asdasdas dsa

                         ----------------------

                                 HAProxy

                          Configuration Manual

                         ----------------------

                             version 1.4.22

                             willy tarreau

                               2012/08/14

 

 

This document covers the configuration language as implemented in the version

specified above. It does not provide any hint, example or advice. For such

documentation, please refer to the Reference Manual or the Architecture Manual.

The summary below is meant to help you search sections by name and navigate

through the document.

 

Note to documentation contributors :

    This document is formated with 80 columns per line, with even number of

    spaces for indentation and without tabs. Please follow these rules strictly

    so that it remains easily printable everywhere. If a line needs to be

    printed verbatim and does not fit, please end each line with a backslash

    ('\') and continue on next line. If you add sections, please update the

    summary below for easier searching.

 

 

Summary

-------

 

1.    Quick reminder about HTTP

1.1.      The HTTP transaction model

1.2.      HTTP request

1.2.1.        The Request line

1.2.2.        The request headers

1.3.      HTTP response

1.3.1.        The Response line

1.3.2.        The response headers

 

2.    Configuring HAProxy

2.1.      Configuration file format

2.2.      Time format

2.3.      Examples

 

3.    Global parameters

3.1.      Process management and security

3.2.      Performance tuning

3.3.      Debugging

3.4.      Userlists

 

4.    Proxies

4.1.      Proxy keywords matrix

4.2.      Alphabetically sorted keywords reference

 

5.    Server and default-server options

 

6.    HTTP header manipulation

 

7.    Using ACLs and pattern extraction

7.1.      Matching integers

7.2.      Matching strings

7.3.      Matching regular expressions (regexes)

7.4.      Matching IPv4 addresses

7.5.      Available matching criteria

7.5.1.        Matching at Layer 4 and below

7.5.2.        Matching contents at Layer 4

7.5.3.        Matching at Layer 7

7.6.      Pre-defined ACLs

7.7.      Using ACLs to form conditions

7.8.      Pattern extraction

 

8.    Logging

8.1.      Log levels

8.2.      Log formats

8.2.1.        Default log format

8.2.2.        TCP log format

8.2.3.        HTTP log format

8.3.      Advanced logging options

8.3.1.        Disabling logging of external tests

8.3.2.        Logging before waiting for the session to terminate

8.3.3.        Raising log level upon errors

8.3.4.        Disabling logging of successful connections

8.4.      Timing events

8.5.      Session state at disconnection

8.6.      Non-printable characters

8.7.      Capturing HTTP cookies

8.8.      Capturing HTTP headers

8.9.      Examples of logs

 

9.    Statistics and monitoring

9.1.      CSV format

9.2.      Unix Socket commands

 

 

1. Quick reminder about HTTP

----------------------------

 

When haproxy is running in HTTP mode, both the request and the response are

fully analyzed and indexed, thus it becomes possible to build matching criteria

on almost anything found in the contents.

 

However, it is important to understand how HTTP requests and responses are

formed, and how HAProxy decomposes them. It will then become easier to write

correct rules and to debug existing configurations.

 

 

1.1. The HTTP transaction model

-------------------------------

 

The HTTP protocol is transaction-driven. This means that each request will lead

to one and only one response. Traditionally, a TCP connection is established

from the client to the server, a request is sent by the client on the

connection, the server responds and the connection is closed. A new request

will involve a new connection :

 

  [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...

 

In this mode, called the "HTTP close" mode, there are as many connection

establishments as there are HTTP transactions. Since the connection is closed

by the server after the response, the client does not need to know the content

length.

 

Due to the transactional nature of the protocol, it was possible to improve it

to avoid closing a connection between two subsequent transactions. In this mode

however, it is mandatory that the server indicates the content length for each

response so that the client does not wait indefinitely. For this, a special

header is used: "Content-length". This mode is called the "keep-alive" mode :

 

  [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...

 

Its advantages are a reduced latency between transactions, and less processing

power required on the server side. It is generally better than the close mode,

but not always because the clients often limit their concurrent connections to

a smaller value.

 

A last improvement in the communications is the pipelining mode. It still uses

keep-alive, but the client does not wait for the first response to send the

second request. This is useful for fetching large number of images composing a

page :

 

  [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...

 

This can obviously have a tremendous benefit on performance because the network

latency is eliminated between subsequent requests. Many HTTP agents do not

correctly support pipelining since there is no way to associate a response with

the corresponding request in HTTP. For this reason, it is mandatory for the

server to reply in the exact same order as the requests were received.

 

By default HAProxy operates in a tunnel-like mode with regards to persistent

connections: for each connection it processes the first request and forwards

everything else (including additional requests) to selected server. Once

established, the connection is persisted both on the client and server

sides. Use "option http-server-close" to preserve client persistent connections

while handling every incoming request individually, dispatching them one after

another to servers, in HTTP close mode. Use "option httpclose" to switch both

sides to HTTP close mode. "option forceclose" and "option

http-pretend-keepalive" help working around servers misbehaving in HTTP close

mode.

 

 

1.2. HTTP request

-----------------

 

First, let's consider this HTTP request :

 

  Line     Contents

  number

     1     GET /serv/login.php?lang=en&profile=2 HTTP/1.1

     2     Host: www.mydomain.com

     3     User-agent: my small browser

     4     Accept: image/jpeg, image/gif

     5     Accept: image/png

 

 

1.2.1. The Request line

-----------------------

 

Line 1 is the "request line". It is always composed of 3 fields :

 

  - a METHOD      : GET

  - a URI         : /serv/login.php?lang=en&profile=2

  - a version tag : HTTP/1.1

 

All of them are delimited by what the standard calls LWS (linear white spaces),

which are commonly spaces, but can also be tabs or line feeds/carriage returns

followed by spaces/tabs. The method itself cannot contain any colon (':') and

is limited to alphabetic letters. All those various combinations make it

desirable that HAProxy performs the splitting itself rather than leaving it to

the user to write a complex or inaccurate regular expression.

 

The URI itself can have several forms :

 

  - A "relative URI" :

 

      /serv/login.php?lang=en&profile=2

 

    It is a complete URL without the host part. This is generally what is

    received by servers, reverse proxies and transparent proxies.

 

  - An "absolute URI", also called a "URL" :

 

      http://192.168.0.12:8080/serv/login.php?lang=en&profile=2

 

    It is composed of a "scheme" (the protocol name followed by '://'), a host

    name or address, optionally a colon (':') followed by a port number, then

    a relative URI beginning at the first slash ('/') after the address part.

    This is generally what proxies receive, but a server supporting HTTP/1.1

    must accept this form too.

 

  - a star ('*') : this form is only accepted in association with the OPTIONS

    method and is not relayable. It is used to inquiry a next hop's

    capabilities.

 

  - an address:port combination : 192.168.0.12:80

    This is used with the CONNECT method, which is used to establish TCP

    tunnels through HTTP proxies, generally for HTTPS, but sometimes for

    other protocols too.

 

In a relative URI, two sub-parts are identified. The part before the question

mark is called the "path". It is typically the relative path to static objects

on the server. The part after the question mark is called the "query string".

It is mostly used with GET requests sent to dynamic scripts and is very

specific to the language, framework or application in use.

 

 

1.2.2. The request headers

--------------------------

 

The headers start at the second line. They are composed of a name at the

beginning of the line, immediately followed by a colon (':'). Traditionally,

an LWS is added after the colon but that's not required. Then come the values.

Multiple identical headers may be folded into one single line, delimiting the

values with commas, provided that their order is respected. This is commonly

encountered in the "Cookie:" field. A header may span over multiple lines if

the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5

define a total of 3 values for the "Accept:" header.

 

Contrary to a common mis-conception, header names are not case-sensitive, and

their values are not either if they refer to other header names (such as the

"Connection:" header).

 

The end of the headers is indicated by the first empty line. People often say

that it's a double line feed, which is not exact, even if a double line feed

is one valid form of empty line.

 

Fortunately, HAProxy takes care of all these complex combinations when indexing

headers, checking values and counting them, so there is no reason to worry

about the way they could be written, but it is important not to accuse an

application of being buggy if it does unusual, valid things.

 

Important note:

   As suggested by RFC2616, HAProxy normalizes headers by replacing line breaks

   in the middle of headers by LWS in order to join multi-line headers. This

   is necessary for proper analysis and helps less capable HTTP parsers to work

   correctly and not to be fooled by such complex constructs.

 

 

1.3. HTTP response

------------------

 

An HTTP response looks very much like an HTTP request. Both are called HTTP

messages. Let's consider this HTTP response :

 

  Line     Contents

  number

     1     HTTP/1.1 200 OK

     2     Content-length: 350

     3     Content-Type: text/html

 

As a special case, HTTP supports so called "Informational responses" as status

codes 1xx. These messages are special in that they don't convey any part of the

response, they're just used as sort of a signaling message to ask a client to

continue to post its request for instance. In the case of a status 100 response

the requested information will be carried by the next non-100 response message

following the informational one. This implies that multiple responses may be

sent to a single request, and that this only works when keep-alive is enabled

(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to

correctly forward and skip them, and only process the next non-100 response. As

such, these messages are neither logged nor transformed, unless explicitly

state otherwise. Status 101 messages indicate that the protocol is changing

over the same connection and that haproxy must switch to tunnel mode, just as

if a CONNECT had occurred. Then the Upgrade header would contain additional

information about the type of protocol the connection is switching to.

 

 

1.3.1. The Response line

------------------------

 

Line 1 is the "response line". It is always composed of 3 fields :

 

  - a version tag : HTTP/1.1

  - a status code : 200

  - a reason      : OK

 

The status code is always 3-digit. The first digit indicates a general status :

 - 1xx = informational message to be skipped (eg: 100, 101)

 - 2xx = OK, content is following   (eg: 200, 206)

 - 3xx = OK, no content following   (eg: 302, 304)

 - 4xx = error caused by the client (eg: 401, 403, 404)

 - 5xx = error caused by the server (eg: 500, 502, 503)

 

Please refer to RFC2616 for the detailed meaning of all such codes. The

"reason" field is just a hint, but is not parsed by clients. Anything can be

found there, but it's a common practice to respect the well-established

messages. It can be composed of one or multiple words, such as "OK", "Found",

or "Authentication Required".

 

Haproxy may emit the following status codes by itself :

 

  Code  When / reason

   200  access to stats page, and when replying to monitoring requests

   301  when performing a redirection, depending on the configured code

   302  when performing a redirection, depending on the configured code

   303  when performing a redirection, depending on the configured code

   400  for an invalid or too large request

   401  when an authentication is required to perform the action (when

        accessing the stats page)

   403  when a request is forbidden by a "block" ACL or "reqdeny" filter

   408  when the request timeout strikes before the request is complete

   500  when haproxy encounters an unrecoverable internal error, such as a

        memory allocation failure, which should never happen

   502  when the server returns an empty, invalid or incomplete response, or

        when an "rspdeny" filter blocks the response.

   503  when no server was available to handle the request, or in response to

        monitoring requests which match the "monitor fail" condition

   504  when the response timeout strikes before the server responds

 

The error 4xx and 5xx codes above may be customized (see "errorloc" in section

4.2).

 

 

1.3.2. The response headers

---------------------------

 

Response headers work exactly like request headers, and as such, HAProxy uses

the same parsing function for both. Please refer to paragraph 1.2.2 for more

details.

 

 

2. Configuring HAProxy

----------------------

 

2.1. Configuration file format

------------------------------

 

HAProxy's configuration process involves 3 major sources of parameters :

 

  - the arguments from the command-line, which always take precedence

  - the "global" section, which sets process-wide parameters

  - the proxies sections which can take form of "defaults", "listen",

    "frontend" and "backend".

 

The configuration file syntax consists in lines beginning with a keyword

referenced in this manual, optionally followed by one or several parameters

delimited by spaces. If spaces have to be entered in strings, then they must be

preceded by a backslash ('\') to be escaped. Backslashes also have to be

escaped by doubling them.

 

 

2.2. Time format

----------------

 

Some parameters involve values representing time, such as timeouts. These

values are generally expressed in milliseconds (unless explicitly stated

otherwise) but may be expressed in any other unit by suffixing the unit to the

numeric value. It is important to consider this because it will not be repeated

for every keyword. Supported units are :

 

  - us : microseconds. 1 microsecond = 1/1000000 second

  - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.

