAProxy in multi-process mode is much
838more efficient when each process uses a distinct listening socket on the same
839IP:port ; this will make the kernel evenly distribute the load across all
840processes instead of waking them all up. Please check the "process" option of
841the "bind" keyword lines in the configuration manual for more information.
842
843
8448. Logging
845----------
846
847For logging, HAProxy always relies on a syslog server since it does not perform
848any file-system access. The standard way of using it is to send logs over UDP
849to the log server (by default on port 514). Very commonly this is configured to
850127.0.0.1 where the local syslog daemon is running, but it's also used over the
851network to log to a central server. The central server provides additional
852benefits especially in active-active scenarios where it is desirable to keep
853the logs merged in arrival order. HAProxy may also make use of a UNIX socket to
854send its logs to the local syslog daemon, but it is not recommended at all,
855because if the syslog server is restarted while haproxy runs, the socket will
856be replaced and new logs will be lost. Since HAProxy will be isolated inside a
857chroot jail, it will not have the ability to reconnect to the new socket. It
858has also been observed in field that the log buffers in use on UNIX sockets are
859very small and lead to lost messages even at very light loads. But this can be
860fine for testing however.
861
862It is recommended to add the following directive to the "global" section to
863make HAProxy log to the local daemon using facility "local0" :
864
865 log 127.0.0.1:514 local0
866
867and then to add the following one to each "defaults" section or to each frontend
868and backend section :
869
870 log global
871
872This way, all logs will be centralized through the global definition of where
873the log server is.
874
875Some syslog daemons do not listen to UDP traffic by default, so depending on
876the daemon being used, the syntax to enable this will vary :
877
878 - on sysklogd, you need to pass argument "-r" on the daemon's command line
879 so that it listens to a UDP socket for "remote" logs ; note that there is
880 no way to limit it to address 127.0.0.1 so it will also receive logs from
881 remote systems ;
882
883 - on rsyslogd, the following lines must be added to the configuration file :
884
885 $ModLoad imudp
886 $UDPServerAddress *
887 $UDPServerRun 514
888
889 - on syslog-ng, a new source can be created the following way, it then needs
890 to be added as a valid source in one of the "log" directives :
891
892 source s_udp {
893 udp(ip(127.0.0.1) port(514));
894 };
895
896Please consult your syslog daemon's manual for more information. If no logs are
897seen in the system's log files, please consider the following tests :
898
899 - restart haproxy. Each frontend and backend logs one line indicating it's
900 starting. If these logs are received, it means logs are working.
901
902 - run "strace -tt -s100 -etrace=sendmsg -p <haproxy's pid>" and perform some
903 activity that you expect to be logged. You should see the log messages
904 being sent using sendmsg() there. If they don't appear, restart using
905 strace on top of haproxy. If you still see no logs, it definitely means
906 that something is wrong in your configuration.
907
908 - run tcpdump to watch for port 514, for example on the loopback interface if
909 the traffic is being sent locally : "tcpdump -As0 -ni lo port 514". If the
910 packets are seen there, it's the proof they're sent then the syslogd daemon
911 needs to be troubleshooted.
912
913While traffic logs are sent from the frontends (where the incoming connections
914are accepted), backends also need to be able to send logs in order to report a
915server state change consecutive to a health check. Please consult HAProxy's
916configuration manual for more information regarding all possible log settings.
917
Dan Lloyd8e48b872016-07-01 21:01:18 -0400918It is convenient to chose a facility that is not used by other daemons. HAProxy
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200919examples often suggest "local0" for traffic logs and "local1" for admin logs
920because they're never seen in field. A single facility would be enough as well.
921Having separate logs is convenient for log analysis, but it's also important to
922remember that logs may sometimes convey confidential information, and as such
Dan Lloyd8e48b872016-07-01 21:01:18 -0400923they must not be mixed with other logs that may accidentally be handed out to
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200924unauthorized people.
925
926For in-field troubleshooting without impacting the server's capacity too much,
927it is recommended to make use of the "halog" utility provided with HAProxy.
928This is sort of a grep-like utility designed to process HAProxy log files at
929a very fast data rate. Typical figures range between 1 and 2 GB of logs per
930second. It is capable of extracting only certain logs (eg: search for some
931classes of HTTP status codes, connection termination status, search by response
932time ranges, look for errors only), count lines, limit the output to a number
933of lines, and perform some more advanced statistics such as sorting servers
934by response time or error counts, sorting URLs by time or count, sorting client
935addresses by access count, and so on. It is pretty convenient to quickly spot
936anomalies such as a bot looping on the site, and block them.
937
938
9399. Statistics and monitoring
940----------------------------
941
Willy Tarreau44aed902015-10-13 14:45:29 +0200942It is possible to query HAProxy about its status. The most commonly used
943mechanism is the HTTP statistics page. This page also exposes an alternative
944CSV output format for monitoring tools. The same format is provided on the
945Unix socket.
946
947
9489.1. CSV format
949---------------
950
951The statistics may be consulted either from the unix socket or from the HTTP
952page. Both means provide a CSV format whose fields follow. The first line
953begins with a sharp ('#') and has one word per comma-delimited field which
954represents the title of the column. All other lines starting at the second one
955use a classical CSV format using a comma as the delimiter, and the double quote
956('"') as an optional text delimiter, but only if the enclosed text is ambiguous
957(if it contains a quote or a comma). The double-quote character ('"') in the
958text is doubled ('""'), which is the format that most tools recognize. Please
959do not insert any column before these ones in order not to break tools which
960use hard-coded column positions.
961
962In brackets after each field name are the types which may have a value for
963that field. The types are L (Listeners), F (Frontends), B (Backends), and
964S (Servers).
965
966 0. pxname [LFBS]: proxy name
967 1. svname [LFBS]: service name (FRONTEND for frontend, BACKEND for backend,
968 any name for server/listener)
969 2. qcur [..BS]: current queued requests. For the backend this reports the
970 number queued without a server assigned.
971 3. qmax [..BS]: max value of qcur
972 4. scur [LFBS]: current sessions
973 5. smax [LFBS]: max sessions
974 6. slim [LFBS]: configured session limit
Willy Tarreauc73810f2016-01-11 13:52:04 +0100975 7. stot [LFBS]: cumulative number of sessions
Willy Tarreau44aed902015-10-13 14:45:29 +0200976 8. bin [LFBS]: bytes in
977 9. bout [LFBS]: bytes out
978 10. dreq [LFB.]: requests denied because of security concerns.
979 - For tcp this is because of a matched tcp-request content rule.
980 - For http this is because of a matched http-request or tarpit rule.
981 11. dresp [LFBS]: responses denied because of security concerns.
982 - For http this is because of a matched http-request rule, or
983 "option checkcache".
984 12. ereq [LF..]: request errors. Some of the possible causes are:
985 - early termination from the client, before the request has been sent.
986 - read error from the client
987 - client timeout
988 - client closed connection
989 - various bad requests from the client.
990 - request was tarpitted.
991 13. econ [..BS]: number of requests that encountered an error trying to
992 connect to a backend server. The backend stat is the sum of the stat
993 for all servers of that backend, plus any connection errors not
994 associated with a particular server (such as the backend having no
995 active servers).
996 14. eresp [..BS]: response errors. srv_abrt will be counted here also.
997 Some other errors are:
998 - write error on the client socket (won't be counted for the server stat)
999 - failure applying filters to the response.
1000 15. wretr [..BS]: number of times a connection to a server was retried.
1001 16. wredis [..BS]: number of times a request was redispatched to another
1002 server. The server value counts the number of times that server was
1003 switched away from.
Willy Tarreaub96dd282016-11-09 14:45:51 +01001004 17. status [LFBS]: status (UP/DOWN/NOLB/MAINT/MAINT(via)/MAINT(resolution)...)
Willy Tarreau44aed902015-10-13 14:45:29 +02001005 18. weight [..BS]: total weight (backend), server weight (server)
1006 19. act [..BS]: number of active servers (backend), server is active (server)
1007 20. bck [..BS]: number of backup servers (backend), server is backup (server)
1008 21. chkfail [...S]: number of failed checks. (Only counts checks failed when
1009 the server is up.)
1010 22. chkdown [..BS]: number of UP->DOWN transitions. The backend counter counts
1011 transitions to the whole backend being down, rather than the sum of the
1012 counters for each server.
1013 23. lastchg [..BS]: number of seconds since the last UP<->DOWN transition
1014 24. downtime [..BS]: total downtime (in seconds). The value for the backend
1015 is the downtime for the whole backend, not the sum of the server downtime.
1016 25. qlimit [...S]: configured maxqueue for the server, or nothing in the
1017 value is 0 (default, meaning no limit)
1018 26. pid [LFBS]: process id (0 for first instance, 1 for second, ...)
1019 27. iid [LFBS]: unique proxy id
1020 28. sid [L..S]: server id (unique inside a proxy)
1021 29. throttle [...S]: current throttle percentage for the server, when
1022 slowstart is active, or no value if not in slowstart.
1023 30. lbtot [..BS]: total number of times a server was selected, either for new
1024 sessions, or when re-dispatching. The server counter is the number
1025 of times that server was selected.
1026 31. tracked [...S]: id of proxy/server if tracking is enabled.
1027 32. type [LFBS]: (0=frontend, 1=backend, 2=server, 3=socket/listener)
1028 33. rate [.FBS]: number of sessions per second over last elapsed second
1029 34. rate_lim [.F..]: configured limit on new sessions per second
1030 35. rate_max [.FBS]: max number of new sessions per second
1031 36. check_status [...S]: status of last health check, one of:
1032 UNK -> unknown
1033 INI -> initializing
1034 SOCKERR -> socket error
1035 L4OK -> check passed on layer 4, no upper layers testing enabled
1036 L4TOUT -> layer 1-4 timeout
1037 L4CON -> layer 1-4 connection problem, for example
1038 "Connection refused" (tcp rst) or "No route to host" (icmp)
1039 L6OK -> check passed on layer 6
1040 L6TOUT -> layer 6 (SSL) timeout
1041 L6RSP -> layer 6 invalid response - protocol error
1042 L7OK -> check passed on layer 7
1043 L7OKC -> check conditionally passed on layer 7, for example 404 with
1044 disable-on-404
1045 L7TOUT -> layer 7 (HTTP/SMTP) timeout
1046 L7RSP -> layer 7 invalid response - protocol error
1047 L7STS -> layer 7 response error, for example HTTP 5xx
Daniel Schnellerb6c8b0d2017-09-01 19:13:55 +02001048 Notice: If a check is currently running, the last known status will be
1049 reported, prefixed with "* ". e. g. "* L7OK".
Willy Tarreau44aed902015-10-13 14:45:29 +02001050 37. check_code [...S]: layer5-7 code, if available
1051 38. check_duration [...S]: time in ms took to finish last health check
1052 39. hrsp_1xx [.FBS]: http responses with 1xx code
1053 40. hrsp_2xx [.FBS]: http responses with 2xx code
1054 41. hrsp_3xx [.FBS]: http responses with 3xx code
1055 42. hrsp_4xx [.FBS]: http responses with 4xx code
1056 43. hrsp_5xx [.FBS]: http responses with 5xx code
1057 44. hrsp_other [.FBS]: http responses with other codes (protocol error)
1058 45. hanafail [...S]: failed health checks details
1059 46. req_rate [.F..]: HTTP requests per second over last elapsed second
1060 47. req_rate_max [.F..]: max number of HTTP requests per second observed
Willy Tarreaufb981bd2016-12-12 14:31:46 +01001061 48. req_tot [.FB.]: total number of HTTP requests received
Willy Tarreau44aed902015-10-13 14:45:29 +02001062 49. cli_abrt [..BS]: number of data transfers aborted by the client
1063 50. srv_abrt [..BS]: number of data transfers aborted by the server
1064 (inc. in eresp)
1065 51. comp_in [.FB.]: number of HTTP response bytes fed to the compressor
1066 52. comp_out [.FB.]: number of HTTP response bytes emitted by the compressor
1067 53. comp_byp [.FB.]: number of bytes that bypassed the HTTP compressor
1068 (CPU/BW limit)
1069 54. comp_rsp [.FB.]: number of HTTP responses that were compressed
1070 55. lastsess [..BS]: number of seconds since last session assigned to
1071 server/backend
1072 56. last_chk [...S]: last health check contents or textual error
1073 57. last_agt [...S]: last agent check contents or textual error
1074 58. qtime [..BS]: the average queue time in ms over the 1024 last requests
1075 59. ctime [..BS]: the average connect time in ms over the 1024 last requests
1076 60. rtime [..BS]: the average response time in ms over the 1024 last requests
1077 (0 for TCP)
1078 61. ttime [..BS]: the average total session time in ms over the 1024 last
1079 requests
Willy Tarreau7f618842016-01-08 11:40:03 +01001080 62. agent_status [...S]: status of last agent check, one of:
1081 UNK -> unknown
1082 INI -> initializing
1083 SOCKERR -> socket error
1084 L4OK -> check passed on layer 4, no upper layers testing enabled
1085 L4TOUT -> layer 1-4 timeout
1086 L4CON -> layer 1-4 connection problem, for example
1087 "Connection refused" (tcp rst) or "No route to host" (icmp)
1088 L7OK -> agent reported "up"
1089 L7STS -> agent reported "fail", "stop", or "down"
1090 63. agent_code [...S]: numeric code reported by agent if any (unused for now)
1091 64. agent_duration [...S]: time in ms taken to finish last check
Willy Tarreaudd7354b2016-01-08 13:47:26 +01001092 65. check_desc [...S]: short human-readable description of check_status
1093 66. agent_desc [...S]: short human-readable description of agent_status
Willy Tarreau3141f592016-01-08 14:25:28 +01001094 67. check_rise [...S]: server's "rise" parameter used by checks
1095 68. check_fall [...S]: server's "fall" parameter used by checks
1096 69. check_health [...S]: server's health check value between 0 and rise+fall-1
1097 70. agent_rise [...S]: agent's "rise" parameter, normally 1
1098 71. agent_fall [...S]: agent's "fall" parameter, normally 1
1099 72. agent_health [...S]: agent's health parameter, between 0 and rise+fall-1
Willy Tarreaua6f5a732016-01-08 16:59:56 +01001100 73. addr [L..S]: address:port or "unix". IPv6 has brackets around the address.
Willy Tarreaue4847c62016-01-08 15:43:54 +01001101 74: cookie [..BS]: server's cookie value or backend's cookie name
Willy Tarreauf8211df2016-01-11 14:09:38 +01001102 75: mode [LFBS]: proxy mode (tcp, http, health, unknown)
Willy Tarreauf1516d92016-01-11 14:48:36 +01001103 76: algo [..B.]: load balancing algorithm
Willy Tarreauc73810f2016-01-11 13:52:04 +01001104 77: conn_rate [.F..]: number of connections over the last elapsed second
1105 78: conn_rate_max [.F..]: highest known conn_rate
1106 79: conn_tot [.F..]: cumulative number of connections
Willy Tarreau5b9bdff2016-01-11 14:40:47 +01001107 80: intercepted [.FB.]: cum. number of intercepted requests (monitor, stats)
Willy Tarreau8a90b8e2016-10-21 18:15:32 +02001108 81: dcon [LF..]: requests denied by "tcp-request connection" rules
Willy Tarreaua5bc36b2016-10-21 18:16:27 +02001109 82: dses [LF..]: requests denied by "tcp-request session" rules
Willy Tarreauea96a822018-05-28 15:15:43 +02001110 83: wrew [LFBS]: cumulative number of failed header rewriting warnings
Jérôme Magnin708eb882019-07-17 09:24:46 +02001111 84: connect [..BS]: cumulative number of connection establishment attempts
1112 85: reuse [..BS]: cumulative number of connection reuses
Willy Tarreau72974292019-11-08 07:29:34 +01001113 86: cache_lookups [.FB.]: cumulative number of cache lookups
Jérôme Magnin34ebb5c2019-07-17 14:04:40 +02001114 87: cache_hits [.FB.]: cumulative number of cache hits
Christopher Faulet2ac25742019-11-08 15:27:27 +01001115 88: srv_icur [...S]: current number of idle connections available for reuse
1116 89: src_ilim [...S]: limit on the number of available idle connections
1117 90. qtime_max [..BS]: the maximum observed queue time in ms
1118 91. ctime_max [..BS]: the maximum observed connect time in ms
1119 92. rtime_max [..BS]: the maximum observed response time in ms (0 for TCP)
1120 93. ttime_max [..BS]: the maximum observed total session time in ms
Christopher Faulet0159ee42019-12-16 14:40:39 +01001121 94. eint [LFBS]: cumulative number of internal errors
Willy Tarreau44aed902015-10-13 14:45:29 +02001122
1123
Willy Tarreau5d8b9792016-03-11 11:09:34 +010011249.2) Typed output format
1125------------------------
1126
1127Both "show info" and "show stat" support a mode where each output value comes
1128with its type and sufficient information to know how the value is supposed to
1129be aggregated between processes and how it evolves.