  - s  : seconds. 1s = 1000ms

  - m  : minutes. 1m = 60s = 60000ms

  - h  : hours.   1h = 60m = 3600s = 3600000ms

  - d  : days.    1d = 24h = 1440m = 86400s = 86400000ms

 

 

2.3. Examples

-------------

 

    # Simple configuration for an HTTP proxy listening on port 80 on all

    # interfaces and forwarding requests to a single backend "servers" with a

    # single server "server1" listening on 127.0.0.1:8000

    global

        daemon

        maxconn 256

 

    defaults

        mode http

        timeout connect 5000ms

        timeout client 50000ms

        timeout server 50000ms

 

    frontend http-in

        bind *:80

        default_backend servers

 

    backend servers

        server server1 127.0.0.1:8000 maxconn 32

 

 

    # The same configuration defined with a single listen block. Shorter but

    # less expressive, especially in HTTP mode.

    global

        daemon

        maxconn 256

 

    defaults

        mode http

        timeout connect 5000ms

        timeout client 50000ms

        timeout server 50000ms

 

    listen http-in

        bind *:80

        server server1 127.0.0.1:8000 maxconn 32

 

 

Assuming haproxy is in $PATH, test these configurations in a shell with:

 

    $ sudo haproxy -f configuration.conf -c

 

 

3. Global parameters

--------------------

 

Parameters in the "global" section are process-wide and often OS-specific. They

are generally set once for all and do not need being changed once correct. Some

of them have command-line equivalents.

 

The following keywords are supported in the "global" section :

 

 * Process management and security

   - chroot

   - daemon

   - gid

   - group

   - log

   - log-send-hostname

   - nbproc

   - pidfile

   - uid

   - ulimit-n

   - user

   - stats

   - node

   - description

 

 * Performance tuning

   - maxconn

   - maxpipes

   - noepoll

   - nokqueue

   - nopoll

   - nosepoll

   - nosplice

   - spread-checks

   - tune.bufsize

   - tune.chksize

   - tune.maxaccept

   - tune.maxpollevents

   - tune.maxrewrite

   - tune.rcvbuf.client

   - tune.rcvbuf.server

   - tune.sndbuf.client

   - tune.sndbuf.server

 

 * Debugging

   - debug

   - quiet

 

 

3.1. Process management and security

------------------------------------

 

chroot <jail dir>

  Changes current directory to <jail dir> and performs a chroot() there before

  dropping privileges. This increases the security level in case an unknown

  vulnerability would be exploited, since it would make it very hard for the

  attacker to exploit the system. This only works when the process is started

  with superuser privileges. It is important to ensure that <jail_dir> is both

  empty and unwritable to anyone.

 

daemon

  Makes the process fork into background. This is the recommended mode of

  operation. It is equivalent to the command line "-D" argument. It can be

  disabled by the command line "-db" argument.

 

gid <number>

  Changes the process' group ID to <number>. It is recommended that the group

  ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must

  be started with a user belonging to this group, or with superuser privileges.

  See also "group" and "uid".

 

group <group name>

  Similar to "gid" but uses the GID of group name <group name> from /etc/group.

  See also "gid" and "user".

 

log <address> <facility> [max level [min level]]

  Adds a global syslog server. Up to two global servers can be defined. They

  will receive logs for startups and exits, as well as all logs from proxies

  configured with "log global".

 

  <address> can be one of:

 

        - An IPv4 address optionally followed by a colon and a UDP port. If

          no port is specified, 514 is used by default (the standard syslog

          port).

 

        - A filesystem path to a UNIX domain socket, keeping in mind

          considerations for chroot (be sure the path is accessible inside

          the chroot) and uid/gid (be sure the path is appropriately

          writeable).

 

  <facility> must be one of the 24 standard syslog facilities :

 

          kern   user   mail   daemon auth   syslog lpr    news

          uucp   cron   auth2  ftp    ntp    audit  alert  cron2

          local0 local1 local2 local3 local4 local5 local6 local7

 

  An optional level can be specified to filter outgoing messages. By default,

  all messages are sent. If a maximum level is specified, only messages with a

  severity at least as important as this level will be sent. An optional minimum

  level can be specified. If it is set, logs emitted with a more severe level

  than this one will be capped to this level. This is used to avoid sending

  "emerg" messages on all terminals on some default syslog configurations.

  Eight levels are known :

 

          emerg  alert  crit   err    warning notice info  debug

 

log-send-hostname [<string>]

  Sets the hostname field in the syslog header. If optional "string" parameter

  is set the header is set to the string contents, otherwise uses the hostname

  of the system. Generally used if one is not relaying logs through an

  intermediate syslog server or for simply customizing the hostname printed in

  the logs.

 

log-tag <string>

  Sets the tag field in the syslog header to this string. It defaults to the

  program name as launched from the command line, which usually is "haproxy".

  Sometimes it can be useful to differentiate between multiple processes

  running on the same host.

 

nbproc <number>

  Creates <number> processes when going daemon. This requires the "daemon"

  mode. By default, only one process is created, which is the recommended mode

  of operation. For systems limited to small sets of file descriptors per

  process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES

  IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".

 

pidfile <pidfile>

  Writes pids of all daemons into file <pidfile>. This option is equivalent to

  the "-p" command line argument. The file must be accessible to the user

  starting the process. See also "daemon".

 

stats socket <path> [{uid | user} <uid>] [{gid | group} <gid>] [mode <mode>]

             [level <level>]

 

  Creates a UNIX socket in stream mode at location <path>. Any previously

  existing socket will be backed up then replaced. Connections to this socket

  will return various statistics outputs and even allow some commands to be

  issued. Please consult section 9.2 "Unix Socket commands" for more details.

 

  An optional "level" parameter can be specified to restrict the nature of

  the commands that can be issued on the socket :

    - "user" is the least privileged level ; only non-sensitive stats can be

      read, and no change is allowed. It would make sense on systems where it

      is not easy to restrict access to the socket.

 

    - "operator" is the default level and fits most common uses. All data can

      be read, and only non-sensitive changes are permitted (eg: clear max

      counters).

 

    - "admin" should be used with care, as everything is permitted (eg: clear

      all counters).

 

  On platforms which support it, it is possible to restrict access to this

  socket by specifying numerical IDs after "uid" and "gid", or valid user and

  group names after the "user" and "group" keywords. It is also possible to

  restrict permissions on the socket by passing an octal value after the "mode"

  keyword (same syntax as chmod). Depending on the platform, the permissions on

  the socket will be inherited from the directory which hosts it, or from the

  user the process is started with.

 

stats timeout <timeout, in milliseconds>

  The default timeout on the stats socket is set to 10 seconds. It is possible

  to change this value with "stats timeout". The value must be passed in

  milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.

 

stats maxconn <connections>

  By default, the stats socket is limited to 10 concurrent connections. It is

  possible to change this value with "stats maxconn".

 

uid <number>

  Changes the process' user ID to <number>. It is recommended that the user ID

  is dedicated to HAProxy or to a small set of similar daemons. HAProxy must

  be started with superuser privileges in order to be able to switch to another

  one. See also "gid" and "user".

 

ulimit-n <number>

  Sets the maximum number of per-process file-descriptors to <number>. By

  default, it is automatically computed, so it is recommended not to use this

  option.

 

user <user name>

  Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.

  See also "uid" and "group".

 

node <name>

  Only letters, digits, hyphen and underscore are allowed, like in DNS names.

 

  This statement is useful in HA configurations where two or more processes or

  servers share the same IP address. By setting a different node-name on all

  nodes, it becomes easy to immediately spot what server is handling the

  traffic.

 

description <text>

  Add a text that describes the instance.

 

  Please note that it is required to escape certain characters (# for example)

  and this text is inserted into a html page so you should avoid using

  "<" and ">" characters.

 

 

3.2. Performance tuning

-----------------------

 

maxconn <number>

  Sets the maximum per-process number of concurrent connections to <number>. It

  is equivalent to the command-line argument "-n". Proxies will stop accepting

  connections when this limit is reached. The "ulimit-n" parameter is

  automatically adjusted according to this value. See also "ulimit-n".

 

maxpipes <number>

  Sets the maximum per-process number of pipes to <number>. Currently, pipes

  are only used by kernel-based tcp splicing. Since a pipe contains two file

  descriptors, the "ulimit-n" value will be increased accordingly. The default

  value is maxconn/4, which seems to be more than enough for most heavy usages.

  The splice code dynamically allocates and releases pipes, and can fall back

  to standard copy, so setting this value too low may only impact performance.

 

noepoll

  Disables the use of the "epoll" event polling system on Linux. It is

  equivalent to the command-line argument "-de". The next polling system

  used will generally be "poll". See also "nosepoll", and "nopoll".

 

nokqueue

  Disables the use of the "kqueue" event polling system on BSD. It is

  equivalent to the command-line argument "-dk". The next polling system

  used will generally be "poll". See also "nopoll".

 

nopoll

  Disables the use of the "poll" event polling system. It is equivalent to the

  command-line argument "-dp". The next polling system used will be "select".

  It should never be needed to disable "poll" since it's available on all

  platforms supported by HAProxy. See also "nosepoll", and "nopoll" and

  "nokqueue".

 

nosepoll

  Disables the use of the "speculative epoll" event polling system on Linux. It

  is equivalent to the command-line argument "-ds". The next polling system

  used will generally be "epoll". See also "nosepoll", and "nopoll".

 

nosplice

  Disables the use of kernel tcp splicing between sockets on Linux. It is

  equivalent to the command line argument "-dS".  Data will then be copied

  using conventional and more portable recv/send calls. Kernel tcp splicing is

  limited to some very recent instances of kernel 2.6. Most versions between

  2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not

  be used. This option makes it easier to globally disable kernel splicing in

  case of doubt. See also "option splice-auto", "option splice-request" and

  "option splice-response".

 

spread-checks <0..50, in percent>

  Sometimes it is desirable to avoid sending health checks to servers at exact

  intervals, for instance when many logical servers are located on the same

  physical server. With the help of this parameter, it becomes possible to add

  some randomness in the check interval between 0 and +/- 50%. A value between

  2 and 5 seems to show good results. The default value remains at 0.

 

tune.bufsize <number>

  Sets the buffer size to this size (in bytes). Lower values allow more

  sessions to coexist in the same amount of RAM, and higher values allow some

  applications with very large cookies to work. The default value is 16384 and

  can be changed at build time. It is strongly recommended not to change this

  from the default value, as very low values will break some services such as

  statistics, and values larger than default size will increase memory usage,

  possibly causing the system to run out of memory. At least the global maxconn

  parameter should be decreased by the same factor as this one is increased.

 

tune.chksize <number>

  Sets the check buffer size to this size (in bytes). Higher values may help

  find string or regex patterns in very large pages, though doing so may imply

  more memory and CPU usage. The default value is 16384 and can be changed at

  build time. It is not recommended to change this value, but to use better

  checks whenever possible.

 

tune.maxaccept <number>

  Sets the maximum number of consecutive accepts that a process may perform on

  a single wake up. High values give higher priority to high connection rates,

  while lower values give higher priority to already established connections.

  This value is limited to 100 by default in single process mode. However, in

  multi-process mode (nbproc > 1), it defaults to 8 so that when one process

  wakes up, it does not take all incoming connections for itself and leaves a

  part of them to other processes. Setting this value to -1 completely disables

  the limitation. It should normally not be needed to tweak this value.

 

tune.maxpollevents <number>

  Sets the maximum amount of events that can be processed at once in a call to

  the polling system. The default value is adapted to the operating system. It

  has been noticed that reducing it below 200 tends to slightly decrease

  latency at the expense of network bandwidth, and increasing it above 200

  tends to trade latency for slightly increased bandwidth.