1130
1131In all cases, the output consists in having a single value per line with all
1132the information split into fields delimited by colons (':').
1133
1134The first column designates the object or metric being dumped. Its format is
1135specific to the command producing this output and will not be described in this
1136section. Usually it will consist in a series of identifiers and field names.
1137
1138The second column contains 3 characters respectively indicating the origin, the
1139nature and the scope of the value being reported. The first character (the
1140origin) indicates where the value was extracted from. Possible characters are :
1141
1142 M The value is a metric. It is valid at one instant any may change depending
1143 on its nature .
1144
1145 S The value is a status. It represents a discrete value which by definition
1146 cannot be aggregated. It may be the status of a server ("UP" or "DOWN"),
1147 the PID of the process, etc.
1148
1149 K The value is a sorting key. It represents an identifier which may be used
1150 to group some values together because it is unique among its class. All
1151 internal identifiers are keys. Some names can be listed as keys if they
1152 are unique (eg: a frontend name is unique). In general keys come from the
Dan Lloyd8e48b872016-07-01 21:01:18 -04001153 configuration, even though some of them may automatically be assigned. For
Willy Tarreau5d8b9792016-03-11 11:09:34 +01001154 most purposes keys may be considered as equivalent to configuration.
1155
1156 C The value comes from the configuration. Certain configuration values make
1157 sense on the output, for example a concurrent connection limit or a cookie
1158 name. By definition these values are the same in all processes started
1159 from the same configuration file.
1160
1161 P The value comes from the product itself. There are very few such values,
1162 most common use is to report the product name, version and release date.
1163 These elements are also the same between all processes.
1164
1165The second character (the nature) indicates the nature of the information
1166carried by the field in order to let an aggregator decide on what operation to
1167use to aggregate multiple values. Possible characters are :
1168
1169 A The value represents an age since a last event. This is a bit different
1170 from the duration in that an age is automatically computed based on the
1171 current date. A typical example is how long ago did the last session
1172 happen on a server. Ages are generally aggregated by taking the minimum
1173 value and do not need to be stored.
1174
1175 a The value represents an already averaged value. The average response times
1176 and server weights are of this nature. Averages can typically be averaged
1177 between processes.
1178
1179 C The value represents a cumulative counter. Such measures perpetually
1180 increase until they wrap around. Some monitoring protocols need to tell
1181 the difference between a counter and a gauge to report a different type.
1182 In general counters may simply be summed since they represent events or
1183 volumes. Examples of metrics of this nature are connection counts or byte
1184 counts.
1185
1186 D The value represents a duration for a status. There are a few usages of
1187 this, most of them include the time taken by the last health check and
1188 the time a server has spent down. Durations are generally not summed,
1189 most of the time the maximum will be retained to compute an SLA.
1190
1191 G The value represents a gauge. It's a measure at one instant. The memory
1192 usage or the current number of active connections are of this nature.
1193 Metrics of this type are typically summed during aggregation.
1194
1195 L The value represents a limit (generally a configured one). By nature,
1196 limits are harder to aggregate since they are specific to the point where
1197 they were retrieved. In certain situations they may be summed or be kept
1198 separate.
1199
1200 M The value represents a maximum. In general it will apply to a gauge and
1201 keep the highest known value. An example of such a metric could be the
1202 maximum amount of concurrent connections that was encountered in the
1203 product's life time. To correctly aggregate maxima, you are supposed to
1204 output a range going from the maximum of all maxima and the sum of all
1205 of them. There is indeed no way to know if they were encountered
1206 simultaneously or not.
1207
1208 m The value represents a minimum. In general it will apply to a gauge and
1209 keep the lowest known value. An example of such a metric could be the
1210 minimum amount of free memory pools that was encountered in the product's
1211 life time. To correctly aggregate minima, you are supposed to output a
1212 range going from the minimum of all minima and the sum of all of them.
1213 There is indeed no way to know if they were encountered simultaneously
1214 or not.
1215
1216 N The value represents a name, so it is a string. It is used to report
1217 proxy names, server names and cookie names. Names have configuration or
1218 keys as their origin and are supposed to be the same among all processes.
1219
1220 O The value represents a free text output. Outputs from various commands,
1221 returns from health checks, node descriptions are of such nature.
1222
1223 R The value represents an event rate. It's a measure at one instant. It is
1224 quite similar to a gauge except that the recipient knows that this measure
1225 moves slowly and may decide not to keep all values. An example of such a
1226 metric is the measured amount of connections per second. Metrics of this
1227 type are typically summed during aggregation.
1228
1229 T The value represents a date or time. A field emitting the current date
1230 would be of this type. The method to aggregate such information is left
1231 as an implementation choice. For now no field uses this type.
1232
1233The third character (the scope) indicates what extent the value reflects. Some
1234elements may be per process while others may be per configuration or per system.
1235The distinction is important to know whether or not a single value should be
1236kept during aggregation or if values have to be aggregated. The following
1237characters are currently supported :
1238
1239 C The value is valid for a whole cluster of nodes, which is the set of nodes
1240 communicating over the peers protocol. An example could be the amount of
1241 entries present in a stick table that is replicated with other peers. At
1242 the moment no metric use this scope.
1243
1244 P The value is valid only for the process reporting it. Most metrics use
1245 this scope.
1246
1247 S The value is valid for the whole service, which is the set of processes
1248 started together from the same configuration file. All metrics originating
1249 from the configuration use this scope. Some other metrics may use it as
1250 well for some shared resources (eg: shared SSL cache statistics).
1251
1252 s The value is valid for the whole system, such as the system's hostname,
1253 current date or resource usage. At the moment this scope is not used by
1254 any metric.
1255
1256Consumers of these information will generally have enough of these 3 characters
1257to determine how to accurately report aggregated information across multiple
1258processes.
1259
1260After this column, the third column indicates the type of the field, among "s32"
1261(signed 32-bit integer), "s64" (signed 64-bit integer), "u32" (unsigned 32-bit
1262integer), "u64" (unsigned 64-bit integer), "str" (string). It is important to
1263know the type before parsing the value in order to properly read it. For example
1264a string containing only digits is still a string an not an integer (eg: an
1265error code extracted by a check).
1266
1267Then the fourth column is the value itself, encoded according to its type.
1268Strings are dumped as-is immediately after the colon without any leading space.
1269If a string contains a colon, it will appear normally. This means that the
1270output should not be exclusively split around colons or some check outputs
1271or server addresses might be truncated.
1272
1273
12749.3. Unix Socket commands
Willy Tarreau44aed902015-10-13 14:45:29 +02001275-------------------------
1276
1277The stats socket is not enabled by default. In order to enable it, it is
1278necessary to add one line in the global section of the haproxy configuration.
1279A second line is recommended to set a larger timeout, always appreciated when
1280issuing commands by hand :
1281
1282 global
1283 stats socket /var/run/haproxy.sock mode 600 level admin
1284 stats timeout 2m
1285
1286It is also possible to add multiple instances of the stats socket by repeating
1287the line, and make them listen to a TCP port instead of a UNIX socket. This is
1288never done by default because this is dangerous, but can be handy in some
1289situations :
1290
1291 global
1292 stats socket /var/run/haproxy.sock mode 600 level admin
1293 stats socket ipv4@192.168.0.1:9999 level admin
1294 stats timeout 2m
1295
1296To access the socket, an external utility such as "socat" is required. Socat is
1297a swiss-army knife to connect anything to anything. We use it to connect
1298terminals to the socket, or a couple of stdin/stdout pipes to it for scripts.
1299The two main syntaxes we'll use are the following :
1300
1301 # socat /var/run/haproxy.sock stdio
1302 # socat /var/run/haproxy.sock readline
1303
1304The first one is used with scripts. It is possible to send the output of a
1305script to haproxy, and pass haproxy's output to another script. That's useful
1306for retrieving counters or attack traces for example.
1307
1308The second one is only useful for issuing commands by hand. It has the benefit
1309that the terminal is handled by the readline library which supports line
1310editing and history, which is very convenient when issuing repeated commands
1311(eg: watch a counter).
1312
1313The socket supports two operation modes :
1314 - interactive
1315 - non-interactive
1316
1317The non-interactive mode is the default when socat connects to the socket. In
1318this mode, a single line may be sent. It is processed as a whole, responses are
1319sent back, and the connection closes after the end of the response. This is the
1320mode that scripts and monitoring tools use. It is possible to send multiple
1321commands in this mode, they need to be delimited by a semi-colon (';'). For
1322example :
1323
1324 # echo "show info;show stat;show table" | socat /var/run/haproxy stdio
1325
Dragan Dosena1c35ab2016-11-24 11:33:12 +01001326If a command needs to use a semi-colon or a backslash (eg: in a value), it
Joseph Herlant71b4b152018-11-13 16:55:16 -08001327must be preceded by a backslash ('\').
Chad Lavoiee3f50312016-05-26 16:42:25 -04001328
Willy Tarreau44aed902015-10-13 14:45:29 +02001329The interactive mode displays a prompt ('>') and waits for commands to be
1330entered on the line, then processes them, and displays the prompt again to wait
1331for a new command. This mode is entered via the "prompt" command which must be
1332sent on the first line in non-interactive mode. The mode is a flip switch, if
1333"prompt" is sent in interactive mode, it is disabled and the connection closes
1334after processing the last command of the same line.
1335
1336For this reason, when debugging by hand, it's quite common to start with the
1337"prompt" command :
1338
1339 # socat /var/run/haproxy readline
1340 prompt
1341 > show info
1342 ...
1343 >
1344
1345Since multiple commands may be issued at once, haproxy uses the empty line as a
1346delimiter to mark an end of output for each command, and takes care of ensuring
1347that no command can emit an empty line on output. A script can thus easily
1348parse the output even when multiple commands were pipelined on a single line.
1349
Aurélien Nephtaliabbf6072018-04-18 13:26:46 +02001350Some commands may take an optional payload. To add one to a command, the first
1351line needs to end with the "<<\n" pattern. The next lines will be treated as
1352the payload and can contain as many lines as needed. To validate a command with
1353a payload, it needs to end with an empty line.
1354
1355Limitations do exist: the length of the whole buffer passed to the CLI must
1356not be greater than tune.bfsize and the pattern "<<" must not be glued to the
1357last word of the line.
1358
1359When entering a paylod while in interactive mode, the prompt will change from
1360"> " to "+ ".
1361
Willy Tarreau44aed902015-10-13 14:45:29 +02001362It is important to understand that when multiple haproxy processes are started
1363on the same sockets, any process may pick up the request and will output its
1364own stats.
1365
1366The list of commands currently supported on the stats socket is provided below.
1367If an unknown command is sent, haproxy displays the usage message which reminds
1368all supported commands. Some commands support a more complex syntax, generally
1369it will explain what part of the command is invalid when this happens.
1370
Olivier Doucetd8703e82017-08-31 11:05:10 +02001371Some commands require a higher level of privilege to work. If you do not have
1372enough privilege, you will get an error "Permission denied". Please check
1373the "level" option of the "bind" keyword lines in the configuration manual
1374for more information.
1375
William Lallemand6ab08b32019-11-29 16:48:43 +01001376abort ssl cert <filename>
1377 Abort and destroy a temporary SSL certificate update transaction.
1378
1379 See also "set ssl cert" and "commit ssl cert".
1380
Willy Tarreau44aed902015-10-13 14:45:29 +02001381add acl <acl> <pattern>
1382 Add an entry into the acl <acl>. <acl> is the #<id> or the <file> returned by
1383 "show acl". This command does not verify if the entry already exists. This
1384 command cannot be used if the reference <acl> is a file also used with a map.
1385 In this case, you must use the command "add map" in place of "add acl".
1386
1387add map <map> <key> <value>
Aurélien Nephtali25650ce2018-04-18 14:04:47 +02001388add map <map> <payload>
Willy Tarreau44aed902015-10-13 14:45:29 +02001389 Add an entry into the map <map> to associate the value <value> to the key
1390 <key>. This command does not verify if the entry already exists. It is
1391 mainly used to fill a map after a clear operation. Note that if the reference
1392 <map> is a file and is shared with a map, this map will contain also a new
Aurélien Nephtali25650ce2018-04-18 14:04:47 +02001393 pattern entry. Using the payload syntax it is possible to add multiple
1394 key/value pairs by entering them on separate lines. On each new line, the
1395 first word is the key and the rest of the line is considered to be the value
1396 which can even contains spaces.
1397
1398 Example:
1399
1400 # socat /tmp/sock1 -
1401 prompt
1402
1403 > add map #-1 <<
1404 + key1 value1
1405 + key2 value2 with spaces
1406 + key3 value3 also with spaces
1407 + key4 value4
1408
1409 >
Willy Tarreau44aed902015-10-13 14:45:29 +02001410
William Lallemandaccac232020-04-02 17:42:51 +02001411add ssl crt-list <crtlist> <certificate>
1412add ssl crt-list <crtlist> <payload>
1413 Add an certificate in a crt-list. It can also be used for directories since
1414 directories are now loaded the same way as the crt-lists. This command allow
1415 you to use a certificate name in parameter, to use SSL options or filters a
1416 crt-list line must sent as a payload instead. Only one crt-list line is
1417 supported in the payload. This command will load the certificate for every
1418 bind lines using the crt-list. To push a new certificate to HAProxy the
1419 commands "new ssl cert" and "set ssl cert" must be used.
1420
1421 Example:
1422 $ echo "new ssl cert foobar.pem" | socat /tmp/sock1 -
1423 $ echo -e "set ssl cert foobar.pem <<\n$(cat foobar.pem)\n" | socat
1424 /tmp/sock1 -
1425 $ echo "commit ssl cert foobar.pem" | socat /tmp/sock1 -
1426 $ echo "add ssl crt-list certlist1 foobar.pem" | socat /tmp/sock1 -
1427
1428 $ echo -e 'add ssl crt-list certlist1 <<\nfoobar.pem [allow-0rtt] foo.bar.com
1429 !test1.com\n' | socat /tmp/sock1 -
1430
Willy Tarreau44aed902015-10-13 14:45:29 +02001431clear counters
1432 Clear the max values of the statistics counters in each proxy (frontend &
Willy Tarreaud80cb4e2018-01-20 19:30:13 +01001433 backend) and in each server. The accumulated counters are not affected. The
1434 internal activity counters reported by "show activity" are also reset. This
Willy Tarreau44aed902015-10-13 14:45:29 +02001435 can be used to get clean counters after an incident, without having to
1436 restart nor to clear traffic counters. This command is restricted and can
1437 only be issued on sockets configured for levels "operator" or "admin".
1438
1439clear counters all
1440 Clear all statistics counters in each proxy (frontend & backend) and in each
1441 server. This has the same effect as restarting. This command is restricted
1442 and can only be issued on sockets configured for level "admin".
1443
1444clear acl <acl>
1445 Remove all entries from the acl <acl>. <acl> is the #<id> or the <file>
1446 returned by "show acl". Note that if the reference <acl> is a file and is
1447 shared with a map, this map will be also cleared.
1448
1449clear map <map>
1450 Remove all entries from the map <map>. <map> is the #<id> or the <file>
1451 returned by "show map". Note that if the reference <map> is a file and is
1452 shared with a acl, this acl will be also cleared.
1453
1454clear table <table> [ data.<type> <operator> <value> ] | [ key <key> ]
1455 Remove entries from the stick-table <table>.
1456
1457 This is typically used to unblock some users complaining they have been
1458 abusively denied access to a service, but this can also be used to clear some
1459 stickiness entries matching a server that is going to be replaced (see "show
1460 table" below for details). Note that sometimes, removal of an entry will be
1461 refused because it is currently tracked by a session. Retrying a few seconds
1462 later after the session ends is usual enough.
1463
1464 In the case where no options arguments are given all entries will be removed.
1465
1466 When the "data." form is used entries matching a filter applied using the
1467 stored data (see "stick-table" in section 4.2) are removed. A stored data
1468 type must be specified in <type>, and this data type must be stored in the
1469 table otherwise an error is reported. The data is compared according to
1470 <operator> with the 64-bit integer <value>. Operators are the same as with
1471 the ACLs :
1472
1473 - eq : match entries whose data is equal to this value
1474 - ne : match entries whose data is not equal to this value
1475 - le : match entries whose data is less than or equal to this value
1476 - ge : match entries whose data is greater than or equal to this value
1477 - lt : match entries whose data is less than this value
1478 - gt : match entries whose data is greater than this value
1479
1480 When the key form is used the entry <key> is removed. The key must be of the
1481 same type as the table, which currently is limited to IPv4, IPv6, integer and
1482 string.
1483
1484 Example :
1485 $ echo "show table http_proxy" | socat stdio /tmp/sock1
1486 >>> # table: http_proxy, type: ip, size:204800, used:2
1487 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
1488 bytes_out_rate(60000)=187
1489 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
1490 bytes_out_rate(60000)=191
1491
1492 $ echo "clear table http_proxy key 127.0.0.1" | socat stdio /tmp/sock1
1493
1494 $ echo "show table http_proxy" | socat stdio /tmp/sock1
1495 >>> # table: http_proxy, type: ip, size:204800, used:1
1496 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
1497 bytes_out_rate(60000)=191
1498 $ echo "clear table http_proxy data.gpc0 eq 1" | socat stdio /tmp/sock1
1499 $ echo "show table http_proxy" | socat stdio /tmp/sock1
1500 >>> # table: http_proxy, type: ip, size:204800, used:1
1501
William Lallemand6ab08b32019-11-29 16:48:43 +01001502commit ssl cert <filename>
William Lallemandc184d872020-06-26 15:39:57 +02001503 Commit a temporary SSL certificate update transaction.
1504
1505 In the case of an existing certificate (in a "Used" state in "show ssl
1506 cert"), generate every SSL contextes and SNIs it need, insert them, and
1507 remove the previous ones. Replace in memory the previous SSL certificates
1508 everywhere the <filename> was used in the configuration. Upon failure it
1509 doesn't remove or insert anything. Once the temporary transaction is
1510 committed, it is destroyed.
1511
1512 In the case of a new certificate (after a "new ssl cert" and in a "Unused"
1513 state in "show ssl cert"), the certificate will be commited in a certificate
1514 storage, but it won't be used anywhere in haproxy. To use it and generate
1515 its SNIs you will need to add it to a crt-list or a directory with "add ssl
1516 crt-list".
William Lallemand6ab08b32019-11-29 16:48:43 +01001517
William Lallemandc184d872020-06-26 15:39:57 +02001518 See also "new ssl cert", "ssl set cert", "abort ssl cert" and
1519 "add ssl crt-list".
William Lallemand6ab08b32019-11-29 16:48:43 +01001520
Willy Tarreau6bdf3e92019-05-20 14:25:05 +02001521debug dev <command> [args]*
Willy Tarreaub24ab222019-10-24 18:03:39 +02001522 Call a developer-specific command. Only supported on a CLI connection running
1523 in expert mode (see "expert-mode on"). Such commands are extremely dangerous
1524 and not forgiving, any misuse may result in a crash of the process. They are
1525 intended for experts only, and must really not be used unless told to do so.
1526 Some of them are only available when haproxy is built with DEBUG_DEV defined
1527 because they may have security implications. All of these commands require
1528 admin privileges, and are purposely not documented to avoid encouraging their
1529 use by people who are not at ease with the source code.
Willy Tarreau6bdf3e92019-05-20 14:25:05 +02001530
Willy Tarreau44aed902015-10-13 14:45:29 +02001531del acl <acl> [<key>|#<ref>]
1532 Delete all the acl entries from the acl <acl> corresponding to the key <key>.
1533 <acl> is the #<id> or the <file> returned by "show acl". If the <ref> is used,
1534 this command delete only the listed reference. The reference can be found with
1535 listing the content of the acl. Note that if the reference <acl> is a file and
1536 is shared with a map, the entry will be also deleted in the map.
1537
1538del map <map> [<key>|#<ref>]
1539 Delete all the map entries from the map <map> corresponding to the key <key>.
1540 <map> is the #<id> or the <file> returned by "show map". If the <ref> is used,
1541 this command delete only the listed reference. The reference can be found with
1542 listing the content of the map. Note that if the reference <map> is a file and
1543 is shared with a acl, the entry will be also deleted in the map.
1544
William Lallemand419e6342020-04-08 12:05:39 +02001545del ssl cert <certfile>
1546 Delete a certificate store from HAProxy. The certificate must be unused and
1547 removed from any crt-list or directory. "show ssl cert" displays the status
1548 of the certificate. The deletion doesn't work with a certificate referenced
1549 directly with the "crt" directive in the configuration.
1550
William Lallemand0a9b9412020-04-06 17:43:05 +02001551del ssl crt-list <filename> <certfile[:line]>
1552 Delete an entry in a crt-list. This will delete every SNIs used for this
1553 entry in the frontends. If a certificate is used several time in a crt-list,
1554 you will need to provide which line you want to delete. To display the line
1555 numbers, use "show ssl crt-list -n <crtlist>".
1556
Willy Tarreau44aed902015-10-13 14:45:29 +02001557disable agent <backend>/<server>
1558 Mark the auxiliary agent check as temporarily stopped.
1559
1560 In the case where an agent check is being run as a auxiliary check, due
1561 to the agent-check parameter of a server directive, new checks are only
Dan Lloyd8e48b872016-07-01 21:01:18 -04001562 initialized when the agent is in the enabled. Thus, disable agent will
Willy Tarreau44aed902015-10-13 14:45:29 +02001563 prevent any new agent checks from begin initiated until the agent
1564 re-enabled using enable agent.
1565
1566 When an agent is disabled the processing of an auxiliary agent check that
1567 was initiated while the agent was set as enabled is as follows: All
1568 results that would alter the weight, specifically "drain" or a weight
1569 returned by the agent, are ignored. The processing of agent check is
1570 otherwise unchanged.
1571
1572 The motivation for this feature is to allow the weight changing effects
1573 of the agent checks to be paused to allow the weight of a server to be
1574 configured using set weight without being overridden by the agent.
1575
1576 This command is restricted and can only be issued on sockets configured for
1577 level "admin".
1578
Olivier Houchard614f8d72017-03-14 20:08:46 +01001579disable dynamic-cookie backend <backend>
Ilya Shipitsin2a950d02020-03-06 13:07:38 +05001580 Disable the generation of dynamic cookies for the backend <backend>
Olivier Houchard614f8d72017-03-14 20:08:46 +01001581
Willy Tarreau44aed902015-10-13 14:45:29 +02001582disable frontend <frontend>
1583 Mark the frontend as temporarily stopped. This corresponds to the mode which
1584 is used during a soft restart : the frontend releases the port but can be
1585 enabled again if needed. This should be used with care as some non-Linux OSes
1586 are unable to enable it back. This is intended to be used in environments
1587 where stopping a proxy is not even imaginable but a misconfigured proxy must
1588 be fixed. That way it's possible to release the port and bind it into another
1589 process to restore operations. The frontend will appear with status "STOP"
1590 on the stats page.
1591
1592 The frontend may be specified either by its name or by its numeric ID,
1593 prefixed with a sharp ('#').
1594
1595 This command is restricted and can only be issued on sockets configured for
1596 level "admin".
1597
1598disable health <backend>/<server>
1599 Mark the primary health check as temporarily stopped. This will disable
1600 sending of health checks, and the last health check result will be ignored.
1601 The server will be in unchecked state and considered UP unless an auxiliary
1602 agent check forces it down.
1603
1604 This command is restricted and can only be issued on sockets configured for
1605 level "admin".
1606
1607disable server <backend>/<server>
1608 Mark the server DOWN for maintenance. In this mode, no more checks will be
1609 performed on the server until it leaves maintenance.
1610 If the server is tracked by other servers, those servers will be set to DOWN
1611 during the maintenance.
1612
1613 In the statistics page, a server DOWN for maintenance will appear with a
1614 "MAINT" status, its tracking servers with the "MAINT(via)" one.
1615
1616 Both the backend and the server may be specified either by their name or by
1617 their numeric ID, prefixed with a sharp ('#').
1618
1619 This command is restricted and can only be issued on sockets configured for
1620 level "admin".
1621
1622enable agent <backend>/<server>
1623 Resume auxiliary agent check that was temporarily stopped.
1624
1625 See "disable agent" for details of the effect of temporarily starting
1626 and stopping an auxiliary agent.
1627
1628 This command is restricted and can only be issued on sockets configured for
1629 level "admin".
1630
Olivier Houchard614f8d72017-03-14 20:08:46 +01001631enable dynamic-cookie backend <backend>
n9@users.noreply.github.com25a1c8e2019-08-23 11:21:05 +02001632 Enable the generation of dynamic cookies for the backend <backend>.
1633 A secret key must also be provided.
Olivier Houchard614f8d72017-03-14 20:08:46 +01001634
Willy Tarreau44aed902015-10-13 14:45:29 +02001635enable frontend <frontend>
1636 Resume a frontend which was temporarily stopped. It is possible that some of
1637 the listening ports won't be able to bind anymore (eg: if another process
1638 took them since the 'disable frontend' operation). If this happens, an error
1639 is displayed. Some operating systems might not be able to resume a frontend
1640 which was disabled.
1641
1642 The frontend may be specified either by its name or by its numeric ID,
1643 prefixed with a sharp ('#').
1644
1645 This command is restricted and can only be issued on sockets configured for
1646 level "admin".
1647
1648enable health <backend>/<server>
1649 Resume a primary health check that was temporarily stopped. This will enable
1650 sending of health checks again. Please see "disable health" for details.
1651
1652 This command is restricted and can only be issued on sockets configured for
1653 level "admin".
1654
1655enable server <backend>/<server>
1656 If the server was previously marked as DOWN for maintenance, this marks the
1657 server UP and checks are re-enabled.
1658
1659 Both the backend and the server may be specified either by their name or by
1660 their numeric ID, prefixed with a sharp ('#').
1661
1662 This command is restricted and can only be issued on sockets configured for
1663 level "admin".
1664
Willy Tarreauabb9f9b2019-10-24 17:55:53 +02001665expert-mode [on|off]
1666 Without options, this indicates whether the expert mode is enabled or
1667 disabled on the current connection. When passed "on", it turns the expert
1668 mode on for the current CLI connection only. With "off" it turns it off. The
1669 expert mode enables displaying of expert commands that can be extremely
1670 dangerous for the process and which may occasionally help developers collect
1671 important information about complex bugs. Any misuse of these features will
1672 likely lead to a process crash. Do not use this option without being invited
1673 to do so. Note that this command is purposely not listed in the help message.
1674 This command is only accessible in admin level. Changing to another level
1675 automatically resets the expert mode.
1676
Willy Tarreau44aed902015-10-13 14:45:29 +02001677get map <map> <value>
1678get acl <acl> <value>
1679 Lookup the value <value> in the map <map> or in the ACL <acl>. <map> or <acl>
1680 are the #<id> or the <file> returned by "show map" or "show acl". This command
1681 returns all the matching patterns associated with this map. This is useful for
1682 debugging maps and ACLs. The output format is composed by one line par
1683 matching type. Each line is composed by space-delimited series of words.
1684
1685 The first two words are:
1686
1687 <match method>: The match method applied. It can be "found", "bool",
1688 "int", "ip", "bin", "len", "str", "beg", "sub", "dir",
1689 "dom", "end" or "reg".
1690
1691 <match result>: The result. Can be "match" or "no-match".
1692
1693 The following words are returned only if the pattern matches an entry.
1694
1695 <index type>: "tree" or "list". The internal lookup algorithm.
1696
1697 <case>: "case-insensitive" or "case-sensitive". The
1698 interpretation of the case.
1699
1700 <entry matched>: match="<entry>". Return the matched pattern. It is
1701 useful with regular expressions.
1702
1703 The two last word are used to show the returned value and its type. With the
1704 "acl" case, the pattern doesn't exist.
1705
1706 return=nothing: No return because there are no "map".
1707 return="<value>": The value returned in the string format.
1708 return=cannot-display: The value cannot be converted as string.
1709
1710 type="<type>": The type of the returned sample.
1711
1712get weight <backend>/<server>
1713 Report the current weight and the initial weight of server <server> in
1714 backend <backend> or an error if either doesn't exist. The initial weight is
1715 the one that appears in the configuration file. Both are normally equal
1716 unless the current weight has been changed. Both the backend and the server
1717 may be specified either by their name or by their numeric ID, prefixed with a
1718 sharp ('#').
1719
1720help
1721 Print the list of known keywords and their basic usage. The same help screen
1722 is also displayed for unknown commands.
1723
William Lallemandaccac232020-04-02 17:42:51 +02001724new ssl cert <filename>
1725 Create a new empty SSL certificate store to be filled with a certificate and
1726 added to a directory or a crt-list. This command should be used in
1727 combination with "set ssl cert" and "add ssl crt-list".
1728
Willy Tarreau44aed902015-10-13 14:45:29 +02001729prompt
1730 Toggle the prompt at the beginning of the line and enter or leave interactive
1731 mode. In interactive mode, the connection is not closed after a command
1732 completes. Instead, the prompt will appear again, indicating the user that
1733 the interpreter is waiting for a new command. The prompt consists in a right
1734 angle bracket followed by a space "> ". This mode is particularly convenient
1735 when one wants to periodically check information such as stats or errors.
1736 It is also a good idea to enter interactive mode before issuing a "help"
1737 command.
1738
1739quit
1740 Close the connection when in interactive mode.
1741
Olivier Houchard614f8d72017-03-14 20:08:46 +01001742set dynamic-cookie-key backend <backend> <value>
1743 Modify the secret key used to generate the dynamic persistent cookies.
1744 This will break the existing sessions.
1745
Willy Tarreau44aed902015-10-13 14:45:29 +02001746set map <map> [<key>|#<ref>] <value>
1747 Modify the value corresponding to each key <key> in a map <map>. <map> is the
1748 #<id> or <file> returned by "show map". If the <ref> is used in place of
1749 <key>, only the entry pointed by <ref> is changed. The new value is <value>.
1750
1751set maxconn frontend <frontend> <value>
1752 Dynamically change the specified frontend's maxconn setting. Any positive
1753 value is allowed including zero, but setting values larger than the global
1754 maxconn does not make much sense. If the limit is increased and connections
1755 were pending, they will immediately be accepted. If it is lowered to a value
1756 below the current number of connections, new connections acceptation will be
1757 delayed until the threshold is reached. The frontend might be specified by
1758 either its name or its numeric ID prefixed with a sharp ('#').
1759
Andrew Hayworthedb93a72015-10-27 21:46:25 +00001760set maxconn server <backend/server> <value>
1761 Dynamically change the specified server's maxconn setting. Any positive
1762 value is allowed including zero, but setting values larger than the global
1763 maxconn does not make much sense.
1764
Willy Tarreau44aed902015-10-13 14:45:29 +02001765set maxconn global <maxconn>
1766 Dynamically change the global maxconn setting within the range defined by the
1767 initial global maxconn setting. If it is increased and connections were
1768 pending, they will immediately be accepted. If it is lowered to a value below
1769 the current number of connections, new connections acceptation will be
1770 delayed until the threshold is reached. A value of zero restores the initial
1771 setting.
1772
Willy Tarreaud2d33482019-04-25 17:09:07 +02001773set profiling { tasks } { auto | on | off }
Willy Tarreau75c62c22018-11-22 11:02:09 +01001774 Enables or disables CPU profiling for the indicated subsystem. This is
1775 equivalent to setting or clearing the "profiling" settings in the "global"
1776 section of the configuration file. Please also see "show profiling".
1777
Willy Tarreau44aed902015-10-13 14:45:29 +02001778set rate-limit connections global <value>
1779 Change the process-wide connection rate limit, which is set by the global
1780 'maxconnrate' setting. A value of zero disables the limitation. This limit
1781 applies to all frontends and the change has an immediate effect. The value
1782 is passed in number of connections per second.
1783
1784set rate-limit http-compression global <value>
1785 Change the maximum input compression rate, which is set by the global
1786 'maxcomprate' setting. A value of zero disables the limitation. The value is
1787 passed in number of kilobytes per second. The value is available in the "show
1788 info" on the line "CompressBpsRateLim" in bytes.
1789
1790set rate-limit sessions global <value>
1791 Change the process-wide session rate limit, which is set by the global
1792 'maxsessrate' setting. A value of zero disables the limitation. This limit
1793 applies to all frontends and the change has an immediate effect. The value
1794 is passed in number of sessions per second.
1795
1796set rate-limit ssl-sessions global <value>
1797 Change the process-wide SSL session rate limit, which is set by the global
1798 'maxsslrate' setting. A value of zero disables the limitation. This limit
1799 applies to all frontends and the change has an immediate effect. The value
1800 is passed in number of sessions per second sent to the SSL stack. It applies
1801 before the handshake in order to protect the stack against handshake abuses.
1802
Baptiste Assmann3749ebf2016-08-03 22:34:12 +02001803set server <backend>/<server> addr <ip4 or ip6 address> [port <port>]
Willy Tarreau44aed902015-10-13 14:45:29 +02001804 Replace the current IP address of a server by the one provided.
Michael Prokop4438c602019-05-24 10:25:45 +02001805 Optionally, the port can be changed using the 'port' parameter.
Baptiste Assmann3749ebf2016-08-03 22:34:12 +02001806 Note that changing the port also support switching from/to port mapping
1807 (notation with +X or -Y), only if a port is configured for the health check.
Willy Tarreau44aed902015-10-13 14:45:29 +02001808
1809set server <backend>/<server> agent [ up | down ]
1810 Force a server's agent to a new state. This can be useful to immediately
1811 switch a server's state regardless of some slow agent checks for example.
1812 Note that the change is propagated to tracking servers if any.
1813
Misiek43972902017-01-09 09:53:06 +01001814set server <backend>/<server> agent-addr <addr>
1815 Change addr for servers agent checks. Allows to migrate agent-checks to
1816 another address at runtime. You can specify both IP and hostname, it will be
1817 resolved.