 

tune.maxrewrite <number>

  Sets the reserved buffer space to this size in bytes. The reserved space is

  used for header rewriting or appending. The first reads on sockets will never

  fill more than bufsize-maxrewrite. Historically it has defaulted to half of

  bufsize, though that does not make much sense since there are rarely large

  numbers of headers to add. Setting it too high prevents processing of large

  requests or responses. Setting it too low prevents addition of new headers

  to already large requests or to POST requests. It is generally wise to set it

  to about 1024. It is automatically readjusted to half of bufsize if it is

  larger than that. This means you don't have to worry about it when changing

  bufsize.

 

tune.rcvbuf.client <number>

tune.rcvbuf.server <number>

  Forces the kernel socket receive buffer size on the client or the server side

  to the specified value in bytes. This value applies to all TCP/HTTP frontends

  and backends. It should normally never be set, and the default size (0) lets

  the kernel autotune this value depending on the amount of available memory.

  However it can sometimes help to set it to very low values (eg: 4096) in

  order to save kernel memory by preventing it from buffering too large amounts

  of received data. Lower values will significantly increase CPU usage though.

 

tune.sndbuf.client <number>

tune.sndbuf.server <number>

  Forces the kernel socket send buffer size on the client or the server side to

  the specified value in bytes. This value applies to all TCP/HTTP frontends

  and backends. It should normally never be set, and the default size (0) lets

  the kernel autotune this value depending on the amount of available memory.

  However it can sometimes help to set it to very low values (eg: 4096) in

  order to save kernel memory by preventing it from buffering too large amounts

  of received data. Lower values will significantly increase CPU usage though.

  Another use case is to prevent write timeouts with extremely slow clients due

  to the kernel waiting for a large part of the buffer to be read before

  notifying haproxy again.

 

 

3.3. Debugging

--------------

 

debug

  Enables debug mode which dumps to stdout all exchanges, and disables forking

  into background. It is the equivalent of the command-line argument "-d". It

  should never be used in a production configuration since it may prevent full

  system startup.

 

quiet

  Do not display any message during startup. It is equivalent to the command-

  line argument "-q".

 

3.4. Userlists

--------------

It is possible to control access to frontend/backend/listen sections or to

http stats by allowing only authenticated and authorized users. To do this,

it is required to create at least one userlist and to define users.

 

userlist <listname>

  Creates new userlist with name <listname>. Many independent userlists can be

  used to store authentication & authorization data for independent customers.

 

group <groupname> [users <user>,<user>,(...)]

  Adds group <groupname> to the current userlist. It is also possible to

  attach users to this group by using a comma separated list of names

  proceeded by "users" keyword.

 

user <username> [password|insecure-password <password>]

                [groups <group>,<group>,(...)]

  Adds user <username> to the current userlist. Both secure (encrypted) and

  insecure (unencrypted) passwords can be used. Encrypted passwords are

  evaluated using the crypt(3) function so depending of the system's

  capabilities, different algorithms are supported. For example modern Glibc

  based Linux system supports MD5, SHA-256, SHA-512 and of course classic,

  DES-based method of crypting passwords.

 

 

  Example:

        userlist L1

          group G1 users tiger,scott

          group G2 users xdb,scott

 

          user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91

          user scott insecure-password elgato

          user xdb insecure-password hello

 

        userlist L2

          group G1

          group G2

 

          user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1

          user scott insecure-password elgato groups G1,G2

          user xdb insecure-password hello groups G2

 

  Please note that both lists are functionally identical.

 

4. Proxies

----------

 

Proxy configuration can be located in a set of sections :

 - defaults <name>

 - frontend <name>

 - backend  <name>

 - listen   <name>

 

A "defaults" section sets default parameters for all other sections following

its declaration. Those default parameters are reset by the next "defaults"

section. See below for the list of parameters which can be set in a "defaults"

section. The name is optional but its use is encouraged for better readability.

 

A "frontend" section describes a set of listening sockets accepting client

connections.

 

A "backend" section describes a set of servers to which the proxy will connect

to forward incoming connections.

 

A "listen" section defines a complete proxy with its frontend and backend

parts combined in one section. It is generally useful for TCP-only traffic.

 

All proxy names must be formed from upper and lower case letters, digits,

'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are

case-sensitive, which means that "www" and "WWW" are two different proxies.

 

Historically, all proxy names could overlap, it just caused troubles in the

logs. Since the introduction of content switching, it is mandatory that two

proxies with overlapping capabilities (frontend/backend) have different names.

However, it is still permitted that a frontend and a backend share the same

name, as this configuration seems to be commonly encountered.

 

Right now, two major proxy modes are supported : "tcp", also known as layer 4,

and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards

bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the

protocol, and can interact with it by allowing, blocking, switching, adding,

modifying, or removing arbitrary contents in requests or responses, based on

arbitrary criteria.

 

 

4.1. Proxy keywords matrix

--------------------------

 

The following list of keywords is supported. Most of them may only be used in a

limited set of section types. Some of them are marked as "deprecated" because

they are inherited from an old syntax which may be confusing or functionally

limited, and there are new recommended keywords to replace them. Keywords

marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no

option contstats". This makes sense when the option has been enabled by default

and must be disabled for a specific instance. Such options may also be prefixed

with "default" in order to restore default settings regardless of what has been

specified in a previous "defaults" section.

 

 

 keyword                              defaults   frontend   listen    backend

------------------------------------+----------+----------+---------+---------

acl                                       -          X         X         X

appsession                                -          -         X         X

backlog                                   X          X         X         -

balance                                   X          -         X         X

bind                                      -          X         X         -

bind-process                              X          X         X         X

block                                     -          X         X         X

capture cookie                            -          X         X         -

capture request header                    -          X         X         -

capture response header                   -          X         X         -

clitimeout                  (deprecated)  X          X         X         -

contimeout                  (deprecated)  X          -         X         X

cookie                                    X          -         X         X

default-server                            X          -         X         X

default_backend                           X          X         X         -

description                               -          X         X         X

disabled                                  X          X         X         X

dispatch                                  -          -         X         X

enabled                                   X          X         X         X

errorfile                                 X          X         X         X

errorloc                                  X          X         X         X

errorloc302                               X          X         X         X

-- keyword -------------------------- defaults - frontend - listen -- backend -

errorloc303                               X          X         X         X

force-persist                             -          X         X         X

fullconn                                  X          -         X         X

grace                                     X          X         X         X

hash-type                                 X          -         X         X

http-check disable-on-404                 X          -         X         X

http-check expect                         -          -         X         X

http-check send-state                     X          -         X         X

http-request                              -          X         X         X

id                                        -          X         X         X

ignore-persist                            -          X         X         X

log                                       X          X         X         X

maxconn                                   X          X         X         -

mode                                      X          X         X         X

monitor fail                              -          X         X         -

monitor-net                               X          X         X         -

monitor-uri                               X          X         X         -

option abortonclose                  (*)  X          -         X         X

option accept-invalid-http-request   (*)  X          X         X         -

option accept-invalid-http-response  (*)  X          -         X         X

option allbackups                    (*)  X          -         X         X

option checkcache                    (*)  X          -         X         X

option clitcpka                      (*)  X          X         X         -

option contstats                     (*)  X          X         X         -

option dontlog-normal                (*)  X          X         X         -

option dontlognull                   (*)  X          X         X         -

option forceclose                    (*)  X          X         X         X

-- keyword -------------------------- defaults - frontend - listen -- backend -

option forwardfor                         X          X         X         X

option http-no-delay                 (*)  X          X         X         X

option http-pretend-keepalive        (*)  X          X         X         X

option http-server-close             (*)  X          X         X         X

option http-use-proxy-header         (*)  X          X         X         -

option httpchk                            X          -         X         X

option httpclose                     (*)  X          X         X         X

option httplog                            X          X         X         X

option http_proxy                    (*)  X          X         X         X

option independant-streams           (*)  X          X         X         X

option ldap-check                         X          -         X         X

option log-health-checks             (*)  X          -         X         X

option log-separate-errors           (*)  X          X         X         -

option logasap                       (*)  X          X         X         -

option mysql-check                        X          -         X         X

option nolinger                      (*)  X          X         X         X

option originalto                         X          X         X         X

option persist                       (*)  X          -         X         X

option redispatch                    (*)  X          -         X         X

option smtpchk                            X          -         X         X

option socket-stats                  (*)  X          X         X         -

option splice-auto                   (*)  X          X         X         X

option splice-request                (*)  X          X         X         X

option splice-response               (*)  X          X         X         X

option srvtcpka                      (*)  X          -         X         X

option ssl-hello-chk                      X          -         X         X

-- keyword -------------------------- defaults - frontend - listen -- backend -

option tcp-smart-accept              (*)  X          X         X         -

option tcp-smart-connect             (*)  X          -         X         X

option tcpka                              X          X         X         X

option tcplog                             X          X         X         X

option transparent                   (*)  X          -         X         X

persist rdp-cookie                        X          -         X         X

rate-limit sessions                       X          X         X         -

redirect                                  -          X         X         X

redisp                      (deprecated)  X          -         X         X

redispatch                  (deprecated)  X          -         X         X

reqadd                                    -          X         X         X

reqallow                                  -          X         X         X

reqdel                                    -          X         X         X

reqdeny                                   -          X         X         X

reqiallow                                 -          X         X         X

reqidel                                   -          X         X         X

reqideny                                  -          X         X         X

reqipass                                  -          X         X         X

reqirep                                   -          X         X         X

reqisetbe                                 -          X         X         X

reqitarpit                                -          X         X         X

reqpass                                   -          X         X         X

reqrep                                    -          X         X         X

-- keyword -------------------------- defaults - frontend - listen -- backend -

reqsetbe                                  -          X         X         X

reqtarpit                                 -          X         X         X

retries                                   X          -         X         X

rspadd                                    -          X         X         X

rspdel                                    -          X         X         X

rspdeny                                   -          X         X         X

rspidel                                   -          X         X         X

rspideny                                  -          X         X         X

rspirep                                   -          X         X         X

rsprep                                    -          X         X         X

server                                    -          -         X         X

source                                    X          -         X         X

srvtimeout                  (deprecated)  X          -         X         X

stats admin                               -          -         X         X

stats auth                                X          -         X         X

stats enable                              X          -         X         X

stats hide-version                        X          -         X         X

stats http-request                        -          -         X         X

stats realm                               X          -         X         X

stats refresh                             X          -         X         X

stats scope                               X          -         X         X

stats show-desc                           X          -         X         X

stats show-legends                        X          -         X         X

stats show-node                           X          -         X         X

stats uri                                 X          -         X         X

-- keyword -------------------------- defaults - frontend - listen -- backend -

stick match                               -          -         X         X

stick on                                  -          -         X         X

stick store-request                       -          -         X         X

stick-table                               -          -         X         X

tcp-request content accept                -          X         X         -

tcp-request content reject                -          X         X         -

tcp-request inspect-delay                 -          X         X         -

timeout check                             X          -         X         X

timeout client                            X          X         X         -

timeout clitimeout          (deprecated)  X          X         X         -

timeout connect                           X          -         X         X

timeout contimeout          (deprecated)  X          -         X         X

timeout http-keep-alive                   X          X         X         X

timeout http-request                      X          X         X         X

timeout queue                             X          -         X         X

timeout server                            X          -         X         X

timeout srvtimeout          (deprecated)  X          -         X         X

timeout tarpit                            X          X         X         X

transparent                 (deprecated)  X          -         X         X

use_backend                               -          X         X         -

------------------------------------+----------+----------+---------+---------

 keyword                              defaults   frontend   listen    backend

 

 

4.2. Alphabetically sorted keywords reference

---------------------------------------------

 

This section provides a description of each keyword and its usage.

 

 

acl <aclname> <criterion> [flags] [operator] <value> ...

  Declare or complete an access list.

  May be used in sections :   defaults | frontend | listen | backend

                                 no    |    yes   |   yes  |   yes

  Example:

        acl invalid_src  src          0.0.0.0/7 224.0.0.0/3

        acl invalid_src  src_port     0:1023

        acl local_dst    hdr(host) -i localhost

 

  See section 7 about ACL usage.

 

 

appsession <cookie> len <length> timeout <holdtime>

           [request-learn] [prefix] [mode <path-parameters|query-string>]

  Define session stickiness on an existing application cookie.