1818
1819set server <backend>/<server> agent-send <value>
1820 Change agent string sent to agent check target. Allows to update string while
1821 changing server address to keep those two matching.
1822
Willy Tarreau44aed902015-10-13 14:45:29 +02001823set server <backend>/<server> health [ up | stopping | down ]
1824 Force a server's health to a new state. This can be useful to immediately
1825 switch a server's state regardless of some slow health checks for example.
1826 Note that the change is propagated to tracking servers if any.
1827
Baptiste Assmann50946562016-08-31 23:26:29 +02001828set server <backend>/<server> check-port <port>
1829 Change the port used for health checking to <port>
1830
Willy Tarreau44aed902015-10-13 14:45:29 +02001831set server <backend>/<server> state [ ready | drain | maint ]
1832 Force a server's administrative state to a new state. This can be useful to
1833 disable load balancing and/or any traffic to a server. Setting the state to
1834 "ready" puts the server in normal mode, and the command is the equivalent of
1835 the "enable server" command. Setting the state to "maint" disables any traffic
1836 to the server as well as any health checks. This is the equivalent of the
1837 "disable server" command. Setting the mode to "drain" only removes the server
1838 from load balancing but still allows it to be checked and to accept new
1839 persistent connections. Changes are propagated to tracking servers if any.
1840
1841set server <backend>/<server> weight <weight>[%]
1842 Change a server's weight to the value passed in argument. This is the exact
1843 equivalent of the "set weight" command below.
1844
Frédéric Lécailleb418c122017-04-26 11:24:02 +02001845set server <backend>/<server> fqdn <FQDN>
Lukas Tribusc5dd5a52018-08-14 11:39:35 +02001846 Change a server's FQDN to the value passed in argument. This requires the
1847 internal run-time DNS resolver to be configured and enabled for this server.
Frédéric Lécailleb418c122017-04-26 11:24:02 +02001848
Andjelko Iharosc4df59e2017-07-20 11:59:48 +02001849set severity-output [ none | number | string ]
1850 Change the severity output format of the stats socket connected to for the
1851 duration of the current session.
1852
William Lallemand6ab08b32019-11-29 16:48:43 +01001853set ssl cert <filename> <payload>
1854 This command is part of a transaction system, the "commit ssl cert" and
1855 "abort ssl cert" commands could be required.
1856 If there is no on-going transaction, it will duplicate the certificate
1857 <filename> in memory to a temporary transaction, then update this
1858 transaction with the PEM file in the payload. If a transaction exists with
1859 the same filename, it will update this transaction. It's also possible to
1860 update the files linked to a certificate (.issuer, .sctl, .oscp etc.)
1861 Once the modification are done, you have to "commit ssl cert" the
1862 transaction.
1863
1864 Example:
1865 echo -e "set ssl cert localhost.pem <<\n$(cat 127.0.0.1.pem)\n" | \
1866 socat /var/run/haproxy.stat -
1867 echo -e \
1868 "set ssl cert localhost.pem.issuer <<\n $(cat 127.0.0.1.pem.issuer)\n" | \
1869 socat /var/run/haproxy.stat -
1870 echo -e \
1871 "set ssl cert localhost.pem.ocsp <<\n$(base64 -w 1000 127.0.0.1.pem.ocsp)\n" | \
1872 socat /var/run/haproxy.stat -
1873 echo "commit ssl cert localhost.pem" | socat /var/run/haproxy.stat -
1874
Aurélien Nephtali1e0867c2018-04-18 14:04:58 +02001875set ssl ocsp-response <response | payload>
Willy Tarreau44aed902015-10-13 14:45:29 +02001876 This command is used to update an OCSP Response for a certificate (see "crt"
1877 on "bind" lines). Same controls are performed as during the initial loading of
1878 the response. The <response> must be passed as a base64 encoded string of the
Emmanuel Hocdet2c32d8f2017-05-22 14:58:00 +02001879 DER encoded response from the OCSP server. This command is not supported with
1880 BoringSSL.
Willy Tarreau44aed902015-10-13 14:45:29 +02001881
1882 Example:
1883 openssl ocsp -issuer issuer.pem -cert server.pem \
1884 -host ocsp.issuer.com:80 -respout resp.der
1885 echo "set ssl ocsp-response $(base64 -w 10000 resp.der)" | \
1886 socat stdio /var/run/haproxy.stat
1887
Aurélien Nephtali1e0867c2018-04-18 14:04:58 +02001888 using the payload syntax:
1889 echo -e "set ssl ocsp-response <<\n$(base64 resp.der)\n" | \
1890 socat stdio /var/run/haproxy.stat
1891
Willy Tarreau44aed902015-10-13 14:45:29 +02001892set ssl tls-key <id> <tlskey>
1893 Set the next TLS key for the <id> listener to <tlskey>. This key becomes the
1894 ultimate key, while the penultimate one is used for encryption (others just
1895 decrypt). The oldest TLS key present is overwritten. <id> is either a numeric
1896 #<id> or <file> returned by "show tls-keys". <tlskey> is a base64 encoded 48
Emeric Brun9e754772019-01-10 17:51:55 +01001897 or 80 bits TLS ticket key (ex. openssl rand 80 | openssl base64 -A).
Willy Tarreau44aed902015-10-13 14:45:29 +02001898
1899set table <table> key <key> [data.<data_type> <value>]*
1900 Create or update a stick-table entry in the table. If the key is not present,
1901 an entry is inserted. See stick-table in section 4.2 to find all possible
1902 values for <data_type>. The most likely use consists in dynamically entering
1903 entries for source IP addresses, with a flag in gpc0 to dynamically block an
1904 IP address or affect its quality of service. It is possible to pass multiple
1905 data_types in a single call.
1906
1907set timeout cli <delay>
1908 Change the CLI interface timeout for current connection. This can be useful
1909 during long debugging sessions where the user needs to constantly inspect
1910 some indicators without being disconnected. The delay is passed in seconds.
1911
1912set weight <backend>/<server> <weight>[%]
1913 Change a server's weight to the value passed in argument. If the value ends
1914 with the '%' sign, then the new weight will be relative to the initially
1915 configured weight. Absolute weights are permitted between 0 and 256.
1916 Relative weights must be positive with the resulting absolute weight is
1917 capped at 256. Servers which are part of a farm running a static
1918 load-balancing algorithm have stricter limitations because the weight
1919 cannot change once set. Thus for these servers, the only accepted values
1920 are 0 and 100% (or 0 and the initial weight). Changes take effect
1921 immediately, though certain LB algorithms require a certain amount of
1922 requests to consider changes. A typical usage of this command is to
1923 disable a server during an update by setting its weight to zero, then to
1924 enable it again after the update by setting it back to 100%. This command
1925 is restricted and can only be issued on sockets configured for level
1926 "admin". Both the backend and the server may be specified either by their
1927 name or by their numeric ID, prefixed with a sharp ('#').
1928
Willy Tarreaud6129fc2017-07-28 16:52:23 +02001929show acl [<acl>]
1930 Dump info about acl converters. Without argument, the list of all available
1931 acls is returned. If a <acl> is specified, its contents are dumped. <acl> if
1932 the #<id> or <file>. The dump format is the same than the map even for the
1933 sample value. The data returned are not a list of available ACL, but are the
1934 list of all patterns composing any ACL. Many of these patterns can be shared
1935 with maps.
1936
1937show backend
1938 Dump the list of backends available in the running process
1939
William Lallemand67a234f2018-12-13 09:05:45 +01001940show cli level
1941 Display the CLI level of the current CLI session. The result could be
1942 'admin', 'operator' or 'user'. See also the 'operator' and 'user' commands.
1943
1944 Example :
1945
1946 $ socat /tmp/sock1 readline
1947 prompt
1948 > operator
1949 > show cli level
1950 operator
1951 > user
1952 > show cli level
1953 user
1954 > operator
1955 Permission denied
1956
1957operator
1958 Decrease the CLI level of the current CLI session to operator. It can't be
Willy Tarreauabb9f9b2019-10-24 17:55:53 +02001959 increased. It also drops expert mode. See also "show cli level".
William Lallemand67a234f2018-12-13 09:05:45 +01001960
1961user
1962 Decrease the CLI level of the current CLI session to user. It can't be
Willy Tarreauabb9f9b2019-10-24 17:55:53 +02001963 increased. It also drops expert mode. See also "show cli level".
William Lallemand67a234f2018-12-13 09:05:45 +01001964
Willy Tarreau4c356932019-05-16 17:39:32 +02001965show activity
1966 Reports some counters about internal events that will help developers and
1967 more generally people who know haproxy well enough to narrow down the causes
1968 of reports of abnormal behaviours. A typical example would be a properly
1969 running process never sleeping and eating 100% of the CPU. The output fields
1970 will be made of one line per metric, and per-thread counters on the same
1971 line. These counters are 32-bit and will wrap during the process' life, which
1972 is not a problem since calls to this command will typically be performed
1973 twice. The fields are purposely not documented so that their exact meaning is
1974 verified in the code where the counters are fed. These values are also reset
1975 by the "clear counters" command.
1976
William Lallemand51132162016-12-16 16:38:58 +01001977show cli sockets
1978 List CLI sockets. The output format is composed of 3 fields separated by
1979 spaces. The first field is the socket address, it can be a unix socket, a
1980 ipv4 address:port couple or a ipv6 one. Socket of other types won't be dump.
1981 The second field describe the level of the socket: 'admin', 'user' or
1982 'operator'. The last field list the processes on which the socket is bound,
1983 separated by commas, it can be numbers or 'all'.
1984
1985 Example :
1986
1987 $ echo 'show cli sockets' | socat stdio /tmp/sock1
1988 # socket lvl processes
1989 /tmp/sock1 admin all
1990 127.0.0.1:9999 user 2,3,4
1991 127.0.0.2:9969 user 2
1992 [::1]:9999 operator 2
1993
William Lallemand86d0df02017-11-24 21:36:45 +01001994show cache
Cyril Bonté7b888f12017-11-26 22:24:31 +01001995 List the configured caches and the objects stored in each cache tree.
William Lallemand86d0df02017-11-24 21:36:45 +01001996
1997 $ echo 'show cache' | socat stdio /tmp/sock1
1998 0x7f6ac6c5b03a: foobar (shctx:0x7f6ac6c5b000, available blocks:3918)
1999 1 2 3 4
2000
2001 1. pointer to the cache structure
2002 2. cache name
2003 3. pointer to the mmap area (shctx)
2004 4. number of blocks available for reuse in the shctx
2005
2006 0x7f6ac6c5b4cc hash:286881868 size:39114 (39 blocks), refcount:9, expire:237
2007 1 2 3 4 5 6
2008
2009 1. pointer to the cache entry
2010 2. first 32 bits of the hash
2011 3. size of the object in bytes
2012 4. number of blocks used for the object
2013 5. number of transactions using the entry
2014 6. expiration time, can be negative if already expired
2015
Willy Tarreauae795722016-02-16 11:27:28 +01002016show env [<name>]
2017 Dump one or all environment variables known by the process. Without any
2018 argument, all variables are dumped. With an argument, only the specified
2019 variable is dumped if it exists. Otherwise "Variable not found" is emitted.
2020 Variables are dumped in the same format as they are stored or returned by the
2021 "env" utility, that is, "<name>=<value>". This can be handy when debugging
2022 certain configuration files making heavy use of environment variables to
2023 ensure that they contain the expected values. This command is restricted and
2024 can only be issued on sockets configured for levels "operator" or "admin".
2025
Willy Tarreau35069f82016-11-25 09:16:37 +01002026show errors [<iid>|<proxy>] [request|response]
Willy Tarreau44aed902015-10-13 14:45:29 +02002027 Dump last known request and response errors collected by frontends and
2028 backends. If <iid> is specified, the limit the dump to errors concerning
Willy Tarreau234ba2d2016-11-25 08:39:10 +01002029 either frontend or backend whose ID is <iid>. Proxy ID "-1" will cause
2030 all instances to be dumped. If a proxy name is specified instead, its ID
Willy Tarreau35069f82016-11-25 09:16:37 +01002031 will be used as the filter. If "request" or "response" is added after the
2032 proxy name or ID, only request or response errors will be dumped. This
2033 command is restricted and can only be issued on sockets configured for
2034 levels "operator" or "admin".
Willy Tarreau44aed902015-10-13 14:45:29 +02002035
2036 The errors which may be collected are the last request and response errors
2037 caused by protocol violations, often due to invalid characters in header
2038 names. The report precisely indicates what exact character violated the
2039 protocol. Other important information such as the exact date the error was
2040 detected, frontend and backend names, the server name (when known), the
2041 internal session ID and the source address which has initiated the session
2042 are reported too.
2043
2044 All characters are returned, and non-printable characters are encoded. The
2045 most common ones (\t = 9, \n = 10, \r = 13 and \e = 27) are encoded as one
2046 letter following a backslash. The backslash itself is encoded as '\\' to
2047 avoid confusion. Other non-printable characters are encoded '\xNN' where
2048 NN is the two-digits hexadecimal representation of the character's ASCII
2049 code.
2050
2051 Lines are prefixed with the position of their first character, starting at 0
2052 for the beginning of the buffer. At most one input line is printed per line,
2053 and large lines will be broken into multiple consecutive output lines so that
2054 the output never goes beyond 79 characters wide. It is easy to detect if a
2055 line was broken, because it will not end with '\n' and the next line's offset
2056 will be followed by a '+' sign, indicating it is a continuation of previous
2057 line.
2058
2059 Example :
Willy Tarreau35069f82016-11-25 09:16:37 +01002060 $ echo "show errors -1 response" | socat stdio /tmp/sock1
Willy Tarreau44aed902015-10-13 14:45:29 +02002061 >>> [04/Mar/2009:15:46:56.081] backend http-in (#2) : invalid response
2062 src 127.0.0.1, session #54, frontend fe-eth0 (#1), server s2 (#1)
2063 response length 213 bytes, error at position 23:
2064
2065 00000 HTTP/1.0 200 OK\r\n
2066 00017 header/bizarre:blah\r\n
2067 00038 Location: blah\r\n
2068 00054 Long-line: this is a very long line which should b
2069 00104+ e broken into multiple lines on the output buffer,
2070 00154+ otherwise it would be too large to print in a ter
2071 00204+ minal\r\n
2072 00211 \r\n
2073
2074 In the example above, we see that the backend "http-in" which has internal
2075 ID 2 has blocked an invalid response from its server s2 which has internal
2076 ID 1. The request was on session 54 initiated by source 127.0.0.1 and
2077 received by frontend fe-eth0 whose ID is 1. The total response length was
2078 213 bytes when the error was detected, and the error was at byte 23. This
2079 is the slash ('/') in header name "header/bizarre", which is not a valid
2080 HTTP character for a header name.
2081
Willy Tarreau1d181e42019-08-30 11:17:01 +02002082show events [<sink>] [-w] [-n]
Willy Tarreau9f830d72019-08-26 18:17:04 +02002083 With no option, this lists all known event sinks and their types. With an
2084 option, it will dump all available events in the designated sink if it is of
Willy Tarreau1d181e42019-08-30 11:17:01 +02002085 type buffer. If option "-w" is passed after the sink name, then once the end
2086 of the buffer is reached, the command will wait for new events and display
2087 them. It is possible to stop the operation by entering any input (which will
2088 be discarded) or by closing the session. Finally, option "-n" is used to
2089 directly seek to the end of the buffer, which is often convenient when
2090 combined with "-w" to only report new events. For convenience, "-wn" or "-nw"
2091 may be used to enable both options at once.
Willy Tarreau9f830d72019-08-26 18:17:04 +02002092
Willy Tarreau7a4a0ac2017-07-25 19:32:50 +02002093show fd [<fd>]
2094 Dump the list of either all open file descriptors or just the one number <fd>
2095 if specified. This is only aimed at developers who need to observe internal
2096 states in order to debug complex issues such as abnormal CPU usages. One fd
2097 is reported per lines, and for each of them, its state in the poller using
2098 upper case letters for enabled flags and lower case for disabled flags, using
2099 "P" for "polled", "R" for "ready", "A" for "active", the events status using
2100 "H" for "hangup", "E" for "error", "O" for "output", "P" for "priority" and
2101 "I" for "input", a few other flags like "N" for "new" (just added into the fd
2102 cache), "U" for "updated" (received an update in the fd cache), "L" for
2103 "linger_risk", "C" for "cloned", then the cached entry position, the pointer
2104 to the internal owner, the pointer to the I/O callback and its name when
2105 known. When the owner is a connection, the connection flags, and the target
2106 are reported (frontend, proxy or server). When the owner is a listener, the
2107 listener's state and its frontend are reported. There is no point in using
2108 this command without a good knowledge of the internals. It's worth noting
2109 that the output format may evolve over time so this output must not be parsed
2110 by tools designed to be durable.
2111
Willy Tarreau6b19b142019-10-09 15:44:21 +02002112show info [typed|json] [desc]
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002113 Dump info about haproxy status on current process. If "typed" is passed as an
2114 optional argument, field numbers, names and types are emitted as well so that
2115 external monitoring products can easily retrieve, possibly aggregate, then
2116 report information found in fields they don't know. Each field is dumped on
Simon Horman05ee2132017-01-04 09:37:25 +01002117 its own line. If "json" is passed as an optional argument then
2118 information provided by "typed" output is provided in JSON format as a
2119 list of JSON objects. By default, the format contains only two columns
2120 delimited by a colon (':'). The left one is the field name and the right
2121 one is the value. It is very important to note that in typed output
2122 format, the dump for a single object is contiguous so that there is no
2123 need for a consumer to store everything at once.
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002124
2125 When using the typed output format, each line is made of 4 columns delimited
2126 by colons (':'). The first column is a dot-delimited series of 3 elements. The
2127 first element is the numeric position of the field in the list (starting at
2128 zero). This position shall not change over time, but holes are to be expected,
2129 depending on build options or if some fields are deleted in the future. The
2130 second element is the field name as it appears in the default "show info"
2131 output. The third element is the relative process number starting at 1.
2132
2133 The rest of the line starting after the first colon follows the "typed output
2134 format" described in the section above. In short, the second column (after the
2135 first ':') indicates the origin, nature and scope of the variable. The third
2136 column indicates the type of the field, among "s32", "s64", "u32", "u64" and
2137 "str". Then the fourth column is the value itself, which the consumer knows
2138 how to parse thanks to column 3 and how to process thanks to column 2.
2139
2140 Thus the overall line format in typed mode is :
2141
2142 <field_pos>.<field_name>.<process_num>:<tags>:<type>:<value>
2143
Willy Tarreau6b19b142019-10-09 15:44:21 +02002144 When "desc" is appended to the command, one extra colon followed by a quoted
2145 string is appended with a description for the metric. At the time of writing,
2146 this is only supported for the "typed" and default output formats.
2147
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002148 Example :
2149
2150 > show info
2151 Name: HAProxy
2152 Version: 1.7-dev1-de52ea-146
2153 Release_date: 2016/03/11
2154 Nbproc: 1
2155 Process_num: 1
2156 Pid: 28105
2157 Uptime: 0d 0h00m04s
2158 Uptime_sec: 4
2159 Memmax_MB: 0
2160 PoolAlloc_MB: 0
2161 PoolUsed_MB: 0
2162 PoolFailed: 0
2163 (...)
2164
2165 > show info typed
2166 0.Name.1:POS:str:HAProxy
2167 1.Version.1:POS:str:1.7-dev1-de52ea-146
2168 2.Release_date.1:POS:str:2016/03/11
2169 3.Nbproc.1:CGS:u32:1
2170 4.Process_num.1:KGP:u32:1
2171 5.Pid.1:SGP:u32:28105
2172 6.Uptime.1:MDP:str:0d 0h00m08s
2173 7.Uptime_sec.1:MDP:u32:8
2174 8.Memmax_MB.1:CLP:u32:0
2175 9.PoolAlloc_MB.1:MGP:u32:0
2176 10.PoolUsed_MB.1:MGP:u32:0
2177 11.PoolFailed.1:MCP:u32:0
2178 (...)
2179
Simon Horman1084a362016-11-21 17:00:24 +01002180 In the typed format, the presence of the process ID at the end of the
2181 first column makes it very easy to visually aggregate outputs from
2182 multiple processes.
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002183 Example :
2184
2185 $ ( echo show info typed | socat /var/run/haproxy.sock1 ; \
2186 echo show info typed | socat /var/run/haproxy.sock2 ) | \
2187 sort -t . -k 1,1n -k 2,2 -k 3,3n
2188 0.Name.1:POS:str:HAProxy
2189 0.Name.2:POS:str:HAProxy
2190 1.Version.1:POS:str:1.7-dev1-868ab3-148
2191 1.Version.2:POS:str:1.7-dev1-868ab3-148
2192 2.Release_date.1:POS:str:2016/03/11
2193 2.Release_date.2:POS:str:2016/03/11
2194 3.Nbproc.1:CGS:u32:2
2195 3.Nbproc.2:CGS:u32:2
2196 4.Process_num.1:KGP:u32:1
2197 4.Process_num.2:KGP:u32:2
2198 5.Pid.1:SGP:u32:30120
2199 5.Pid.2:SGP:u32:30121
2200 6.Uptime.1:MDP:str:0d 0h01m28s
2201 6.Uptime.2:MDP:str:0d 0h01m28s
2202 (...)
Willy Tarreau44aed902015-10-13 14:45:29 +02002203
Simon Horman05ee2132017-01-04 09:37:25 +01002204 The format of JSON output is described in a schema which may be output
Simon Horman6f6bb382017-01-04 09:37:26 +01002205 using "show schema json".
Simon Horman05ee2132017-01-04 09:37:25 +01002206
2207 The JSON output contains no extra whitespace in order to reduce the
2208 volume of output. For human consumption passing the output through a
2209 pretty printer may be helpful. Example :
2210
2211 $ echo "show info json" | socat /var/run/haproxy.sock stdio | \
2212 python -m json.tool
2213
Simon Horman6f6bb382017-01-04 09:37:26 +01002214 The JSON output contains no extra whitespace in order to reduce the
2215 volume of output. For human consumption passing the output through a
2216 pretty printer may be helpful. Example :
2217
2218 $ echo "show info json" | socat /var/run/haproxy.sock stdio | \
2219 python -m json.tool
2220
Willy Tarreau44aed902015-10-13 14:45:29 +02002221show map [<map>]
2222 Dump info about map converters. Without argument, the list of all available
2223 maps is returned. If a <map> is specified, its contents are dumped. <map> is
2224 the #<id> or <file>. The first column is a unique identifier. It can be used
2225 as reference for the operation "del map" and "set map". The second column is
2226 the pattern and the third column is the sample if available. The data returned
2227 are not directly a list of available maps, but are the list of all patterns
2228 composing any map. Many of these patterns can be shared with ACL.
2229
Frédéric Lécaille21dde502019-04-15 13:50:23 +02002230show peers [<peers section>]
2231 Dump info about the peers configured in "peers" sections. Without argument,
2232 the list of the peers belonging to all the "peers" sections are listed. If
2233 <peers section> is specified, only the information about the peers belonging
2234 to this "peers" section are dumped.
2235
Michael Prokop4438c602019-05-24 10:25:45 +02002236 Here are two examples of outputs where hostA, hostB and hostC peers belong to
Frédéric Lécaille21dde502019-04-15 13:50:23 +02002237 "sharedlb" peers sections. Only hostA and hostB are connected. Only hostA has
2238 sent data to hostB.
2239
2240 $ echo "show peers" | socat - /tmp/hostA
2241 0x55deb0224320: [15/Apr/2019:11:28:01] id=sharedlb state=0 flags=0x3 \
Emeric Brun0bbec0f2019-04-18 11:39:43 +02002242 resync_timeout=<PAST> task_calls=45122
Frédéric Lécaille21dde502019-04-15 13:50:23 +02002243 0x55deb022b540: id=hostC(remote) addr=127.0.0.12:10002 status=CONN \
2244 reconnect=4s confirm=0
2245 flags=0x0
2246 0x55deb022a440: id=hostA(local) addr=127.0.0.10:10000 status=NONE \
2247 reconnect=<NEVER> confirm=0
2248 flags=0x0
2249 0x55deb0227d70: id=hostB(remote) addr=127.0.0.11:10001 status=ESTA
2250 reconnect=2s confirm=0
Emeric Brun0bbec0f2019-04-18 11:39:43 +02002251 flags=0x20000200 appctx:0x55deb028fba0 st0=7 st1=0 task_calls=14456 \
2252 state=EST
Frédéric Lécaille21dde502019-04-15 13:50:23 +02002253 xprt=RAW src=127.0.0.1:37257 addr=127.0.0.10:10000
2254 remote_table:0x55deb0224a10 id=stkt local_id=1 remote_id=1
2255 last_local_table:0x55deb0224a10 id=stkt local_id=1 remote_id=1
2256 shared tables:
2257 0x55deb0224a10 local_id=1 remote_id=1 flags=0x0 remote_data=0x65
2258 last_acked=0 last_pushed=3 last_get=0 teaching_origin=0 update=3
2259 table:0x55deb022d6a0 id=stkt update=3 localupdate=3 \
2260 commitupdate=3 syncing=0
2261
2262 $ echo "show peers" | socat - /tmp/hostB
2263 0x55871b5ab320: [15/Apr/2019:11:28:03] id=sharedlb state=0 flags=0x3 \
Emeric Brun0bbec0f2019-04-18 11:39:43 +02002264 resync_timeout=<PAST> task_calls=3
Frédéric Lécaille21dde502019-04-15 13:50:23 +02002265 0x55871b5b2540: id=hostC(remote) addr=127.0.0.12:10002 status=CONN \
2266 reconnect=3s confirm=0
2267 flags=0x0
2268 0x55871b5b1440: id=hostB(local) addr=127.0.0.11:10001 status=NONE \
2269 reconnect=<NEVER> confirm=0
2270 flags=0x0
2271 0x55871b5aed70: id=hostA(remote) addr=127.0.0.10:10000 status=ESTA \
2272 reconnect=2s confirm=0
Emeric Brun0bbec0f2019-04-18 11:39:43 +02002273 flags=0x20000200 appctx:0x7fa46800ee00 st0=7 st1=0 task_calls=62356 \
2274 state=EST
Frédéric Lécaille21dde502019-04-15 13:50:23 +02002275 remote_table:0x55871b5ab960 id=stkt local_id=1 remote_id=1
2276 last_local_table:0x55871b5ab960 id=stkt local_id=1 remote_id=1
2277 shared tables:
2278 0x55871b5ab960 local_id=1 remote_id=1 flags=0x0 remote_data=0x65
2279 last_acked=3 last_pushed=0 last_get=3 teaching_origin=0 update=0
2280 table:0x55871b5b46a0 id=stkt update=1 localupdate=0 \
2281 commitupdate=0 syncing=0
2282
Willy Tarreau44aed902015-10-13 14:45:29 +02002283show pools
2284 Dump the status of internal memory pools. This is useful to track memory
2285 usage when suspecting a memory leak for example. It does exactly the same
2286 as the SIGQUIT when running in foreground except that it does not flush
2287 the pools.
2288
Willy Tarreau75c62c22018-11-22 11:02:09 +01002289show profiling
2290 Dumps the current profiling settings, one per line, as well as the command
2291 needed to change them.
2292
Willy Tarreau69f591e2020-07-01 07:00:59 +02002293show servers conn [<backend>]
2294 Dump the current and idle connections state of the servers belonging to the
2295 designated backend (or all backends if none specified). A backend name or
2296 identifier may be used.
2297
2298 The output consists in a header line showing the fields titles, then one
2299 server per line with for each, the backend name and ID, server name and ID,
2300 the address, port and a series or values. The number of fields varies
2301 depending on thread count.
2302
2303 Given the threaded nature of idle connections, it's important to understand
2304 that some values may change once read, and that as such, consistency within a
2305 line isn't granted. This output is mostly provided as a debugging tool and is
2306 not relevant to be routinely monitored nor graphed.
2307
Willy Tarreau44aed902015-10-13 14:45:29 +02002308show servers state [<backend>]
2309 Dump the state of the servers found in the running configuration. A backend
2310 name or identifier may be provided to limit the output to this backend only.
2311
2312 The dump has the following format:
2313 - first line contains the format version (1 in this specification);
2314 - second line contains the column headers, prefixed by a sharp ('#');
2315 - third line and next ones contain data;
2316 - each line starting by a sharp ('#') is considered as a comment.
2317
Dan Lloyd8e48b872016-07-01 21:01:18 -04002318 Since multiple versions of the output may co-exist, below is the list of
Willy Tarreau44aed902015-10-13 14:45:29 +02002319 fields and their order per file format version :
2320 1:
2321 be_id: Backend unique id.
2322 be_name: Backend label.
2323 srv_id: Server unique id (in the backend).
2324 srv_name: Server label.
2325 srv_addr: Server IP address.
2326 srv_op_state: Server operational state (UP/DOWN/...).
Cyril Bonté5b2ce8a2016-11-02 00:19:58 +01002327 0 = SRV_ST_STOPPED
2328 The server is down.
2329 1 = SRV_ST_STARTING
2330 The server is warming up (up but
2331 throttled).
2332 2 = SRV_ST_RUNNING
2333 The server is fully up.
2334 3 = SRV_ST_STOPPING
2335 The server is up but soft-stopping
2336 (eg: 404).
Willy Tarreau44aed902015-10-13 14:45:29 +02002337 srv_admin_state: Server administrative state (MAINT/DRAIN/...).
Cyril Bonté5b2ce8a2016-11-02 00:19:58 +01002338 The state is actually a mask of values :
2339 0x01 = SRV_ADMF_FMAINT
2340 The server was explicitly forced into
2341 maintenance.
2342 0x02 = SRV_ADMF_IMAINT
2343 The server has inherited the maintenance
2344 status from a tracked server.
2345 0x04 = SRV_ADMF_CMAINT
2346 The server is in maintenance because of
2347 the configuration.
2348 0x08 = SRV_ADMF_FDRAIN
2349 The server was explicitly forced into
2350 drain state.
2351 0x10 = SRV_ADMF_IDRAIN
2352 The server has inherited the drain status
2353 from a tracked server.
Baptiste Assmann89aa7f32016-11-02 21:31:27 +01002354 0x20 = SRV_ADMF_RMAINT
2355 The server is in maintenance because of an
2356 IP address resolution failure.
Frédéric Lécailleb418c122017-04-26 11:24:02 +02002357 0x40 = SRV_ADMF_HMAINT
2358 The server FQDN was set from stats socket.
2359
Willy Tarreau44aed902015-10-13 14:45:29 +02002360 srv_uweight: User visible server's weight.
2361 srv_iweight: Server's initial weight.
2362 srv_time_since_last_change: Time since last operational change.
2363 srv_check_status: Last health check status.
2364 srv_check_result: Last check result (FAILED/PASSED/...).
Cyril Bonté5b2ce8a2016-11-02 00:19:58 +01002365 0 = CHK_RES_UNKNOWN
2366 Initialized to this by default.
2367 1 = CHK_RES_NEUTRAL
2368 Valid check but no status information.
2369 2 = CHK_RES_FAILED
2370 Check failed.
2371 3 = CHK_RES_PASSED
2372 Check succeeded and server is fully up
2373 again.
2374 4 = CHK_RES_CONDPASS
2375 Check reports the server doesn't want new
2376 sessions.
Willy Tarreau44aed902015-10-13 14:45:29 +02002377 srv_check_health: Checks rise / fall current counter.
2378 srv_check_state: State of the check (ENABLED/PAUSED/...).
Cyril Bonté5b2ce8a2016-11-02 00:19:58 +01002379 The state is actually a mask of values :
2380 0x01 = CHK_ST_INPROGRESS
2381 A check is currently running.
2382 0x02 = CHK_ST_CONFIGURED
2383 This check is configured and may be
2384 enabled.
2385 0x04 = CHK_ST_ENABLED
2386 This check is currently administratively
2387 enabled.
2388 0x08 = CHK_ST_PAUSED
2389 Checks are paused because of maintenance
2390 (health only).
Willy Tarreau44aed902015-10-13 14:45:29 +02002391 srv_agent_state: State of the agent check (ENABLED/PAUSED/...).
Cyril Bonté5b2ce8a2016-11-02 00:19:58 +01002392 This state uses the same mask values as
2393 "srv_check_state", adding this specific one :
2394 0x10 = CHK_ST_AGENT
2395 Check is an agent check (otherwise it's a
2396 health check).
Willy Tarreau44aed902015-10-13 14:45:29 +02002397 bk_f_forced_id: Flag to know if the backend ID is forced by
2398 configuration.
2399 srv_f_forced_id: Flag to know if the server's ID is forced by
2400 configuration.
Frédéric Lécailleb418c122017-04-26 11:24:02 +02002401 srv_fqdn: Server FQDN.
Frédéric Lécaille31694712017-08-01 08:47:19 +02002402 srv_port: Server port.
Baptiste Assmann6d0f38f2018-07-02 17:00:54 +02002403 srvrecord: DNS SRV record associated to this SRV.
Willy Tarreau44aed902015-10-13 14:45:29 +02002404
2405show sess
2406 Dump all known sessions. Avoid doing this on slow connections as this can
2407 be huge. This command is restricted and can only be issued on sockets
Willy Tarreauc6e7a1b2020-06-28 01:24:12 +02002408 configured for levels "operator" or "admin". Note that on machines with
2409 quickly recycled connections, it is possible that this output reports less
2410 entries than really exist because it will dump all existing sessions up to
2411 the last one that was created before the command was entered; those which
2412 die in the mean time will not appear.