  May be used in sections :   defaults | frontend | listen | backend

                                 no    |    no    |   yes  |   yes

  Arguments :

    <cookie>   this is the name of the cookie used by the application and which

               HAProxy will have to learn for each new session.

 

    <length>   this is the max number of characters that will be memorized and

               checked in each cookie value.

 

    <holdtime> this is the time after which the cookie will be removed from

               memory if unused. If no unit is specified, this time is in

               milliseconds.

 

    request-learn

               If this option is specified, then haproxy will be able to learn

               the cookie found in the request in case the server does not

               specify any in response. This is typically what happens with

               PHPSESSID cookies, or when haproxy's session expires before

               the application's session and the correct server is selected.

               It is recommended to specify this option to improve reliability.

 

    prefix     When this option is specified, haproxy will match on the cookie

               prefix (or URL parameter prefix). The appsession value is the

               data following this prefix.

 

               Example :

               appsession ASPSESSIONID len 64 timeout 3h prefix

 

               This will match the cookie ASPSESSIONIDXXXX=XXXXX,

               the appsession value will be XXXX=XXXXX.

 

    mode       This option allows to change the URL parser mode.

               2 modes are currently supported :

               - path-parameters :

                 The parser looks for the appsession in the path parameters

                 part (each parameter is separated by a semi-colon), which is

                 convenient for JSESSIONID for example.

                 This is the default mode if the option is not set.

               - query-string :

                 In this mode, the parser will look for the appsession in the

                 query string.

 

  When an application cookie is defined in a backend, HAProxy will check when

  the server sets such a cookie, and will store its value in a table, and

  associate it with the server's identifier. Up to <length> characters from

  the value will be retained. On each connection, haproxy will look for this

  cookie both in the "Cookie:" headers, and as a URL parameter (depending on

  the mode used). If a known value is found, the client will be directed to the

  server associated with this value. Otherwise, the load balancing algorithm is

  applied. Cookies are automatically removed from memory when they have been

  unused for a duration longer than <holdtime>.

 

  The definition of an application cookie is limited to one per backend.

 

  Note : Consider not using this feature in multi-process mode (nbproc > 1)

         unless you know what you do : memory is not shared between the

         processes, which can result in random behaviours.

 

  Example :

        appsession JSESSIONID len 52 timeout 3h

 

  See also : "cookie", "capture cookie", "balance", "stick", "stick-table",

             "ignore-persist", "nbproc" and "bind-process".

 

 

backlog <conns>

  Give hints to the system about the approximate listen backlog desired size

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments :

    <conns>   is the number of pending connections. Depending on the operating

              system, it may represent the number of already acknowledged

              connections, of non-acknowledged ones, or both.

 

  In order to protect against SYN flood attacks, one solution is to increase

  the system's SYN backlog size. Depending on the system, sometimes it is just

  tunable via a system parameter, sometimes it is not adjustable at all, and

  sometimes the system relies on hints given by the application at the time of

  the listen() syscall. By default, HAProxy passes the frontend's maxconn value

  to the listen() syscall. On systems which can make use of this value, it can

  sometimes be useful to be able to specify a different value, hence this

  backlog parameter.

 

  On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is

  used as a hint and the system accepts up to the smallest greater power of

  two, and never more than some limits (usually 32768).

 

  See also : "maxconn" and the target operating system's tuning guide.

 

 

balance <algorithm> [ <arguments> ]

balance url_param <param> [check_post [<max_wait>]]

  Define the load balancing algorithm to be used in a backend.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <algorithm> is the algorithm used to select a server when doing load

                balancing. This only applies when no persistence information

                is available, or when a connection is redispatched to another

                server. <algorithm> may be one of the following :

 

      roundrobin  Each server is used in turns, according to their weights.

                  This is the smoothest and fairest algorithm when the server's

                  processing time remains equally distributed. This algorithm

                  is dynamic, which means that server weights may be adjusted

                  on the fly for slow starts for instance. It is limited by

                  design to 4128 active servers per backend. Note that in some

                  large farms, when a server becomes up after having been down

                  for a very short time, it may sometimes take a few hundreds

                  requests for it to be re-integrated into the farm and start

                  receiving traffic. This is normal, though very rare. It is

                  indicated here in case you would have the chance to observe

                  it, so that you don't worry.

 

      static-rr   Each server is used in turns, according to their weights.

                  This algorithm is as similar to roundrobin except that it is

                  static, which means that changing a server's weight on the

                  fly will have no effect. On the other hand, it has no design

                  limitation on the number of servers, and when a server goes

                  up, it is always immediately reintroduced into the farm, once

                  the full map is recomputed. It also uses slightly less CPU to

                  run (around -1%).

 

      leastconn   The server with the lowest number of connections receives the

                  connection. Round-robin is performed within groups of servers

                  of the same load to ensure that all servers will be used. Use

                  of this algorithm is recommended where very long sessions are

                  expected, such as LDAP, SQL, TSE, etc... but is not very well

                  suited for protocols using short sessions such as HTTP. This

                  algorithm is dynamic, which means that server weights may be

                  adjusted on the fly for slow starts for instance.

 

      source      The source IP address is hashed and divided by the total

                  weight of the running servers to designate which server will

                  receive the request. This ensures that the same client IP

                  address will always reach the same server as long as no

                  server goes down or up. If the hash result changes due to the

                  number of running servers changing, many clients will be

                  directed to a different server. This algorithm is generally

                  used in TCP mode where no cookie may be inserted. It may also

                  be used on the Internet to provide a best-effort stickiness

                  to clients which refuse session cookies. This algorithm is

                  static by default, which means that changing a server's

                  weight on the fly will have no effect, but this can be

                  changed using "hash-type".

 

      uri         This algorithm hashes either the left part of the URI (before

                  the question mark) or the whole URI (if the "whole" parameter

                  is present) and divides the hash value by the total weight of

                  the running servers. The result designates which server will

                  receive the request. This ensures that the same URI will

                  always be directed to the same server as long as no server

                  goes up or down. This is used with proxy caches and

                  anti-virus proxies in order to maximize the cache hit rate.

                  Note that this algorithm may only be used in an HTTP backend.

                  This algorithm is static by default, which means that

                  changing a server's weight on the fly will have no effect,

                  but this can be changed using "hash-type".

 

                  This algorithm supports two optional parameters "len" and

                  "depth", both followed by a positive integer number. These

                  options may be helpful when it is needed to balance servers

                  based on the beginning of the URI only. The "len" parameter

                  indicates that the algorithm should only consider that many

                  characters at the beginning of the URI to compute the hash.

                  Note that having "len" set to 1 rarely makes sense since most

                  URIs start with a leading "/".

 

                  The "depth" parameter indicates the maximum directory depth

                  to be used to compute the hash. One level is counted for each

                  slash in the request. If both parameters are specified, the

                  evaluation stops when either is reached.

 

      url_param   The URL parameter specified in argument will be looked up in

                  the query string of each HTTP GET request.

 

                  If the modifier "check_post" is used, then an HTTP POST

                  request entity will be searched for the parameter argument,

                  when it is not found in a query string after a question mark

                  ('?') in the URL. Optionally, specify a number of octets to

                  wait for before attempting to search the message body. If the

                  entity can not be searched, then round robin is used for each

                  request. For instance, if your clients always send the LB

                  parameter in the first 128 bytes, then specify that. The

                  default is 48. The entity data will not be scanned until the

                  required number of octets have arrived at the gateway, this

                  is the minimum of: (default/max_wait, Content-Length or first

                  chunk length). If Content-Length is missing or zero, it does

                  not need to wait for more data than the client promised to

                  send. When Content-Length is present and larger than

                  <max_wait>, then waiting is limited to <max_wait> and it is

                  assumed that this will be enough data to search for the

                  presence of the parameter. In the unlikely event that

                  Transfer-Encoding: chunked is used, only the first chunk is

                  scanned. Parameter values separated by a chunk boundary, may

                  be randomly balanced if at all.

 

                  If the parameter is found followed by an equal sign ('=') and

                  a value, then the value is hashed and divided by the total

                  weight of the running servers. The result designates which

                  server will receive the request.

 

                  This is used to track user identifiers in requests and ensure

                  that a same user ID will always be sent to the same server as

                  long as no server goes up or down. If no value is found or if

                  the parameter is not found, then a round robin algorithm is

                  applied. Note that this algorithm may only be used in an HTTP

                  backend. This algorithm is static by default, which means

                  that changing a server's weight on the fly will have no

                  effect, but this can be changed using "hash-type".

 

      hdr(<name>) The HTTP header <name> will be looked up in each HTTP request.

                  Just as with the equivalent ACL 'hdr()' function, the header

                  name in parenthesis is not case sensitive. If the header is

                  absent or if it does not contain any value, the roundrobin

                  algorithm is applied instead.

 

                  An optional 'use_domain_only' parameter is available, for

                  reducing the hash algorithm to the main domain part with some

                  specific headers such as 'Host'. For instance, in the Host

                  value "haproxy.1wt.eu", only "1wt" will be considered.

 

                  This algorithm is static by default, which means that

                  changing a server's weight on the fly will have no effect,

                  but this can be changed using "hash-type".

 

      rdp-cookie

      rdp-cookie(name)

                  The RDP cookie <name> (or "mstshash" if omitted) will be

                  looked up and hashed for each incoming TCP request. Just as

                  with the equivalent ACL 'req_rdp_cookie()' function, the name

                  is not case-sensitive. This mechanism is useful as a degraded

                  persistence mode, as it makes it possible to always send the

                  same user (or the same session ID) to the same server. If the

                  cookie is not found, the normal roundrobin algorithm is

                  used instead.

 

                  Note that for this to work, the frontend must ensure that an

                  RDP cookie is already present in the request buffer. For this

                  you must use 'tcp-request content accept' rule combined with

                  a 'req_rdp_cookie_cnt' ACL.

 

                  This algorithm is static by default, which means that

                  changing a server's weight on the fly will have no effect,

                  but this can be changed using "hash-type".

 

    <arguments> is an optional list of arguments which may be needed by some

                algorithms. Right now, only "url_param" and "uri" support an

                optional argument.

 

                balance uri [len <len>] [depth <depth>]

                balance url_param <param> [check_post [<max_wait>]]

 

  The load balancing algorithm of a backend is set to roundrobin when no other

  algorithm, mode nor option have been set. The algorithm may only be set once

  for each backend.

 

  Examples :

        balance roundrobin

        balance url_param userid

        balance url_param session_id check_post 64

        balance hdr(User-Agent)

        balance hdr(host)

        balance hdr(Host) use_domain_only

 

  Note: the following caveats and limitations on using the "check_post"

  extension with "url_param" must be considered :

 

    - all POST requests are eligible for consideration, because there is no way

      to determine if the parameters will be found in the body or entity which

      may contain binary data. Therefore another method may be required to

      restrict consideration of POST requests that have no URL parameters in

      the body. (see acl reqideny http_end)

 

    - using a <max_wait> value larger than the request buffer size does not

      make sense and is useless. The buffer size is set at build time, and

      defaults to 16 kB.

 

    - Content-Encoding is not supported, the parameter search will probably

      fail; and load balancing will fall back to Round Robin.

 

    - Expect: 100-continue is not supported, load balancing will fall back to

      Round Robin.

 

    - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.

      If the entire parameter value is not present in the first chunk, the

      selection of server is undefined (actually, defined by how little

      actually appeared in the first chunk).

 

    - This feature does not support generation of a 100, 411 or 501 response.

 

    - In some cases, requesting "check_post" MAY attempt to scan the entire

      contents of a message body. Scanning normally terminates when linear

      white space or control characters are found, indicating the end of what

      might be a URL parameter list. This is probably not a concern with SGML

      type message bodies.

 

  See also : "dispatch", "cookie", "appsession", "transparent", "hash-type" and

             "http_proxy".

 

 

bind [<address>]:<port_range> [, ...]

bind [<address>]:<port_range> [, ...] interface <interface>

bind [<address>]:<port_range> [, ...] mss <maxseg>

bind [<address>]:<port_range> [, ...] transparent

bind [<address>]:<port_range> [, ...] id <id>

bind [<address>]:<port_range> [, ...] name <name>

bind [<address>]:<port_range> [, ...] defer-accept

  Define one or several listening addresses and/or ports in a frontend.