Willy Tarreau44aed902015-10-13 14:45:29 +02002413
2414show sess <id>
2415 Display a lot of internal information about the specified session identifier.
2416 This identifier is the first field at the beginning of the lines in the dumps
2417 of "show sess" (it corresponds to the session pointer). Those information are
2418 useless to most users but may be used by haproxy developers to troubleshoot a
2419 complex bug. The output format is intentionally not documented so that it can
2420 freely evolve depending on demands. You may find a description of all fields
2421 returned in src/dumpstats.c
2422
2423 The special id "all" dumps the states of all sessions, which must be avoided
2424 as much as possible as it is highly CPU intensive and can take a lot of time.
2425
Willy Tarreau6b19b142019-10-09 15:44:21 +02002426show stat [{<iid>|<proxy>} <type> <sid>] [typed|json] [desc]
Simon Horman05ee2132017-01-04 09:37:25 +01002427 Dump statistics using the CSV format; using the extended typed output
2428 format described in the section above if "typed" is passed after the
2429 other arguments; or in JSON if "json" is passed after the other arguments
2430 . By passing <id>, <type> and <sid>, it is possible to dump only selected
2431 items :
Willy Tarreaua1b1ed52016-11-25 08:50:58 +01002432 - <iid> is a proxy ID, -1 to dump everything. Alternatively, a proxy name
2433 <proxy> may be specified. In this case, this proxy's ID will be used as
2434 the ID selector.
Willy Tarreau44aed902015-10-13 14:45:29 +02002435 - <type> selects the type of dumpable objects : 1 for frontends, 2 for
2436 backends, 4 for servers, -1 for everything. These values can be ORed,
2437 for example:
2438 1 + 2 = 3 -> frontend + backend.
2439 1 + 2 + 4 = 7 -> frontend + backend + server.
2440 - <sid> is a server ID, -1 to dump everything from the selected proxy.
2441
2442 Example :
2443 $ echo "show info;show stat" | socat stdio unix-connect:/tmp/sock1
2444 >>> Name: HAProxy
2445 Version: 1.4-dev2-49
2446 Release_date: 2009/09/23
2447 Nbproc: 1
2448 Process_num: 1
2449 (...)
2450
2451 # pxname,svname,qcur,qmax,scur,smax,slim,stot,bin,bout,dreq, (...)
2452 stats,FRONTEND,,,0,0,1000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,1,0, (...)
2453 stats,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,0,0,0,,0,250,(...)
2454 (...)
2455 www1,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,1,1,0,,0,250, (...)
2456
2457 $
2458
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002459 In this example, two commands have been issued at once. That way it's easy to
2460 find which process the stats apply to in multi-process mode. This is not
2461 needed in the typed output format as the process number is reported on each
2462 line. Notice the empty line after the information output which marks the end
2463 of the first block. A similar empty line appears at the end of the second
2464 block (stats) so that the reader knows the output has not been truncated.
2465
2466 When "typed" is specified, the output format is more suitable to monitoring
2467 tools because it provides numeric positions and indicates the type of each
2468 output field. Each value stands on its own line with process number, element
2469 number, nature, origin and scope. This same format is available via the HTTP
2470 stats by passing ";typed" after the URI. It is very important to note that in
Dan Lloyd8e48b872016-07-01 21:01:18 -04002471 typed output format, the dump for a single object is contiguous so that there
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002472 is no need for a consumer to store everything at once.
2473
2474 When using the typed output format, each line is made of 4 columns delimited
2475 by colons (':'). The first column is a dot-delimited series of 5 elements. The
2476 first element is a letter indicating the type of the object being described.
2477 At the moment the following object types are known : 'F' for a frontend, 'B'
2478 for a backend, 'L' for a listener, and 'S' for a server. The second element
2479 The second element is a positive integer representing the unique identifier of
2480 the proxy the object belongs to. It is equivalent to the "iid" column of the
2481 CSV output and matches the value in front of the optional "id" directive found
2482 in the frontend or backend section. The third element is a positive integer
2483 containing the unique object identifier inside the proxy, and corresponds to
2484 the "sid" column of the CSV output. ID 0 is reported when dumping a frontend
2485 or a backend. For a listener or a server, this corresponds to their respective
2486 ID inside the proxy. The fourth element is the numeric position of the field
2487 in the list (starting at zero). This position shall not change over time, but
2488 holes are to be expected, depending on build options or if some fields are
2489 deleted in the future. The fifth element is the field name as it appears in
2490 the CSV output. The sixth element is a positive integer and is the relative
2491 process number starting at 1.
2492
2493 The rest of the line starting after the first colon follows the "typed output
2494 format" described in the section above. In short, the second column (after the
2495 first ':') indicates the origin, nature and scope of the variable. The third
2496 column indicates the type of the field, among "s32", "s64", "u32", "u64" and
2497 "str". Then the fourth column is the value itself, which the consumer knows
2498 how to parse thanks to column 3 and how to process thanks to column 2.
2499
Willy Tarreau6b19b142019-10-09 15:44:21 +02002500 When "desc" is appended to the command, one extra colon followed by a quoted
2501 string is appended with a description for the metric. At the time of writing,
2502 this is only supported for the "typed" output format.
2503
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002504 Thus the overall line format in typed mode is :
2505
2506 <obj>.<px_id>.<id>.<fpos>.<fname>.<process_num>:<tags>:<type>:<value>
2507
2508 Here's an example of typed output format :
2509
2510 $ echo "show stat typed" | socat stdio unix-connect:/tmp/sock1
2511 F.2.0.0.pxname.1:MGP:str:private-frontend
2512 F.2.0.1.svname.1:MGP:str:FRONTEND
2513 F.2.0.8.bin.1:MGP:u64:0
2514 F.2.0.9.bout.1:MGP:u64:0
2515 F.2.0.40.hrsp_2xx.1:MGP:u64:0
2516 L.2.1.0.pxname.1:MGP:str:private-frontend
2517 L.2.1.1.svname.1:MGP:str:sock-1
2518 L.2.1.17.status.1:MGP:str:OPEN
2519 L.2.1.73.addr.1:MGP:str:0.0.0.0:8001
2520 S.3.13.60.rtime.1:MCP:u32:0
2521 S.3.13.61.ttime.1:MCP:u32:0
2522 S.3.13.62.agent_status.1:MGP:str:L4TOUT
2523 S.3.13.64.agent_duration.1:MGP:u64:2001
2524 S.3.13.65.check_desc.1:MCP:str:Layer4 timeout
2525 S.3.13.66.agent_desc.1:MCP:str:Layer4 timeout
2526 S.3.13.67.check_rise.1:MCP:u32:2
2527 S.3.13.68.check_fall.1:MCP:u32:3
2528 S.3.13.69.check_health.1:SGP:u32:0
2529 S.3.13.70.agent_rise.1:MaP:u32:1
2530 S.3.13.71.agent_fall.1:SGP:u32:1
2531 S.3.13.72.agent_health.1:SGP:u32:1
2532 S.3.13.73.addr.1:MCP:str:1.255.255.255:8888
2533 S.3.13.75.mode.1:MAP:str:http
2534 B.3.0.0.pxname.1:MGP:str:private-backend
2535 B.3.0.1.svname.1:MGP:str:BACKEND
2536 B.3.0.2.qcur.1:MGP:u32:0
2537 B.3.0.3.qmax.1:MGP:u32:0
2538 B.3.0.4.scur.1:MGP:u32:0
2539 B.3.0.5.smax.1:MGP:u32:0
2540 B.3.0.6.slim.1:MGP:u32:1000
2541 B.3.0.55.lastsess.1:MMP:s32:-1
2542 (...)
2543
Simon Horman1084a362016-11-21 17:00:24 +01002544 In the typed format, the presence of the process ID at the end of the
2545 first column makes it very easy to visually aggregate outputs from
2546 multiple processes, as show in the example below where each line appears
2547 for each process :
Willy Tarreau5d8b9792016-03-11 11:09:34 +01002548
2549 $ ( echo show stat typed | socat /var/run/haproxy.sock1 - ; \
2550 echo show stat typed | socat /var/run/haproxy.sock2 - ) | \
2551 sort -t . -k 1,1 -k 2,2n -k 3,3n -k 4,4n -k 5,5 -k 6,6n
2552 B.3.0.0.pxname.1:MGP:str:private-backend
2553 B.3.0.0.pxname.2:MGP:str:private-backend
2554 B.3.0.1.svname.1:MGP:str:BACKEND
2555 B.3.0.1.svname.2:MGP:str:BACKEND
2556 B.3.0.2.qcur.1:MGP:u32:0
2557 B.3.0.2.qcur.2:MGP:u32:0
2558 B.3.0.3.qmax.1:MGP:u32:0
2559 B.3.0.3.qmax.2:MGP:u32:0
2560 B.3.0.4.scur.1:MGP:u32:0
2561 B.3.0.4.scur.2:MGP:u32:0
2562 B.3.0.5.smax.1:MGP:u32:0
2563 B.3.0.5.smax.2:MGP:u32:0
2564 B.3.0.6.slim.1:MGP:u32:1000
2565 B.3.0.6.slim.2:MGP:u32:1000
2566 (...)
Willy Tarreau44aed902015-10-13 14:45:29 +02002567
Simon Horman05ee2132017-01-04 09:37:25 +01002568 The format of JSON output is described in a schema which may be output
Simon Horman6f6bb382017-01-04 09:37:26 +01002569 using "show schema json".
2570
2571 The JSON output contains no extra whitespace in order to reduce the
2572 volume of output. For human consumption passing the output through a
2573 pretty printer may be helpful. Example :
2574
2575 $ echo "show stat json" | socat /var/run/haproxy.sock stdio | \
2576 python -m json.tool
Simon Horman05ee2132017-01-04 09:37:25 +01002577
2578 The JSON output contains no extra whitespace in order to reduce the
2579 volume of output. For human consumption passing the output through a
2580 pretty printer may be helpful. Example :
2581
2582 $ echo "show stat json" | socat /var/run/haproxy.sock stdio | \
2583 python -m json.tool
2584
William Lallemandd4f946c2019-12-05 10:26:40 +01002585show ssl cert [<filename>]
Ilya Shipitsin2a950d02020-03-06 13:07:38 +05002586 Display the list of certificates used on frontends. If a filename is prefixed
2587 by an asterisk, it is a transaction which is not committed yet. If a
William Lallemandd4f946c2019-12-05 10:26:40 +01002588 filename is specified, it will show details about the certificate. This
2589 command can be useful to check if a certificate was well updated. You can
2590 also display details on a transaction by prefixing the filename by an
2591 asterisk.
2592
2593 Example :
2594
2595 $ echo "@1 show ssl cert" | socat /var/run/haproxy.master -
2596 # transaction
2597 *test.local.pem
2598 # filename
2599 test.local.pem
2600
2601 $ echo "@1 show ssl cert test.local.pem" | socat /var/run/haproxy.master -
2602 Filename: test.local.pem
2603 Serial: 03ECC19BA54B25E85ABA46EE561B9A10D26F
2604 notBefore: Sep 13 21:20:24 2019 GMT
2605 notAfter: Dec 12 21:20:24 2019 GMT
2606 Issuer: /C=US/O=Let's Encrypt/CN=Let's Encrypt Authority X3
2607 Subject: /CN=test.local
2608 Subject Alternative Name: DNS:test.local, DNS:imap.test.local
2609 Algorithm: RSA2048
2610 SHA1 FingerPrint: 417A11CAE25F607B24F638B4A8AEE51D1E211477
2611
2612 $ echo "@1 show ssl cert *test.local.pem" | socat /var/run/haproxy.master -
2613 Filename: *test.local.pem
2614 [...]
2615
William Lallemandc69f02d2020-04-06 19:07:03 +02002616show ssl crt-list [-n] [<filename>]
William Lallemandaccac232020-04-02 17:42:51 +02002617 Display the list of crt-list and directories used in the HAProxy
William Lallemandc69f02d2020-04-06 19:07:03 +02002618 configuration. If a filename is specified, dump the content of a crt-list or
2619 a directory. Once dumped the output can be used as a crt-list file.
2620 The '-n' option can be used to display the line number, which is useful when
2621 combined with the 'del ssl crt-list' option when a entry is duplicated. The
2622 output with the '-n' option is not compatible with the crt-list format and
2623 not loadable by haproxy.
William Lallemandaccac232020-04-02 17:42:51 +02002624
2625 Example:
William Lallemandc69f02d2020-04-06 19:07:03 +02002626 echo "show ssl crt-list -n localhost.crt-list" | socat /tmp/sock1 -
William Lallemandaccac232020-04-02 17:42:51 +02002627 # localhost.crt-list
William Lallemandc69f02d2020-04-06 19:07:03 +02002628 common.pem:1 !not.test1.com *.test1.com !localhost
2629 common.pem:2
2630 ecdsa.pem:3 [verify none allow-0rtt ssl-min-ver TLSv1.0 ssl-max-ver TLSv1.3] localhost !www.test1.com
2631 ecdsa.pem:4 [verify none allow-0rtt ssl-min-ver TLSv1.0 ssl-max-ver TLSv1.3]
William Lallemandaccac232020-04-02 17:42:51 +02002632
Christopher Faulet78c43062019-09-27 10:45:47 +02002633show resolvers [<resolvers section id>]
Willy Tarreau44aed902015-10-13 14:45:29 +02002634 Dump statistics for the given resolvers section, or all resolvers sections
2635 if no section is supplied.
2636
2637 For each name server, the following counters are reported:
2638 sent: number of DNS requests sent to this server
2639 valid: number of DNS valid responses received from this server
2640 update: number of DNS responses used to update the server's IP address
2641 cname: number of CNAME responses
2642 cname_error: CNAME errors encountered with this server
2643 any_err: number of empty response (IE: server does not support ANY type)
2644 nx: non existent domain response received from this server
2645 timeout: how many time this server did not answer in time
2646 refused: number of requests refused by this server
2647 other: any other DNS errors
2648 invalid: invalid DNS response (from a protocol point of view)
2649 too_big: too big response
2650 outdated: number of response arrived too late (after an other name server)
2651
2652show table
2653 Dump general information on all known stick-tables. Their name is returned
2654 (the name of the proxy which holds them), their type (currently zero, always
2655 IP), their size in maximum possible number of entries, and the number of
2656 entries currently in use.
2657
2658 Example :
2659 $ echo "show table" | socat stdio /tmp/sock1
2660 >>> # table: front_pub, type: ip, size:204800, used:171454
2661 >>> # table: back_rdp, type: ip, size:204800, used:0
2662
Adis Nezirovic1a693fc2020-01-16 15:19:29 +01002663show table <name> [ data.<type> <operator> <value> [data.<type> ...]] | [ key <key> ]
Willy Tarreau44aed902015-10-13 14:45:29 +02002664 Dump contents of stick-table <name>. In this mode, a first line of generic
2665 information about the table is reported as with "show table", then all
2666 entries are dumped. Since this can be quite heavy, it is possible to specify
2667 a filter in order to specify what entries to display.
2668
2669 When the "data." form is used the filter applies to the stored data (see
2670 "stick-table" in section 4.2). A stored data type must be specified
2671 in <type>, and this data type must be stored in the table otherwise an
2672 error is reported. The data is compared according to <operator> with the
2673 64-bit integer <value>. Operators are the same as with the ACLs :
2674
2675 - eq : match entries whose data is equal to this value
2676 - ne : match entries whose data is not equal to this value
2677 - le : match entries whose data is less than or equal to this value
2678 - ge : match entries whose data is greater than or equal to this value
2679 - lt : match entries whose data is less than this value
2680 - gt : match entries whose data is greater than this value
2681
Adis Nezirovic1a693fc2020-01-16 15:19:29 +01002682 In this form, you can use multiple data filter entries, up to a maximum
2683 defined during build time (4 by default).
Willy Tarreau44aed902015-10-13 14:45:29 +02002684
2685 When the key form is used the entry <key> is shown. The key must be of the
2686 same type as the table, which currently is limited to IPv4, IPv6, integer,
2687 and string.
2688
2689 Example :
2690 $ echo "show table http_proxy" | socat stdio /tmp/sock1
2691 >>> # table: http_proxy, type: ip, size:204800, used:2
2692 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
2693 bytes_out_rate(60000)=187
2694 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
2695 bytes_out_rate(60000)=191
2696
2697 $ echo "show table http_proxy data.gpc0 gt 0" | socat stdio /tmp/sock1
2698 >>> # table: http_proxy, type: ip, size:204800, used:2
2699 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
2700 bytes_out_rate(60000)=191
2701
2702 $ echo "show table http_proxy data.conn_rate gt 5" | \
2703 socat stdio /tmp/sock1
2704 >>> # table: http_proxy, type: ip, size:204800, used:2
2705 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
2706 bytes_out_rate(60000)=191
2707
2708 $ echo "show table http_proxy key 127.0.0.2" | \
2709 socat stdio /tmp/sock1
2710 >>> # table: http_proxy, type: ip, size:204800, used:2
2711 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
2712 bytes_out_rate(60000)=191
2713
2714 When the data criterion applies to a dynamic value dependent on time such as
2715 a bytes rate, the value is dynamically computed during the evaluation of the
2716 entry in order to decide whether it has to be dumped or not. This means that
2717 such a filter could match for some time then not match anymore because as
2718 time goes, the average event rate drops.
2719
2720 It is possible to use this to extract lists of IP addresses abusing the
2721 service, in order to monitor them or even blacklist them in a firewall.
2722 Example :
2723 $ echo "show table http_proxy data.gpc0 gt 0" \
2724 | socat stdio /tmp/sock1 \
2725 | fgrep 'key=' | cut -d' ' -f2 | cut -d= -f2 > abusers-ip.txt
2726 ( or | awk '/key/{ print a[split($2,a,"=")]; }' )
2727
Willy Tarreau4e2b6462019-05-16 17:44:30 +02002728show threads
2729 Dumps some internal states and structures for each thread, that may be useful
2730 to help developers understand a problem. The output tries to be readable by
Willy Tarreauc7091d82019-05-17 10:08:49 +02002731 showing one block per thread. When haproxy is built with USE_THREAD_DUMP=1,
2732 an advanced dump mechanism involving thread signals is used so that each
2733 thread can dump its own state in turn. Without this option, the thread
2734 processing the command shows all its details but the other ones are less
Willy Tarreaue6a02fa2019-05-22 07:06:44 +02002735 detailed. A star ('*') is displayed in front of the thread handling the
2736 command. A right angle bracket ('>') may also be displayed in front of
2737 threads which didn't make any progress since last invocation of this command,
2738 indicating a bug in the code which must absolutely be reported. When this
2739 happens between two threads it usually indicates a deadlock. If a thread is
2740 alone, it's a different bug like a corrupted list. In all cases the process
2741 needs is not fully functional anymore and needs to be restarted.
2742
2743 The output format is purposely not documented so that it can easily evolve as
2744 new needs are identified, without having to maintain any form of backwards
2745 compatibility, and just like with "show activity", the values are meaningless
2746 without the code at hand.
Willy Tarreau4e2b6462019-05-16 17:44:30 +02002747
William Lallemandbb933462016-05-31 21:09:53 +02002748show tls-keys [id|*]
2749 Dump all loaded TLS ticket keys references. The TLS ticket key reference ID
2750 and the file from which the keys have been loaded is shown. Both of those
2751 can be used to update the TLS keys using "set ssl tls-key". If an ID is
2752 specified as parameter, it will dump the tickets, using * it will dump every
2753 keys from every references.
Willy Tarreau44aed902015-10-13 14:45:29 +02002754
Simon Horman6f6bb382017-01-04 09:37:26 +01002755show schema json
2756 Dump the schema used for the output of "show info json" and "show stat json".
2757
2758 The contains no extra whitespace in order to reduce the volume of output.
2759 For human consumption passing the output through a pretty printer may be
2760 helpful. Example :
2761
2762 $ echo "show schema json" | socat /var/run/haproxy.sock stdio | \
2763 python -m json.tool
2764
2765 The schema follows "JSON Schema" (json-schema.org) and accordingly
2766 verifiers may be used to verify the output of "show info json" and "show
2767 stat json" against the schema.
2768
Willy Tarreauf909c912019-08-22 20:06:04 +02002769show trace [<source>]
2770 Show the current trace status. For each source a line is displayed with a
2771 single-character status indicating if the trace is stopped, waiting, or
2772 running. The output sink used by the trace is indicated (or "none" if none
2773 was set), as well as the number of dropped events in this sink, followed by a
2774 brief description of the source. If a source name is specified, a detailed
2775 list of all events supported by the source, and their status for each action
2776 (report, start, pause, stop), indicated by a "+" if they are enabled, or a
2777 "-" otherwise. All these events are independent and an event might trigger
2778 a start without being reported and conversely.
Simon Horman6f6bb382017-01-04 09:37:26 +01002779
Willy Tarreau44aed902015-10-13 14:45:29 +02002780shutdown frontend <frontend>
2781 Completely delete the specified frontend. All the ports it was bound to will
2782 be released. It will not be possible to enable the frontend anymore after
2783 this operation. This is intended to be used in environments where stopping a
2784 proxy is not even imaginable but a misconfigured proxy must be fixed. That
2785 way it's possible to release the port and bind it into another process to
2786 restore operations. The frontend will not appear at all on the stats page
2787 once it is terminated.
2788
2789 The frontend may be specified either by its name or by its numeric ID,
2790 prefixed with a sharp ('#').
2791
2792 This command is restricted and can only be issued on sockets configured for
2793 level "admin".
2794
2795shutdown session <id>
2796 Immediately terminate the session matching the specified session identifier.
2797 This identifier is the first field at the beginning of the lines in the dumps
2798 of "show sess" (it corresponds to the session pointer). This can be used to
2799 terminate a long-running session without waiting for a timeout or when an
2800 endless transfer is ongoing. Such terminated sessions are reported with a 'K'
2801 flag in the logs.
2802
2803shutdown sessions server <backend>/<server>
2804 Immediately terminate all the sessions attached to the specified server. This
2805 can be used to terminate long-running sessions after a server is put into
2806 maintenance mode, for instance. Such terminated sessions are reported with a
2807 'K' flag in the logs.
2808
Willy Tarreauf909c912019-08-22 20:06:04 +02002809trace
2810 The "trace" command alone lists the trace sources, their current status, and
2811 their brief descriptions. It is only meant as a menu to enter next levels,
2812 see other "trace" commands below.
2813
2814trace 0
2815 Immediately stops all traces. This is made to be used as a quick solution
2816 to terminate a debugging session or as an emergency action to be used in case
2817 complex traces were enabled on multiple sources and impact the service.
2818
2819trace <source> event [ [+|-|!]<name> ]
2820 Without argument, this will list all the events supported by the designated
2821 source. They are prefixed with a "-" if they are not enabled, or a "+" if
2822 they are enabled. It is important to note that a single trace may be labelled
2823 with multiple events, and as long as any of the enabled events matches one of
2824 the events labelled on the trace, the event will be passed to the trace
2825 subsystem. For example, receiving an HTTP/2 frame of type HEADERS may trigger
2826 a frame event and a stream event since the frame creates a new stream. If
2827 either the frame event or the stream event are enabled for this source, the
2828 frame will be passed to the trace framework.
2829
2830 With an argument, it is possible to toggle the state of each event and
2831 individually enable or disable them. Two special keywords are supported,
2832 "none", which matches no event, and is used to disable all events at once,
2833 and "any" which matches all events, and is used to enable all events at
2834 once. Other events are specific to the event source. It is possible to
2835 enable one event by specifying its name, optionally prefixed with '+' for
2836 better readability. It is possible to disable one event by specifying its
2837 name prefixed by a '-' or a '!'.
2838
2839 One way to completely disable a trace source is to pass "event none", and
2840 this source will instantly be totally ignored.
2841
2842trace <source> level [<level>]
Willy Tarreau2ea549b2019-08-29 08:01:48 +02002843 Without argument, this will list all trace levels for this source, and the
Willy Tarreauf909c912019-08-22 20:06:04 +02002844 current one will be indicated by a star ('*') prepended in front of it. With
Willy Tarreau2ea549b2019-08-29 08:01:48 +02002845 an argument, this will change the trace level to the specified level. Detail
Willy Tarreauf909c912019-08-22 20:06:04 +02002846 levels are a form of filters that are applied before reporting the events.
Willy Tarreau2ea549b2019-08-29 08:01:48 +02002847 These filters are used to selectively include or exclude events depending on
2848 their level of importance. For example a developer might need to know
2849 precisely where in the code an HTTP header was considered invalid while the
2850 end user may not even care about this header's validity at all. There are
2851 currently 5 distinct levels for a trace :
Willy Tarreauf909c912019-08-22 20:06:04 +02002852
2853 user this will report information that are suitable for use by a
2854 regular haproxy user who wants to observe his traffic.
2855 Typically some HTTP requests and responses will be reported
2856 without much detail. Most sources will set this as the
2857 default level to ease operations.
2858
Willy Tarreau2ea549b2019-08-29 08:01:48 +02002859 proto in addition to what is reported at the "user" level, it also
2860 displays protocol-level updates. This can for example be the
2861 frame types or HTTP headers after decoding.
Willy Tarreauf909c912019-08-22 20:06:04 +02002862
2863 state in addition to what is reported at the "proto" level, it
2864 will also display state transitions (or failed transitions)
2865 which happen in parsers, so this will show attempts to
2866 perform an operation while the "proto" level only shows
2867 the final operation.
2868
Willy Tarreau2ea549b2019-08-29 08:01:48 +02002869 data in addition to what is reported at the "state" level, it
2870 will also include data transfers between the various layers.
2871
Willy Tarreauf909c912019-08-22 20:06:04 +02002872 developer it reports everything available, which can include advanced
2873 information such as "breaking out of this loop" that are
2874 only relevant to a developer trying to understand a bug that
Willy Tarreau09fb0df2019-08-29 08:40:59 +02002875 only happens once in a while in field. Function names are
2876 only reported at this level.
Willy Tarreauf909c912019-08-22 20:06:04 +02002877
2878 It is highly recommended to always use the "user" level only and switch to
2879 other levels only if instructed to do so by a developer. Also it is a good
2880 idea to first configure the events before switching to higher levels, as it
2881 may save from dumping many lines if no filter is applied.
2882
2883trace <source> lock [criterion]
2884 Without argument, this will list all the criteria supported by this source
2885 for lock-on processing, and display the current choice by a star ('*') in
2886 front of it. Lock-on means that the source will focus on the first matching
2887 event and only stick to the criterion which triggered this event, and ignore
2888 all other ones until the trace stops. This allows for example to take a trace
2889 on a single connection or on a single stream. The following criteria are
2890 supported by some traces, though not necessarily all, since some of them
2891 might not be available to the source :
2892
2893 backend lock on the backend that started the trace
2894 connection lock on the connection that started the trace
2895 frontend lock on the frontend that started the trace
2896 listener lock on the listener that started the trace
2897 nothing do not lock on anything
2898 server lock on the server that started the trace
2899 session lock on the session that started the trace
2900 thread lock on the thread that started the trace
2901
2902 In addition to this, each source may provide up to 4 specific criteria such
2903 as internal states or connection IDs. For example in HTTP/2 it is possible
2904 to lock on the H2 stream and ignore other streams once a strace starts.
2905
2906 When a criterion is passed in argument, this one is used instead of the
2907 other ones and any existing tracking is immediately terminated so that it can
2908 restart with the new criterion. The special keyword "nothing" is supported by
2909 all sources to permanently disable tracking.
2910
2911trace <source> { pause | start | stop } [ [+|-|!]event]
2912 Without argument, this will list the events enabled to automatically pause,
2913 start, or stop a trace for this source. These events are specific to each
2914 trace source. With an argument, this will either enable the event for the
2915 specified action (if optionally prefixed by a '+') or disable it (if
2916 prefixed by a '-' or '!'). The special keyword "now" is not an event and
2917 requests to take the action immediately. The keywords "none" and "any" are
2918 supported just like in "trace event".
2919
2920 The 3 supported actions are respectively "pause", "start" and "stop". The
2921 "pause" action enumerates events which will cause a running trace to stop and
2922 wait for a new start event to restart it. The "start" action enumerates the
2923 events which switch the trace into the waiting mode until one of the start
2924 events appears. And the "stop" action enumerates the events which definitely
2925 stop the trace until it is manually enabled again. In practice it makes sense
2926 to manually start a trace using "start now" without caring about events, and
2927 to stop it using "stop now". In order to capture more subtle event sequences,
2928 setting "start" to a normal event (like receiving an HTTP request) and "stop"
2929 to a very rare event like emitting a certain error, will ensure that the last
2930 captured events will match the desired criteria. And the pause event is
2931 useful to detect the end of a sequence, disable the lock-on and wait for
2932 another opportunity to take a capture. In this case it can make sense to
2933 enable lock-on to spot only one specific criterion (e.g. a stream), and have
2934 "start" set to anything that starts this criterion (e.g. all events which
2935 create a stream), "stop" set to the expected anomaly, and "pause" to anything
2936 that ends that criterion (e.g. any end of stream event). In this case the
2937 trace log will contain complete sequences of perfectly clean series affecting
2938 a single object, until the last sequence containing everything from the
2939 beginning to the anomaly.
2940
2941trace <source> sink [<sink>]
2942 Without argument, this will list all event sinks available for this source,
2943 and the currently configured one will have a star ('*') prepended in front
2944 of it. Sink "none" is always available and means that all events are simply
2945 dropped, though their processing is not ignored (e.g. lock-on does occur).
2946 Other sinks are available depending on configuration and build options, but
2947 typically "stdout" and "stderr" will be usable in debug mode, and in-memory
2948 ring buffers should be available as well. When a name is specified, the sink
2949 instantly changes for the specified source. Events are not changed during a
2950 sink change. In the worst case some may be lost if an invalid sink is used
2951 (or "none"), but operations do continue to a different destination.
2952
Willy Tarreau370a6942019-08-29 08:24:16 +02002953trace <source> verbosity [<level>]
2954 Without argument, this will list all verbosity levels for this source, and the
2955 current one will be indicated by a star ('*') prepended in front of it. With
2956 an argument, this will change the verbosity level to the specified one.
2957
2958 Verbosity levels indicate how far the trace decoder should go to provide
2959 detailed information. It depends on the trace source, since some sources will
2960 not even provide a specific decoder. Level "quiet" is always available and
2961 disables any decoding. It can be useful when trying to figure what's
2962 happening before trying to understand the details, since it will have a very
2963 low impact on performance and trace size. When no verbosity levels are
2964 declared by a source, level "default" is available and will cause a decoder
2965 to be called when specified in the traces. It is an opportunistic decoding.
2966 When the source declares some verbosity levels, these ones are listed with
2967 a description of what they correspond to. In this case the trace decoder
2968 provided by the source will be as accurate as possible based on the
2969 information available at the trace point. The first level above "quiet" is
2970 set by default.
2971
Willy Tarreau2212e6a2015-10-13 14:40:55 +02002972
William Lallemand142db372018-12-11 18:56:45 +010029739.4. Master CLI
2974---------------
2975
2976The master CLI is a socket bound to the master process in master-worker mode.
2977This CLI gives access to the unix socket commands in every running or leaving
2978processes and allows a basic supervision of those processes.
2979
2980The master CLI is configurable only from the haproxy program arguments with
2981the -S option. This option also takes bind options separated by commas.
2982
2983Example:
2984
2985 # haproxy -W -S 127.0.0.1:1234 -f test1.cfg
2986 # haproxy -Ws -S /tmp/master-socket,uid,1000,gid,1000,mode,600 -f test1.cfg
William Lallemandb7ea1412018-12-13 09:05:47 +01002987 # haproxy -W -S /tmp/master-socket,level,user -f test1.cfg
William Lallemand142db372018-12-11 18:56:45 +01002988
2989The master CLI introduces a new 'show proc' command to surpervise the
2990processes:
2991
2992Example:
2993
2994 $ echo 'show proc' | socat /var/run/haproxy-master.sock -
William Lallemand1dc69632019-06-12 19:11:33 +02002995 #<PID> <type> <relative PID> <reloads> <uptime> <version>
2996 1162 master 0 5 0d00h02m07s 2.0-dev7-0124c9-7
William Lallemand142db372018-12-11 18:56:45 +01002997 # workers
William Lallemand1dc69632019-06-12 19:11:33 +02002998 1271 worker 1 0 0d00h00m00s 2.0-dev7-0124c9-7
2999 1272 worker 2 0 0d00h00m00s 2.0-dev7-0124c9-7
William Lallemand142db372018-12-11 18:56:45 +01003000 # old workers
William Lallemand1dc69632019-06-12 19:11:33 +02003001 1233 worker [was: 1] 3 0d00h00m43s 2.0-dev3-6019f6-289
William Lallemand142db372018-12-11 18:56:45 +01003002
3003
3004In this example, the master has been reloaded 5 times but one of the old
3005worker is still running and survived 3 reloads. You could access the CLI of
3006this worker to understand what's going on.
3007
Willy Tarreau52880f92018-12-15 13:30:03 +01003008When the prompt is enabled (via the "prompt" command), the context the CLI is
3009working on is displayed in the prompt. The master is identified by the "master"
3010string, and other processes are identified with their PID. In case the last
3011reload failed, the master prompt will be changed to "master[ReloadFailed]>" so
3012that it becomes visible that the process is still running on the previous
3013configuration and that the new configuration is not operational.
3014
William Lallemand142db372018-12-11 18:56:45 +01003015The master CLI uses a special prefix notation to access the multiple
3016processes. This notation is easily identifiable as it begins by a @.
3017
3018A @ prefix can be followed by a relative process number or by an exclamation
3019point and a PID. (e.g. @1 or @!1271). A @ alone could be use to specify the
3020master. Leaving processes are only accessible with the PID as relative process
3021number are only usable with the current processes.
3022
3023Examples:
3024
3025 $ socat /var/run/haproxy-master.sock readline
3026 prompt
3027 master> @1 show info; @2 show info
3028 [...]
3029 Process_num: 1
3030 Pid: 1271
3031 [...]