  May be used in sections :   defaults | frontend | listen | backend

                                  no   |    yes   |   yes  |   no

  Arguments :

    <address>     is optional and can be a host name, an IPv4 address, an IPv6

                  address, or '*'. It designates the address the frontend will

                  listen on. If unset, all IPv4 addresses of the system will be

                  listened on. The same will apply for '*' or the system's

                  special address "0.0.0.0".

 

    <port_range>  is either a unique TCP port, or a port range for which the

                  proxy will accept connections for the IP address specified

                  above. The port is mandatory. Note that in the case of an

                  IPv6 address, the port is always the number after the last

                  colon (':'). A range can either be :

                   - a numerical port (ex: '80')

                   - a dash-delimited ports range explicitly stating the lower

                     and upper bounds (ex: '2000-2100') which are included in

                     the range.

 

                  Particular care must be taken against port ranges, because

                  every <address:port> couple consumes one socket (= a file

                  descriptor), so it's easy to consume lots of descriptors

                  with a simple range, and to run out of sockets. Also, each

                  <address:port> couple must be used only once among all

                  instances running on a same system. Please note that binding

                  to ports lower than 1024 generally require particular

                  privileges to start the program, which are independant of

                  the 'uid' parameter.

 

    <interface>   is an optional physical interface name. This is currently

                  only supported on Linux. The interface must be a physical

                  interface, not an aliased interface. When specified, all

                  addresses on the same line will only be accepted if the

                  incoming packet physically come through the designated

                  interface. It is also possible to bind multiple frontends to

                  the same address if they are bound to different interfaces.

                  Note that binding to a physical interface requires root

                  privileges.

 

    <maxseg>      is an optional TCP Maximum Segment Size (MSS) value to be

                  advertised on incoming connections. This can be used to force

                  a lower MSS for certain specific ports, for instance for

                  connections passing through a VPN. Note that this relies on a

                  kernel feature which is theorically supported under Linux but

                  was buggy in all versions prior to 2.6.28. It may or may not

                  work on other operating systems. The commonly advertised

                  value on Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP).

 

    <id>          is a persistent value for socket ID. Must be positive and

                  unique in the proxy. An unused value will automatically be

                  assigned if unset. Can only be used when defining only a

                  single socket.

 

    <name>        is an optional name provided for stats

 

    transparent   is an optional keyword which is supported only on certain

                  Linux kernels. It indicates that the addresses will be bound

                  even if they do not belong to the local machine. Any packet

                  targeting any of these addresses will be caught just as if

                  the address was locally configured. This normally requires

                  that IP forwarding is enabled. Caution! do not use this with

                  the default address '*', as it would redirect any traffic for

                  the specified port. This keyword is available only when

                  HAProxy is built with USE_LINUX_TPROXY=1.

 

    defer-accept  is an optional keyword which is supported only on certain

                  Linux kernels. It states that a connection will only be

                  accepted once some data arrive on it, or at worst after the

                  first retransmit. This should be used only on protocols for

                  which the client talks first (eg: HTTP). It can slightly

                  improve performance by ensuring that most of the request is

                  already available when the connection is accepted. On the

                  other hand, it will not be able to detect connections which

                  don't talk. It is important to note that this option is

                  broken in all kernels up to 2.6.31, as the connection is

                  never accepted until the client talks. This can cause issues

                  with front firewalls which would see an established

                  connection while the proxy will only see it in SYN_RECV.

 

  It is possible to specify a list of address:port combinations delimited by

  commas. The frontend will then listen on all of these addresses. There is no

  fixed limit to the number of addresses and ports which can be listened on in

  a frontend, as well as there is no limit to the number of "bind" statements

  in a frontend.

 

  Example :

        listen http_proxy

            bind :80,:443

            bind 10.0.0.1:10080,10.0.0.1:10443

 

  See also : "source".

 

 

bind-process [ all | odd | even | <number 1-32> ] ...

  Limit visibility of an instance to a certain set of processes numbers.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    all           All process will see this instance. This is the default. It

                  may be used to override a default value.

 

    odd           This instance will be enabled on processes 1,3,5,...31. This

                  option may be combined with other numbers.

 

    even          This instance will be enabled on processes 2,4,6,...32. This

                  option may be combined with other numbers. Do not use it

                  with less than 2 processes otherwise some instances might be

                  missing from all processes.

 

    number        The instance will be enabled on this process number, between

                  1 and 32. You must be careful not to reference a process

                  number greater than the configured global.nbproc, otherwise

                  some instances might be missing from all processes.

 

  This keyword limits binding of certain instances to certain processes. This

  is useful in order not to have too many processes listening to the same

  ports. For instance, on a dual-core machine, it might make sense to set

  'nbproc 2' in the global section, then distributes the listeners among 'odd'

  and 'even' instances.

 

  At the moment, it is not possible to reference more than 32 processes using

  this keyword, but this should be more than enough for most setups. Please

  note that 'all' really means all processes and is not limited to the first

  32.

 

  If some backends are referenced by frontends bound to other processes, the

  backend automatically inherits the frontend's processes.

 

  Example :

        listen app_ip1

            bind 10.0.0.1:80

            bind-process odd

 

        listen app_ip2

            bind 10.0.0.2:80

            bind-process even

 

        listen management

            bind 10.0.0.3:80

            bind-process 1 2 3 4

 

  See also : "nbproc" in global section.

 

 

block { if | unless } <condition>

  Block a layer 7 request if/unless a condition is matched

  May be used in sections :   defaults | frontend | listen | backend

                                 no    |    yes   |   yes  |   yes

 

  The HTTP request will be blocked very early in the layer 7 processing

  if/unless <condition> is matched. A 403 error will be returned if the request

  is blocked. The condition has to reference ACLs (see section 7). This is

  typically used to deny access to certain sensitive resources if some

  conditions are met or not met. There is no fixed limit to the number of

  "block" statements per instance.

 

  Example:

        acl invalid_src  src          0.0.0.0/7 224.0.0.0/3

        acl invalid_src  src_port     0:1023

        acl local_dst    hdr(host) -i localhost

        block if invalid_src || local_dst

 

  See section 7 about ACL usage.

 

 

capture cookie <name> len <length>

  Capture and log a cookie in the request and in the response.

  May be used in sections :   defaults | frontend | listen | backend

                                  no   |    yes   |   yes  |   no

  Arguments :

    <name>    is the beginning of the name of the cookie to capture. In order

              to match the exact name, simply suffix the name with an equal

              sign ('='). The full name will appear in the logs, which is

              useful with application servers which adjust both the cookie name

              and value (eg: ASPSESSIONXXXXX).

 

    <length>  is the maximum number of characters to report in the logs, which

              include the cookie name, the equal sign and the value, all in the

              standard "name=value" form. The string will be truncated on the

              right if it exceeds <length>.

 

  Only the first cookie is captured. Both the "cookie" request headers and the

  "set-cookie" response headers are monitored. This is particularly useful to

  check for application bugs causing session crossing or stealing between

  users, because generally the user's cookies can only change on a login page.

 

  When the cookie was not presented by the client, the associated log column

  will report "-". When a request does not cause a cookie to be assigned by the

  server, a "-" is reported in the response column.

 

  The capture is performed in the frontend only because it is necessary that

  the log format does not change for a given frontend depending on the

  backends. This may change in the future. Note that there can be only one

  "capture cookie" statement in a frontend. The maximum capture length is

  configured in the sources by default to 64 characters. It is not possible to

  specify a capture in a "defaults" section.

 

  Example:

        capture cookie ASPSESSION len 32

 

  See also : "capture request header", "capture response header" as well as

            section 8 about logging.

 

 

capture request header <name> len <length>

  Capture and log the first occurrence of the specified request header.

  May be used in sections :   defaults | frontend | listen | backend

                                  no   |    yes   |   yes  |   no

  Arguments :

    <name>    is the name of the header to capture. The header names are not

              case-sensitive, but it is a common practice to write them as they

              appear in the requests, with the first letter of each word in

              upper case. The header name will not appear in the logs, only the

              value is reported, but the position in the logs is respected.

 

    <length>  is the maximum number of characters to extract from the value and

              report in the logs. The string will be truncated on the right if

              it exceeds <length>.

 

  Only the first value of the last occurrence of the header is captured. The

  value will be added to the logs between braces ('{}'). If multiple headers

  are captured, they will be delimited by a vertical bar ('|') and will appear

  in the same order they were declared in the configuration. Non-existent

  headers will be logged just as an empty string. Common uses for request

  header captures include the "Host" field in virtual hosting environments, the

  "Content-length" when uploads are supported, "User-agent" to quickly

  differentiate between real users and robots, and "X-Forwarded-For" in proxied

  environments to find where the request came from.

 

  Note that when capturing headers such as "User-agent", some spaces may be

  logged, making the log analysis more difficult. Thus be careful about what

  you log if you know your log parser is not smart enough to rely on the

  braces.

 

  There is no limit to the number of captured request headers, but each capture

  is limited to 64 characters. In order to keep log format consistent for a

  same frontend, header captures can only be declared in a frontend. It is not

  possible to specify a capture in a "defaults" section.

 

  Example:

        capture request header Host len 15

        capture request header X-Forwarded-For len 15

        capture request header Referrer len 15

 

  See also : "capture cookie", "capture response header" as well as section 8

             about logging.

 

 

capture response header <name> len <length>

  Capture and log the first occurrence of the specified response header.

  May be used in sections :   defaults | frontend | listen | backend

                                  no   |    yes   |   yes  |   no

  Arguments :

    <name>    is the name of the header to capture. The header names are not

              case-sensitive, but it is a common practice to write them as they

              appear in the response, with the first letter of each word in

              upper case. The header name will not appear in the logs, only the

              value is reported, but the position in the logs is respected.

 

    <length>  is the maximum number of characters to extract from the value and

              report in the logs. The string will be truncated on the right if

              it exceeds <length>.

 

  Only the first value of the last occurrence of the header is captured. The

  result will be added to the logs between braces ('{}') after the captured

  request headers. If multiple headers are captured, they will be delimited by

  a vertical bar ('|') and will appear in the same order they were declared in

  the configuration. Non-existent headers will be logged just as an empty

  string. Common uses for response header captures include the "Content-length"

  header which indicates how many bytes are expected to be returned, the

  "Location" header to track redirections.

 

  There is no limit to the number of captured response headers, but each

  capture is limited to 64 characters. In order to keep log format consistent

  for a same frontend, header captures can only be declared in a frontend. It

  is not possible to specify a capture in a "defaults" section.

 

  Example:

        capture response header Content-length len 9

        capture response header Location len 15

 

  See also : "capture cookie", "capture request header" as well as section 8

             about logging.

 

 

clitimeout <timeout> (deprecated)

  Set the maximum inactivity time on the client side.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments :

    <timeout> is the timeout value is specified in milliseconds by default, but

              can be in any other unit if the number is suffixed by the unit,

              as explained at the top of this document.

 

  The inactivity timeout applies when the client is expected to acknowledge or

  send data. In HTTP mode, this timeout is particularly important to consider

  during the first phase, when the client sends the request, and during the

  response while it is reading data sent by the server. The value is specified

  in milliseconds by default, but can be in any other unit if the number is

  suffixed by the unit, as specified at the top of this document. In TCP mode

  (and to a lesser extent, in HTTP mode), it is highly recommended that the

  client timeout remains equal to the server timeout in order to avoid complex

  situations to debug. It is a good practice to cover one or several TCP packet

  losses by specifying timeouts that are slightly above multiples of 3 seconds

  (eg: 4 or 5 seconds).

 

  This parameter is specific to frontends, but can be specified once for all in

  "defaults" sections. This is in fact one of the easiest solutions not to

  forget about it. An unspecified timeout results in an infinite timeout, which

  is not recommended. Such a usage is accepted and works but reports a warning

  during startup because it may results in accumulation of expired sessions in

  the system if the system's timeouts are not configured either.

 

  This parameter is provided for compatibility but is currently deprecated.

  Please use "timeout client" instead.

 

  See also : "timeout client", "timeout http-request", "timeout server", and

             "srvtimeout".

 

 

contimeout <timeout> (deprecated)

  Set the maximum time to wait for a connection attempt to a server to succeed.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <timeout> is the timeout value is specified in milliseconds by default, but

              can be in any other unit if the number is suffixed by the unit,

              as explained at the top of this document.

 

  If the server is located on the same LAN as haproxy, the connection should be

  immediate (less than a few milliseconds). Anyway, it is a good practice to

  cover one or several TCP packet losses by specifying timeouts that are

  slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the

  connect timeout also presets the queue timeout to the same value if this one

  has not been specified. Historically, the contimeout was also used to set the

  tarpit timeout in a listen section, which is not possible in a pure frontend.

 

  This parameter is specific to backends, but can be specified once for all in

  "defaults" sections. This is in fact one of the easiest solutions not to

  forget about it. An unspecified timeout results in an infinite timeout, which

  is not recommended. Such a usage is accepted and works but reports a warning

  during startup because it may results in accumulation of failed sessions in

  the system if the system's timeouts are not configured either.

 

  This parameter is provided for backwards compatibility but is currently

  deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"

  instead.

 

  See also : "timeout connect", "timeout queue", "timeout tarpit",

             "timeout server", "contimeout".

 

 

cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]

              [ postonly ] [ preserve ] [ httponly ] [ secure ]

              [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]

  Enable cookie-based persistence in a backend.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <name>    is the name of the cookie which will be monitored, modified or

              inserted in order to bring persistence. This cookie is sent to

              the client via a "Set-Cookie" header in the response, and is

              brought back by the client in a "Cookie" header in all requests.

              Special care should be taken to choose a name which does not

              conflict with any likely application cookie. Also, if the same

              backends are subject to be used by the same clients (eg:

              HTTP/HTTPS), care should be taken to use different cookie names

              between all backends if persistence between them is not desired.

 

    rewrite   This keyword indicates that the cookie will be provided by the

              server and that haproxy will have to modify its value to set the

              server's identifier in it. This mode is handy when the management

              of complex combinations of "Set-cookie" and "Cache-control"

              headers is left to the application. The application can then

              decide whether or not it is appropriate to emit a persistence

              cookie. Since all responses should be monitored, this mode only

              works in HTTP close mode. Unless the application behaviour is

              very complex and/or broken, it is advised not to start with this

              mode for new deployments. This keyword is incompatible with

              "insert" and "prefix".

 

    insert    This keyword indicates that the persistence cookie will have to

              be inserted by haproxy in server responses if the client did not

 

              already have a cookie that would have permitted it to access this

              server. When used without the "preserve" option, if the server

              emits a cookie with the same name, it will be remove before

              processing.  For this reason, this mode can be used to upgrade

              existing configurations running in the "rewrite" mode. The cookie

              will only be a session cookie and will not be stored on the

              client's disk. By default, unless the "indirect" option is added,

              the server will see the cookies emitted by the client. Due to

              caching effects, it is generally wise to add the "nocache" or

              "postonly" keywords (see below). The "insert" keyword is not

              compatible with "rewrite" and "prefix".

 

    prefix    This keyword indicates that instead of relying on a dedicated

              cookie for the persistence, an existing one will be completed.

              This may be needed in some specific environments where the client

              does not support more than one single cookie and the application

              already needs it. In this case, whenever the server sets a cookie

              named <name>, it will be prefixed with the server's identifier

              and a delimiter. The prefix will be removed from all client

              requests so that the server still finds the cookie it emitted.

              Since all requests and responses are subject to being modified,

              this mode requires the HTTP close mode. The "prefix" keyword is

              not compatible with "rewrite" and "insert".

 

    indirect  When this option is specified, no cookie will be emitted to a

              client which already has a valid one for the server which has

              processed the request. If the server sets such a cookie itself,

              it will be removed, unless the "preserve" option is also set. In

              "insert" mode, this will additionally remove cookies from the

              requests transmitted to the server, making the persistence

              mechanism totally transparent from an application point of view.

 

    nocache   This option is recommended in conjunction with the insert mode

              when there is a cache between the client and HAProxy, as it

              ensures that a cacheable response will be tagged non-cacheable if

              a cookie needs to be inserted. This is important because if all

              persistence cookies are added on a cacheable home page for

              instance, then all customers will then fetch the page from an

              outer cache and will all share the same persistence cookie,

              leading to one server receiving much more traffic than others.

              See also the "insert" and "postonly" options.

 

    postonly  This option ensures that cookie insertion will only be performed

              on responses to POST requests. It is an alternative to the

              "nocache" option, because POST responses are not cacheable, so

              this ensures that the persistence cookie will never get cached.

              Since most sites do not need any sort of persistence before the

              first POST which generally is a login request, this is a very

              efficient method to optimize caching without risking to find a

              persistence cookie in the cache.

              See also the "insert" and "nocache" options.

 

    preserve  This option may only be used with "insert" and/or "indirect". It

              allows the server to emit the persistence cookie itself. In this

              case, if a cookie is found in the response, haproxy will leave it

              untouched. This is useful in order to end persistence after a

              logout request for instance. For this, the server just has to

              emit a cookie with an invalid value (eg: empty) or with a date in

              the past. By combining this mechanism with the "disable-on-404"

              check option, it is possible to perform a completely graceful

              shutdown because users will definitely leave the server after

              they logout.

 

    httponly  This option tells haproxy to add an "HttpOnly" cookie attribute

              when a cookie is inserted. This attribute is used so that a

              user agent doesn't share the cookie with non-HTTP components.

              Please check RFC6265 for more information on this attribute.

 

    secure    This option tells haproxy to add a "Secure" cookie attribute when

              a cookie is inserted. This attribute is used so that a user agent

              never emits this cookie over non-secure channels, which means

              that a cookie learned with this flag will be presented only over

              SSL/TLS connections. Please check RFC6265 for more information on

              this attribute.

 

    domain    This option allows to specify the domain at which a cookie is

              inserted. It requires exactly one parameter: a valid domain

              name. If the domain begins with a dot, the browser is allowed to

              use it for any host ending with that name. It is also possible to

              specify several domain names by invoking this option multiple

              times. Some browsers might have small limits on the number of

              domains, so be careful when doing that. For the record, sending

              10 domains to MSIE 6 or Firefox 2 works as expected.

 

    maxidle   This option allows inserted cookies to be ignored after some idle

              time. It only works with insert-mode cookies. When a cookie is

              sent to the client, the date this cookie was emitted is sent too.

              Upon further presentations of this cookie, if the date is older

              than the delay indicated by the parameter (in seconds), it will

              be ignored. Otherwise, it will be refreshed if needed when the

              response is sent to the client. This is particularly useful to

              prevent users who never close their browsers from remaining for

              too long on the same server (eg: after a farm size change). When

              this option is set and a cookie has no date, it is always

              accepted, but gets refreshed in the response. This maintains the

              ability for admins to access their sites. Cookies that have a

              date in the future further than 24 hours are ignored. Doing so

              lets admins fix timezone issues without risking kicking users off

              the site.

 

    maxlife   This option allows inserted cookies to be ignored after some life

              time, whether they're in use or not. It only works with insert

              mode cookies. When a cookie is first sent to the client, the date

              this cookie was emitted is sent too. Upon further presentations

              of this cookie, if the date is older than the delay indicated by

              the parameter (in seconds), it will be ignored. If the cookie in

              the request has no date, it is accepted and a date will be set.

              Cookies that have a date in the future further than 24 hours are

              ignored. Doing so lets admins fix timezone issues without risking

              kicking users off the site. Contrary to maxidle, this value is

              not refreshed, only the first visit date counts. Both maxidle and

              maxlife may be used at the time. This is particularly useful to

              prevent users who never close their browsers from remaining for

              too long on the same server (eg: after a farm size change). This

              is stronger than the maxidle method in that it forces a

              redispatch after some absolute delay.

 

  There can be only one persistence cookie per HTTP backend, and it can be

  declared in a defaults section. The value of the cookie will be the value

  indicated after the "cookie" keyword in a "server" statement. If no cookie

  is declared for a given server, the cookie is not set.

 

  Examples :

        cookie JSESSIONID prefix

        cookie SRV insert indirect nocache

        cookie SRV insert postonly indirect

        cookie SRV insert indirect nocache maxidle 30m maxlife 8h

 

  See also : "appsession", "balance source", "capture cookie", "server"

             and "ignore-persist".

 

 

default-server [param*]

  Change default options for a server in a backend

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments:

    <param*>  is a list of parameters for this server. The "default-server"

              keyword accepts an important number of options and has a complete

              section dedicated to it. Please refer to section 5 for more

              details.

 

  Example :

        default-server inter 1000 weight 13

 

  See also: "server" and section 5 about server options

 

 

default_backend <backend>

  Specify the backend to use when no "use_backend" rule has been matched.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments :

    <backend> is the name of the backend to use.

 

  When doing content-switching between frontend and backends using the

  "use_backend" keyword, it is often useful to indicate which backend will be

  used when no rule has matched. It generally is the dynamic backend which

  will catch all undetermined requests.

 

  Example :

 

        use_backend     dynamic  if  url_dyn

        use_backend     static   if  url_css url_img extension_img

        default_backend dynamic

 

  See also : "use_backend", "reqsetbe", "reqisetbe"

 

 

disabled

  Disable a proxy, frontend or backend.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

 

  The "disabled" keyword is used to disable an instance, mainly in order to

  liberate a listening port or to temporarily disable a service. The instance

  will still be created and its configuration will be checked, but it will be

  created in the "stopped" state and will appear as such in the statistics. It

  will not receive any traffic nor will it send any health-checks or logs. It

  is possible to disable many instances at once by adding the "disabled"

  keyword in a "defaults" section.

 

  See also : "enabled"

 

 

dispatch <address>:<port>

  Set a default server address

  May be used in sections :   defaults | frontend | listen | backend

                                 no    |    no    |   yes  |   yes

  Arguments : none

 

    <address> is the IPv4 address of the default server. Alternatively, a

              resolvable hostname is supported, but this name will be resolved

              during start-up.

 

    <ports>   is a mandatory port specification. All connections will be sent

              to this port, and it is not permitted to use port offsets as is

              possible with normal servers.

 

  The "dispatch" keyword designates a default server for use when no other

  server can take the connection. In the past it was used to forward non

  persistent connections to an auxiliary load balancer. Due to its simple

  syntax, it has also been used for simple TCP relays. It is recommended not to

  use it for more clarity, and to use the "server" directive instead.

 

  See also : "server"

 

 

enabled

  Enable a proxy, frontend or backend.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  The "enabled" keyword is used to explicitly enable an instance, when the

  defaults has been set to "disabled". This is very rarely used.

 

  See also : "disabled"

 

 

errorfile <code> <file>

  Return a file contents instead of errors generated by HAProxy

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    <code>    is the HTTP status code. Currently, HAProxy is capable of

              generating codes 200, 400, 403, 408, 500, 502, 503, and 504.

 

    <file>    designates a file containing the full HTTP response. It is

              recommended to follow the common practice of appending ".http" to

              the filename so that people do not confuse the response with HTML

              error pages, and to use absolute paths, since files are read

              before any chroot is performed.

 

  It is important to understand that this keyword is not meant to rewrite

  errors returned by the server, but errors detected and returned by HAProxy.

  This is why the list of supported errors is limited to a small set.

 

  Code 200 is emitted in response to requests matching a "monitor-uri" rule.

 

  The files are returned verbatim on the TCP socket. This allows any trick such

  as redirections to another URL or site, as well as tricks to clean cookies,

  force enable or disable caching, etc... The package provides default error

  files returning the same contents as default errors.

 

  The files should not exceed the configured buffer size (BUFSIZE), which

  generally is 8 or 16 kB, otherwise they will be truncated. It is also wise

  not to put any reference to local contents (eg: images) in order to avoid

  loops between the client and HAProxy when all servers are down, causing an

  error to be returned instead of an image. For better HTTP compliance, it is

  recommended that all header lines end with CR-LF and not LF alone.

 

  The files are read at the same time as the configuration and kept in memory.