3032 Process_num: 2
3033 Pid: 1272
3034 [...]
3035 master>
3036
3037 $ echo '@!1271 show info; @!1272 show info' | socat /var/run/haproxy-master.sock -
3038 [...]
3039
3040A prefix could be use as a command, which will send every next commands to
3041the specified process.
3042
3043Examples:
3044
3045 $ socat /var/run/haproxy-master.sock readline
3046 prompt
3047 master> @1
3048 1271> show info
3049 [...]
3050 1271> show stat
3051 [...]
3052 1271> @
3053 master>
3054
3055 $ echo '@1; show info; show stat; @2; show info; show stat' | socat /var/run/haproxy-master.sock -
3056 [...]
3057
William Lallemanda57b7e32018-12-14 21:11:31 +01003058You can also reload the HAProxy master process with the "reload" command which
3059does the same as a `kill -USR2` on the master process, provided that the user
3060has at least "operator" or "admin" privileges.
3061
3062Example:
3063
3064 $ echo "reload" | socat /var/run/haproxy-master.sock
3065
3066Note that a reload will close the connection to the master CLI.
3067
William Lallemand142db372018-12-11 18:56:45 +01003068
Willy Tarreau2212e6a2015-10-13 14:40:55 +0200306910. Tricks for easier configuration management
3070----------------------------------------------
3071
3072It is very common that two HAProxy nodes constituting a cluster share exactly
3073the same configuration modulo a few addresses. Instead of having to maintain a
3074duplicate configuration for each node, which will inevitably diverge, it is
3075possible to include environment variables in the configuration. Thus multiple
3076configuration may share the exact same file with only a few different system
3077wide environment variables. This started in version 1.5 where only addresses
3078were allowed to include environment variables, and 1.6 goes further by
3079supporting environment variables everywhere. The syntax is the same as in the
3080UNIX shell, a variable starts with a dollar sign ('$'), followed by an opening
3081curly brace ('{'), then the variable name followed by the closing brace ('}').
3082Except for addresses, environment variables are only interpreted in arguments
3083surrounded with double quotes (this was necessary not to break existing setups
3084using regular expressions involving the dollar symbol).
3085
3086Environment variables also make it convenient to write configurations which are
3087expected to work on various sites where only the address changes. It can also
3088permit to remove passwords from some configs. Example below where the the file
3089"site1.env" file is sourced by the init script upon startup :
3090
3091 $ cat site1.env
3092 LISTEN=192.168.1.1
3093 CACHE_PFX=192.168.11
3094 SERVER_PFX=192.168.22
3095 LOGGER=192.168.33.1
3096 STATSLP=admin:pa$$w0rd
3097 ABUSERS=/etc/haproxy/abuse.lst
3098 TIMEOUT=10s
3099
3100 $ cat haproxy.cfg
3101 global
3102 log "${LOGGER}:514" local0
3103
3104 defaults
3105 mode http
3106 timeout client "${TIMEOUT}"
3107 timeout server "${TIMEOUT}"
3108 timeout connect 5s
3109
3110 frontend public
3111 bind "${LISTEN}:80"
3112 http-request reject if { src -f "${ABUSERS}" }
3113 stats uri /stats
3114 stats auth "${STATSLP}"
3115 use_backend cache if { path_end .jpg .css .ico }
3116 default_backend server
3117
3118 backend cache
3119 server cache1 "${CACHE_PFX}.1:18080" check
3120 server cache2 "${CACHE_PFX}.2:18080" check
3121
3122 backend server
3123 server cache1 "${SERVER_PFX}.1:8080" check
3124 server cache2 "${SERVER_PFX}.2:8080" check
3125
3126
312711. Well-known traps to avoid
3128-----------------------------
3129
3130Once in a while, someone reports that after a system reboot, the haproxy
3131service wasn't started, and that once they start it by hand it works. Most
3132often, these people are running a clustered IP address mechanism such as
3133keepalived, to assign the service IP address to the master node only, and while
3134it used to work when they used to bind haproxy to address 0.0.0.0, it stopped
3135working after they bound it to the virtual IP address. What happens here is
3136that when the service starts, the virtual IP address is not yet owned by the
3137local node, so when HAProxy wants to bind to it, the system rejects this
3138because it is not a local IP address. The fix doesn't consist in delaying the
3139haproxy service startup (since it wouldn't stand a restart), but instead to
3140properly configure the system to allow binding to non-local addresses. This is
3141easily done on Linux by setting the net.ipv4.ip_nonlocal_bind sysctl to 1. This
3142is also needed in order to transparently intercept the IP traffic that passes
3143through HAProxy for a specific target address.
3144
3145Multi-process configurations involving source port ranges may apparently seem
3146to work but they will cause some random failures under high loads because more
3147than one process may try to use the same source port to connect to the same
3148server, which is not possible. The system will report an error and a retry will
3149happen, picking another port. A high value in the "retries" parameter may hide
3150the effect to a certain extent but this also comes with increased CPU usage and
3151processing time. Logs will also report a certain number of retries. For this
3152reason, port ranges should be avoided in multi-process configurations.
3153
Dan Lloyd8e48b872016-07-01 21:01:18 -04003154Since HAProxy uses SO_REUSEPORT and supports having multiple independent
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003155processes bound to the same IP:port, during troubleshooting it can happen that
3156an old process was not stopped before a new one was started. This provides
3157absurd test results which tend to indicate that any change to the configuration
3158is ignored. The reason is that in fact even the new process is restarted with a
3159new configuration, the old one also gets some incoming connections and
3160processes them, returning unexpected results. When in doubt, just stop the new
3161process and try again. If it still works, it very likely means that an old
3162process remains alive and has to be stopped. Linux's "netstat -lntp" is of good
3163help here.
3164
3165When adding entries to an ACL from the command line (eg: when blacklisting a
3166source address), it is important to keep in mind that these entries are not
3167synchronized to the file and that if someone reloads the configuration, these
3168updates will be lost. While this is often the desired effect (for blacklisting)
3169it may not necessarily match expectations when the change was made as a fix for
3170a problem. See the "add acl" action of the CLI interface.
3171
3172
317312. Debugging and performance issues
3174------------------------------------
3175
3176When HAProxy is started with the "-d" option, it will stay in the foreground
3177and will print one line per event, such as an incoming connection, the end of a
3178connection, and for each request or response header line seen. This debug
3179output is emitted before the contents are processed, so they don't consider the
3180local modifications. The main use is to show the request and response without
3181having to run a network sniffer. The output is less readable when multiple
3182connections are handled in parallel, though the "debug2ansi" and "debug2html"
3183scripts found in the examples/ directory definitely help here by coloring the
3184output.
3185
3186If a request or response is rejected because HAProxy finds it is malformed, the
3187best thing to do is to connect to the CLI and issue "show errors", which will
3188report the last captured faulty request and response for each frontend and
3189backend, with all the necessary information to indicate precisely the first
3190character of the input stream that was rejected. This is sometimes needed to
3191prove to customers or to developers that a bug is present in their code. In
3192this case it is often possible to relax the checks (but still keep the
3193captures) using "option accept-invalid-http-request" or its equivalent for
3194responses coming from the server "option accept-invalid-http-response". Please
3195see the configuration manual for more details.
3196
3197Example :
3198
3199 > show errors
3200 Total events captured on [13/Oct/2015:13:43:47.169] : 1
3201
3202 [13/Oct/2015:13:43:40.918] frontend HAProxyLocalStats (#2): invalid request
3203 backend <NONE> (#-1), server <NONE> (#-1), event #0
3204 src 127.0.0.1:51981, session #0, session flags 0x00000080
3205 HTTP msg state 26, msg flags 0x00000000, tx flags 0x00000000
3206 HTTP chunk len 0 bytes, HTTP body len 0 bytes
3207 buffer flags 0x00808002, out 0 bytes, total 31 bytes
3208 pending 31 bytes, wrapping at 8040, error at position 13:
3209
3210 00000 GET /invalid request HTTP/1.1\r\n
3211
3212
3213The output of "show info" on the CLI provides a number of useful information
3214regarding the maximum connection rate ever reached, maximum SSL key rate ever
3215reached, and in general all information which can help to explain temporary
3216issues regarding CPU or memory usage. Example :
3217
3218 > show info
3219 Name: HAProxy
3220 Version: 1.6-dev7-e32d18-17
3221 Release_date: 2015/10/12
3222 Nbproc: 1
3223 Process_num: 1
3224 Pid: 7949
3225 Uptime: 0d 0h02m39s
3226 Uptime_sec: 159
3227 Memmax_MB: 0
3228 Ulimit-n: 120032
3229 Maxsock: 120032
3230 Maxconn: 60000
3231 Hard_maxconn: 60000
3232 CurrConns: 0
3233 CumConns: 3
3234 CumReq: 3
3235 MaxSslConns: 0
3236 CurrSslConns: 0
3237 CumSslConns: 0
3238 Maxpipes: 0
3239 PipesUsed: 0
3240 PipesFree: 0
3241 ConnRate: 0
3242 ConnRateLimit: 0
3243 MaxConnRate: 1
3244 SessRate: 0
3245 SessRateLimit: 0
3246 MaxSessRate: 1
3247 SslRate: 0
3248 SslRateLimit: 0
3249 MaxSslRate: 0
3250 SslFrontendKeyRate: 0
3251 SslFrontendMaxKeyRate: 0
3252 SslFrontendSessionReuse_pct: 0
3253 SslBackendKeyRate: 0
3254 SslBackendMaxKeyRate: 0
3255 SslCacheLookups: 0
3256 SslCacheMisses: 0
3257 CompressBpsIn: 0
3258 CompressBpsOut: 0
3259 CompressBpsRateLim: 0
3260 ZlibMemUsage: 0
3261 MaxZlibMemUsage: 0
3262 Tasks: 5
3263 Run_queue: 1
3264 Idle_pct: 100
3265 node: wtap
3266 description:
3267
3268When an issue seems to randomly appear on a new version of HAProxy (eg: every
3269second request is aborted, occasional crash, etc), it is worth trying to enable
Dan Lloyd8e48b872016-07-01 21:01:18 -04003270memory poisoning so that each call to malloc() is immediately followed by the
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003271filling of the memory area with a configurable byte. By default this byte is
32720x50 (ASCII for 'P'), but any other byte can be used, including zero (which
3273will have the same effect as a calloc() and which may make issues disappear).
Dan Lloyd8e48b872016-07-01 21:01:18 -04003274Memory poisoning is enabled on the command line using the "-dM" option. It
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003275slightly hurts performance and is not recommended for use in production. If
Dan Lloyd8e48b872016-07-01 21:01:18 -04003276an issue happens all the time with it or never happens when poisoning uses
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003277byte zero, it clearly means you've found a bug and you definitely need to
3278report it. Otherwise if there's no clear change, the problem it is not related.
3279
3280When debugging some latency issues, it is important to use both strace and
3281tcpdump on the local machine, and another tcpdump on the remote system. The
3282reason for this is that there are delays everywhere in the processing chain and
3283it is important to know which one is causing latency to know where to act. In
3284practice, the local tcpdump will indicate when the input data come in. Strace
3285will indicate when haproxy receives these data (using recv/recvfrom). Warning,
3286openssl uses read()/write() syscalls instead of recv()/send(). Strace will also
3287show when haproxy sends the data, and tcpdump will show when the system sends
3288these data to the interface. Then the external tcpdump will show when the data
3289sent are really received (since the local one only shows when the packets are
3290queued). The benefit of sniffing on the local system is that strace and tcpdump
3291will use the same reference clock. Strace should be used with "-tts200" to get
3292complete timestamps and report large enough chunks of data to read them.
3293Tcpdump should be used with "-nvvttSs0" to report full packets, real sequence
3294numbers and complete timestamps.
3295
3296In practice, received data are almost always immediately received by haproxy
3297(unless the machine has a saturated CPU or these data are invalid and not
3298delivered). If these data are received but not sent, it generally is because
3299the output buffer is saturated (ie: recipient doesn't consume the data fast
3300enough). This can be confirmed by seeing that the polling doesn't notify of
3301the ability to write on the output file descriptor for some time (it's often
3302easier to spot in the strace output when the data finally leave and then roll
3303back to see when the write event was notified). It generally matches an ACK
3304received from the recipient, and detected by tcpdump. Once the data are sent,
3305they may spend some time in the system doing nothing. Here again, the TCP
3306congestion window may be limited and not allow these data to leave, waiting for
3307an ACK to open the window. If the traffic is idle and the data take 40 ms or
3308200 ms to leave, it's a different issue (which is not an issue), it's the fact
3309that the Nagle algorithm prevents empty packets from leaving immediately, in
3310hope that they will be merged with subsequent data. HAProxy automatically
3311disables Nagle in pure TCP mode and in tunnels. However it definitely remains
3312enabled when forwarding an HTTP body (and this contributes to the performance
3313improvement there by reducing the number of packets). Some HTTP non-compliant
3314applications may be sensitive to the latency when delivering incomplete HTTP
3315response messages. In this case you will have to enable "option http-no-delay"
3316to disable Nagle in order to work around their design, keeping in mind that any
3317other proxy in the chain may similarly be impacted. If tcpdump reports that data
3318leave immediately but the other end doesn't see them quickly, it can mean there
Dan Lloyd8e48b872016-07-01 21:01:18 -04003319is a congested WAN link, a congested LAN with flow control enabled and
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003320preventing the data from leaving, or more commonly that HAProxy is in fact
3321running in a virtual machine and that for whatever reason the hypervisor has
3322decided that the data didn't need to be sent immediately. In virtualized
3323environments, latency issues are almost always caused by the virtualization
3324layer, so in order to save time, it's worth first comparing tcpdump in the VM
3325and on the external components. Any difference has to be credited to the
3326hypervisor and its accompanying drivers.
3327
3328When some TCP SACK segments are seen in tcpdump traces (using -vv), it always
3329means that the side sending them has got the proof of a lost packet. While not
3330seeing them doesn't mean there are no losses, seeing them definitely means the
3331network is lossy. Losses are normal on a network, but at a rate where SACKs are
3332not noticeable at the naked eye. If they appear a lot in the traces, it is
3333worth investigating exactly what happens and where the packets are lost. HTTP
3334doesn't cope well with TCP losses, which introduce huge latencies.
3335
3336The "netstat -i" command will report statistics per interface. An interface
3337where the Rx-Ovr counter grows indicates that the system doesn't have enough
3338resources to receive all incoming packets and that they're lost before being
3339processed by the network driver. Rx-Drp indicates that some received packets
3340were lost in the network stack because the application doesn't process them
3341fast enough. This can happen during some attacks as well. Tx-Drp means that
3342the output queues were full and packets had to be dropped. When using TCP it
Dan Lloyd8e48b872016-07-01 21:01:18 -04003343should be very rare, but will possibly indicate a saturated outgoing link.
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003344
3345
334613. Security considerations
3347---------------------------
3348
3349HAProxy is designed to run with very limited privileges. The standard way to
3350use it is to isolate it into a chroot jail and to drop its privileges to a
3351non-root user without any permissions inside this jail so that if any future
3352vulnerability were to be discovered, its compromise would not affect the rest
3353of the system.
3354
Dan Lloyd8e48b872016-07-01 21:01:18 -04003355In order to perform a chroot, it first needs to be started as a root user. It is
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003356pointless to build hand-made chroots to start the process there, these ones are
3357painful to build, are never properly maintained and always contain way more
3358bugs than the main file-system. And in case of compromise, the intruder can use
3359the purposely built file-system. Unfortunately many administrators confuse
3360"start as root" and "run as root", resulting in the uid change to be done prior
3361to starting haproxy, and reducing the effective security restrictions.
3362
3363HAProxy will need to be started as root in order to :
3364 - adjust the file descriptor limits
3365 - bind to privileged port numbers
3366 - bind to a specific network interface
3367 - transparently listen to a foreign address
3368 - isolate itself inside the chroot jail
3369 - drop to another non-privileged UID
3370
3371HAProxy may require to be run as root in order to :
3372 - bind to an interface for outgoing connections
3373 - bind to privileged source ports for outgoing connections
Dan Lloyd8e48b872016-07-01 21:01:18 -04003374 - transparently bind to a foreign address for outgoing connections
Willy Tarreau2212e6a2015-10-13 14:40:55 +02003375
3376Most users will never need the "run as root" case. But the "start as root"
3377covers most usages.
3378
3379A safe configuration will have :
3380
3381 - a chroot statement pointing to an empty location without any access
3382 permissions. This can be prepared this way on the UNIX command line :
3383
3384 # mkdir /var/empty && chmod 0 /var/empty || echo "Failed"
3385
3386 and referenced like this in the HAProxy configuration's global section :
3387
3388 chroot /var/empty
3389
3390 - both a uid/user and gid/group statements in the global section :
3391
3392 user haproxy
3393 group haproxy
3394
3395 - a stats socket whose mode, uid and gid are set to match the user and/or
3396 group allowed to access the CLI so that nobody may access it :
3397
3398 stats socket /var/run/haproxy.stat uid hatop gid hatop mode 600
3399