  For this reason, the errors continue to be returned even when the process is

  chrooted, and no file change is considered while the process is running. A

  simple method for developing those files consists in associating them to the

  403 status code and interrogating a blocked URL.

 

  See also : "errorloc", "errorloc302", "errorloc303"

 

  Example :

        errorfile 400 /etc/haproxy/errorfiles/400badreq.http

        errorfile 403 /etc/haproxy/errorfiles/403forbid.http

        errorfile 503 /etc/haproxy/errorfiles/503sorry.http

 

 

errorloc <code> <url>

errorloc302 <code> <url>

  Return an HTTP redirection to a URL instead of errors generated by HAProxy

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    <code>    is the HTTP status code. Currently, HAProxy is capable of

              generating codes 200, 400, 403, 408, 500, 502, 503, and 504.

 

    <url>     it is the exact contents of the "Location" header. It may contain

              either a relative URI to an error page hosted on the same site,

              or an absolute URI designating an error page on another site.

              Special care should be given to relative URIs to avoid redirect

              loops if the URI itself may generate the same error (eg: 500).

 

  It is important to understand that this keyword is not meant to rewrite

  errors returned by the server, but errors detected and returned by HAProxy.

  This is why the list of supported errors is limited to a small set.

 

  Code 200 is emitted in response to requests matching a "monitor-uri" rule.

 

  Note that both keyword return the HTTP 302 status code, which tells the

  client to fetch the designated URL using the same HTTP method. This can be

  quite problematic in case of non-GET methods such as POST, because the URL

  sent to the client might not be allowed for something other than GET. To

  workaround this problem, please use "errorloc303" which send the HTTP 303

  status code, indicating to the client that the URL must be fetched with a GET

  request.

 

  See also : "errorfile", "errorloc303"

 

 

errorloc303 <code> <url>

  Return an HTTP redirection to a URL instead of errors generated by HAProxy

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    <code>    is the HTTP status code. Currently, HAProxy is capable of

              generating codes 400, 403, 408, 500, 502, 503, and 504.

 

    <url>     it is the exact contents of the "Location" header. It may contain

              either a relative URI to an error page hosted on the same site,

              or an absolute URI designating an error page on another site.

              Special care should be given to relative URIs to avoid redirect

              loops if the URI itself may generate the same error (eg: 500).

 

  It is important to understand that this keyword is not meant to rewrite

  errors returned by the server, but errors detected and returned by HAProxy.

  This is why the list of supported errors is limited to a small set.

 

  Code 200 is emitted in response to requests matching a "monitor-uri" rule.

 

  Note that both keyword return the HTTP 303 status code, which tells the

  client to fetch the designated URL using the same HTTP GET method. This

  solves the usual problems associated with "errorloc" and the 302 code. It is

  possible that some very old browsers designed before HTTP/1.1 do not support

  it, but no such problem has been reported till now.

 

  See also : "errorfile", "errorloc", "errorloc302"

 

 

force-persist { if | unless } <condition>

  Declare a condition to force persistence on down servers

  May be used in sections:    defaults | frontend | listen | backend

                                  no   |    yes   |   yes  |   yes

 

  By default, requests are not dispatched to down servers. It is possible to

  force this using "option persist", but it is unconditional and redispatches

  to a valid server if "option redispatch" is set. That leaves with very little

  possibilities to force some requests to reach a server which is artificially

  marked down for maintenance operations.

 

  The "force-persist" statement allows one to declare various ACL-based

  conditions which, when met, will cause a request to ignore the down status of

  a server and still try to connect to it. That makes it possible to start a

  server, still replying an error to the health checks, and run a specially

  configured browser to test the service. Among the handy methods, one could

  use a specific source IP address, or a specific cookie. The cookie also has

  the advantage that it can easily be added/removed on the browser from a test

  page. Once the service is validated, it is then possible to open the service

  to the world by returning a valid response to health checks.

 

  The forced persistence is enabled when an "if" condition is met, or unless an

  "unless" condition is met. The final redispatch is always disabled when this

  is used.

 

  See also : "option redispatch", "ignore-persist", "persist",

             and section 7 about ACL usage.

 

 

fullconn <conns>

  Specify at what backend load the servers will reach their maxconn

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <conns>   is the number of connections on the backend which will make the

              servers use the maximal number of connections.

 

  When a server has a "maxconn" parameter specified, it means that its number

  of concurrent connections will never go higher. Additionally, if it has a

  "minconn" parameter, it indicates a dynamic limit following the backend's

  load. The server will then always accept at least <minconn> connections,

  never more than <maxconn>, and the limit will be on the ramp between both

  values when the backend has less than <conns> concurrent connections. This

  makes it possible to limit the load on the servers during normal loads, but

  push it further for important loads without overloading the servers during

  exceptional loads.

 

  Example :

     # The servers will accept between 100 and 1000 concurrent connections each

     # and the maximum of 1000 will be reached when the backend reaches 10000

     # connections.

     backend dynamic

        fullconn   10000

        server     srv1   dyn1:80 minconn 100 maxconn 1000

        server     srv2   dyn2:80 minconn 100 maxconn 1000

 

  See also : "maxconn", "server"

 

 

grace <time>

  Maintain a proxy operational for some time after a soft stop

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    <time>    is the time (by default in milliseconds) for which the instance

              will remain operational with the frontend sockets still listening

              when a soft-stop is received via the SIGUSR1 signal.

 

  This may be used to ensure that the services disappear in a certain order.

  This was designed so that frontends which are dedicated to monitoring by an

  external equipment fail immediately while other ones remain up for the time

  needed by the equipment to detect the failure.

 

  Note that currently, there is very little benefit in using this parameter,

  and it may in fact complicate the soft-reconfiguration process more than

  simplify it.

 

 

hash-type <method>

  Specify a method to use for mapping hashes to servers

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    map-based   the hash table is a static array containing all alive servers.

                The hashes will be very smooth, will consider weights, but will

                be static in that weight changes while a server is up will be

                ignored. This means that there will be no slow start. Also,

                since a server is selected by its position in the array, most

                mappings are changed when the server count changes. This means

                that when a server goes up or down, or when a server is added

                to a farm, most connections will be redistributed to different

                servers. This can be inconvenient with caches for instance.

 

    consistent  the hash table is a tree filled with many occurrences of each

                server. The hash key is looked up in the tree and the closest

                server is chosen. This hash is dynamic, it supports changing

                weights while the servers are up, so it is compatible with the

                slow start feature. It has the advantage that when a server

                goes up or down, only its associations are moved. When a server

                is added to the farm, only a few part of the mappings are

                redistributed, making it an ideal algorithm for caches.

                However, due to its principle, the algorithm will never be very

                smooth and it may sometimes be necessary to adjust a server's

                weight or its ID to get a more balanced distribution. In order

                to get the same distribution on multiple load balancers, it is

                important that all servers have the same IDs.

 

  The default hash type is "map-based" and is recommended for most usages.

 

  See also : "balance", "server"

 

 

http-check disable-on-404

  Enable a maintenance mode upon HTTP/404 response to health-checks

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  When this option is set, a server which returns an HTTP code 404 will be

  excluded from further load-balancing, but will still receive persistent

  connections. This provides a very convenient method for Web administrators

  to perform a graceful shutdown of their servers. It is also important to note

  that a server which is detected as failed while it was in this mode will not

  generate an alert, just a notice. If the server responds 2xx or 3xx again, it

  will immediately be reinserted into the farm. The status on the stats page

  reports "NOLB" for a server in this mode. It is important to note that this

  option only works in conjunction with the "httpchk" option. If this option

  is used with "http-check expect", then it has precedence over it so that 404

  responses will still be considered as soft-stop.

 

  See also : "option httpchk", "http-check expect"

 

 

http-check expect [!] <match> <pattern>

  Make HTTP health checks consider reponse contents or specific status codes

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <match>   is a keyword indicating how to look for a specific pattern in the

              response. The keyword may be one of "status", "rstatus",

              "string", or "rstring". The keyword may be preceeded by an

              exclamation mark ("!") to negate the match. Spaces are allowed

              between the exclamation mark and the keyword. See below for more

              details on the supported keywords.

 

    <pattern> is the pattern to look for. It may be a string or a regular

              expression. If the pattern contains spaces, they must be escaped

              with the usual backslash ('\').

 

  By default, "option httpchk" considers that response statuses 2xx and 3xx

  are valid, and that others are invalid. When "http-check expect" is used,

  it defines what is considered valid or invalid. Only one "http-check"

  statement is supported in a backend. If a server fails to respond or times

  out, the check obviously fails. The available matches are :

 

    status <string> : test the exact string match for the HTTP status code.

                      A health check respose will be considered valid if the

                      response's status code is exactly this string. If the

                      "status" keyword is prefixed with "!", then the response

                      will be considered invalid if the status code matches.

 

    rstatus <regex> : test a regular expression for the HTTP status code.

                      A health check respose will be considered valid if the

                      response's status code matches the expression. If the

                      "rstatus" keyword is prefixed with "!", then the response

                      will be considered invalid if the status code matches.

                      This is mostly used to check for multiple codes.

 

    string <string> : test the exact string match in the HTTP response body.

                      A health check respose will be considered valid if the

                      response's body contains this exact string. If the

                      "string" keyword is prefixed with "!", then the response

                      will be considered invalid if the body contains this

                      string. This can be used to look for a mandatory word at

                      the end of a dynamic page, or to detect a failure when a

                      specific error appears on the check page (eg: a stack

                      trace).

 

    rstring <regex> : test a regular expression on the HTTP response body.

                      A health check respose will be considered valid if the

                      response's body matches this expression. If the "rstring"

                      keyword is prefixed with "!", then the response will be

                      considered invalid if the body matches the expression.

                      This can be used to look for a mandatory word at the end

                      of a dynamic page, or to detect a failure when a specific

                      error appears on the check page (eg: a stack trace).

 

  It is important to note that the responses will be limited to a certain size

  defined by the global "tune.chksize" option, which defaults to 16384 bytes.

  Thus, too large responses may not contain the mandatory pattern when using

  "string" or "rstring". If a large response is absolutely required, it is

  possible to change the default max size by setting the global variable.

  However, it is worth keeping in mind that parsing very large responses can

  waste some CPU cycles, especially when regular expressions are used, and that

  it is always better to focus the checks on smaller resources.

 

  Last, if "http-check expect" is combined with "http-check disable-on-404",

  then this last one has precedence when the server responds with 404.

 

  Examples :

         # only accept status 200 as valid

         http-check expect status 200

 

         # consider SQL errors as errors

         http-check expect ! string SQL\ Error

 

         # consider status 5xx only as errors

         http-check expect ! rstatus ^5

 

         # check that we have a correct hexadecimal tag before /html

         http-check expect rstring <!--tag:[0-9a-f]*</html>

 

  See also : "option httpchk", "http-check disable-on-404"

 

 

http-check send-state

  Enable emission of a state header with HTTP health checks

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  When this option is set, haproxy will systematically send a special header

  "X-Haproxy-Server-State" with a list of parameters indicating to each server

  how they are seen by haproxy. This can be used for instance when a server is

  manipulated without access to haproxy and the operator needs to know whether

  haproxy still sees it up or not, or if the server is the last one in a farm.

 

  The header is composed of fields delimited by semi-colons, the first of which

  is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid

  checks on the total number before transition, just as appears in the stats

  interface. Next headers are in the form "<variable>=<value>", indicating in

  no specific order some values available in the stats interface :

    - a variable "name", containing the name of the backend followed by a slash

      ("/") then the name of the server. This can be used when a server is

      checked in multiple backends.

 

    - a variable "node" containing the name of the haproxy node, as set in the

      global "node" variable, otherwise the system's hostname if unspecified.

 

    - a variable "weight" indicating the weight of the server, a slash ("/")

      and the total weight of the farm (just counting usable servers). This

      helps to know if other servers are available to handle the load when this

      one fails.

 

    - a variable "scur" indicating the current number of concurrent connections

      on the server, followed by a slash ("/") then the total number of

      connections on all servers of the same backend.

 

    - a variable "qcur" indicating the current number of requests in the

      server's queue.

 

  Example of a header received by the application server :

    >>>  X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \

           scur=13/22; qcur=0

 

  See also : "option httpchk", "http-check disable-on-404"

 

http-request { allow | deny | auth [realm <realm>] }

             [ { if | unless } <condition> ]

  Access control for Layer 7 requests

 

  May be used in sections:   defaults | frontend | listen | backend

                                no    |    yes   |   yes  |   yes

 

  These set of options allow to fine control access to a

  frontend/listen/backend. Each option may be followed by if/unless and acl.

  First option with matched condition (or option without condition) is final.

  For "deny" a 403 error will be returned, for "allow" normal processing is

  performed, for "auth" a 401/407 error code is returned so the client

  should be asked to enter a username and password.

 

  There is no fixed limit to the number of http-request statements per

  instance.

 

  Example:

        acl nagios src 192.168.129.3

        acl local_net src 192.168.0.0/16

        acl auth_ok http_auth(L1)

 

        http-request allow if nagios

        http-request allow if local_net auth_ok

        http-request auth realm Gimme if local_net auth_ok

        http-request deny

 

  Example:

        acl auth_ok http_auth_group(L1) G1

 

        http-request auth unless auth_ok

 

  See also : "stats http-request", section 3.4 about userlists and section 7

             about ACL usage.

 

http-send-name-header [<header>]

  Add the server name to a request. Use the header string given by <header>

 

  May be used in sections:   defaults | frontend | listen | backend

                               yes    |    no    |   yes  |   yes

 

  Arguments :

 

    <header>  The header string to use to send the server name

 

  The "http-send-name-header" statement causes the name of the target

  server to be added to the headers of an HTTP request.  The name

  is added with the header string proved.

 

  See also : "server"

 

id <value>

  Set a persistent ID to a proxy.

  May be used in sections :   defaults | frontend | listen | backend

                                  no   |    yes   |   yes  |   yes

  Arguments : none

 

  Set a persistent ID for the proxy. This ID must be unique and positive.

  An unused ID will automatically be assigned if unset. The first assigned

  value will be 1. This ID is currently only returned in statistics.

 

 

ignore-persist { if | unless } <condition>

  Declare a condition to ignore persistence

  May be used in sections:    defaults | frontend | listen | backend

                                  no   |    yes   |   yes  |   yes

 

  By default, when cookie persistence is enabled, every requests containing

  the cookie are unconditionally persistent (assuming the target server is up

  and running).

 

  The "ignore-persist" statement allows one to declare various ACL-based

  conditions which, when met, will cause a request to ignore persistence.

  This is sometimes useful to load balance requests for static files, which

  oftenly don't require persistence. This can also be used to fully disable

  persistence for a specific User-Agent (for example, some web crawler bots).

 

  Combined with "appsession", it can also help reduce HAProxy memory usage, as

  the appsession table won't grow if persistence is ignored.

 

  The persistence is ignored when an "if" condition is met, or unless an

  "unless" condition is met.

 

  See also : "force-persist", "cookie", and section 7 about ACL usage.

 

 

log global

log <address> <facility> [<level> [<minlevel>]]

  Enable per-instance logging of events and traffic.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    global     should be used when the instance's logging parameters are the

               same as the global ones. This is the most common usage. "global"

               replaces <address>, <facility> and <level> with those of the log

               entries found in the "global" section. Only one "log global"

               statement may be used per instance, and this form takes no other

               parameter.

 

    <address>  indicates where to send the logs. It takes the same format as

               for the "global" section's logs, and can be one of :

 

               - An IPv4 address optionally followed by a colon (':') and a UDP

                 port. If no port is specified, 514 is used by default (the

                 standard syslog port).

 

               - A filesystem path to a UNIX domain socket, keeping in mind

                 considerations for chroot (be sure the path is accessible

                 inside the chroot) and uid/gid (be sure the path is

                 appropriately writeable).

 

    <facility> must be one of the 24 standard syslog facilities :

 

                 kern   user   mail   daemon auth   syslog lpr    news

                 uucp   cron   auth2  ftp    ntp    audit  alert  cron2

                 local0 local1 local2 local3 local4 local5 local6 local7

 

    <level>    is optional and can be specified to filter outgoing messages. By

               default, all messages are sent. If a level is specified, only

               messages with a severity at least as important as this level

               will be sent. An optional minimum level can be specified. If it

               is set, logs emitted with a more severe level than this one will

               be capped to this level. This is used to avoid sending "emerg"

               messages on all terminals on some default syslog configurations.

               Eight levels are known :

 

                 emerg  alert  crit   err    warning notice info  debug

 

  Note that up to two "log" entries may be specified per instance. However, if

  "log global" is used and if the "global" section already contains 2 log

  entries, then additional log entries will be ignored.

 

  Also, it is important to keep in mind that it is the frontend which decides

  what to log from a connection, and that in case of content switching, the log

  entries from the backend will be ignored. Connections are logged at level

  "info".

 

  However, backend log declaration define how and where servers status changes

  will be logged. Level "notice" will be used to indicate a server going up,

  "warning" will be used for termination signals and definitive service

  termination, and "alert" will be used for when a server goes down.

 

  Note : According to RFC3164, messages are truncated to 1024 bytes before

         being emitted.

 

  Example :

    log global

    log 127.0.0.1:514 local0 notice         # only send important events

    log 127.0.0.1:514 local0 notice notice  # same but limit output level

 

 

maxconn <conns>

  Fix the maximum number of concurrent connections on a frontend

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments :

    <conns>   is the maximum number of concurrent connections the frontend will

              accept to serve. Excess connections will be queued by the system

              in the socket's listen queue and will be served once a connection

              closes.

 

  If the system supports it, it can be useful on big sites to raise this limit

  very high so that haproxy manages connection queues, instead of leaving the

  clients with unanswered connection attempts. This value should not exceed the

  global maxconn. Also, keep in mind that a connection contains two buffers

  of 8kB each, as well as some other data resulting in about 17 kB of RAM being

  consumed per established connection. That means that a medium system equipped

  with 1GB of RAM can withstand around 40000-50000 concurrent connections if

  properly tuned.

 

  Also, when <conns> is set to large values, it is possible that the servers

  are not sized to accept such loads, and for this reason it is generally wise

  to assign them some reasonable connection limits.

 

  By default, this value is set to 2000.

 

  See also : "server", global section's "maxconn", "fullconn"

 

 

mode { tcp|http|health }

  Set the running mode or protocol of the instance

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    tcp       The instance will work in pure TCP mode. A full-duplex connection

              will be established between clients and servers, and no layer 7

              examination will be performed. This is the default mode. It

              should be used for SSL, SSH, SMTP, ...

 

    http      The instance will work in HTTP mode. The client request will be

              analyzed in depth before connecting to any server. Any request

              which is not RFC-compliant will be rejected. Layer 7 filtering,

              processing and switching will be possible. This is the mode which

              brings HAProxy most of its value.

 

    health    The instance will work in "health" mode. It will just reply "OK"

              to incoming connections and close the connection. Nothing will be

              logged. This mode is used to reply to external components health

              checks. This mode is deprecated and should not be used anymore as

              it is possible to do the same and even better by combining TCP or

              HTTP modes with the "monitor" keyword.

 

  When doing content switching, it is mandatory that the frontend and the

  backend are in the same mode (generally HTTP), otherwise the configuration

  will be refused.

 

  Example :

     defaults http_instances

         mode http

 

  See also : "monitor", "monitor-net"

 

 

monitor fail { if | unless } <condition>

  Add a condition to report a failure to a monitor HTTP request.

  May be used in sections :   defaults | frontend | listen | backend

                                 no    |    yes   |   yes  |   no

  Arguments :

    if <cond>     the monitor request will fail if the condition is satisfied,

                  and will succeed otherwise. The condition should describe a

                  combined test which must induce a failure if all conditions

                  are met, for instance a low number of servers both in a

                  backend and its backup.

 

    unless <cond> the monitor request will succeed only if the condition is

                  satisfied, and will fail otherwise. Such a condition may be

                  based on a test on the presence of a minimum number of active

                  servers in a list of backends.

 

  This statement adds a condition which can force the response to a monitor

  request to report a failure. By default, when an external component queries

  the URI dedicated to monitoring, a 200 response is returned. When one of the

  conditions above is met, haproxy will return 503 instead of 200. This is

  very useful to report a site failure to an external component which may base

  routing advertisements between multiple sites on the availability reported by

  haproxy. In this case, one would rely on an ACL involving the "nbsrv"

  criterion. Note that "monitor fail" only works in HTTP mode. Both status

  messages may be tweaked using "errorfile" or "errorloc" if needed.

 

  Example:

     frontend www

        mode http

        acl site_dead nbsrv(dynamic) lt 2

        acl site_dead nbsrv(static)  lt 2

        monitor-uri   /site_alive

        monitor fail  if site_dead

 

  See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"

 

 

monitor-net <source>

  Declare a source network which is limited to monitor requests

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments :

    <source>  is the source IPv4 address or network which will only be able to

              get monitor responses to any request. It can be either an IPv4

              address, a host name, or an address followed by a slash ('/')

              followed by a mask.

 

  In TCP mode, any connection coming from a source matching <source> will cause

  the connection to be immediately closed without any log. This allows another

  equipment to probe the port and verify that it is still listening, without

  forwarding the connection to a remote server.

 

  In HTTP mode, a connection coming from a source matching <source> will be

  accepted, the following response will be sent without waiting for a request,

  then the connection will be closed : "HTTP/1.0 200 OK". This is normally

  enough for any front-end HTTP probe to detect that the service is UP and

  running without forwarding the request to a backend server.

 

  Monitor requests are processed very early. It is not possible to block nor

  divert them using ACLs. They cannot be logged either, and it is the intended

  purpose. They are only used to report HAProxy's health to an upper component,

  nothing more. Right now, it is not possible to set failure conditions on

  requests caught by "monitor-net".

 

  Last, please note that only one "monitor-net" statement can be specified in

  a frontend. If more than one is found, only the last one will be considered.

 

  Example :

    # addresses .252 and .253 are just probing us.

    frontend www

        monitor-net 192.168.0.252/31

 

  See also : "monitor fail", "monitor-uri"

 

 

monitor-uri <uri>

  Intercept a URI used by external components' monitor requests

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments :

    <uri>     is the exact URI which we want to intercept to return HAProxy's

              health status instead of forwarding the request.

 

  When an HTTP request referencing <uri> will be received on a frontend,

  HAProxy will not forward it nor log it, but instead will return either

  "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure

  conditions defined with "monitor fail". This is normally enough for any

  front-end HTTP probe to detect that the service is UP and running without

  forwarding the request to a backend server. Note that the HTTP method, the

  version and all headers are ignored, but the request must at least be valid

  at the HTTP level. This keyword may only be used with an HTTP-mode frontend.

 

  Monitor requests are processed very early. It is not possible to block nor

  divert them using ACLs. They cannot be logged either, and it is the intended

  purpose. They are only used to report HAProxy's health to an upper component,

  nothing more. However, it is possible to add any number of conditions using

  "monitor fail" and ACLs so that the result can be adjusted to whatever check

  can be imagined (most often the number of available servers in a backend).

 

  Example :

    # Use /haproxy_test to report haproxy's status

    frontend www

        mode http

        monitor-uri /haproxy_test

 

  See also : "monitor fail", "monitor-net"

 

 

option abortonclose

no option abortonclose

  Enable or disable early dropping of aborted requests pending in queues.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |     no   |   yes  |   yes

  Arguments : none

 

  In presence of very high loads, the servers will take some time to respond.

  The per-instance connection queue will inflate, and the response time will

  increase respective to the size of the queue times the average per-session

  response time. When clients will wait for more than a few seconds, they will

  often hit the "STOP" button on their browser, leaving a useless request in

  the queue, and slowing down other users, and the servers as well, because the

  request will eventually be served, then aborted at the first error

  encountered while delivering the response.

 

  As there is no way to distinguish between a full STOP and a simple output

  close on the client side, HTTP agents should be conservative and consider

  that the client might only have closed its output channel while waiting for

  the response. However, this introduces risks of congestion when lots of users

  do the same, and is completely useless nowadays because probably no client at

  all will close the session while waiting for the response. Some HTTP agents

  support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most

  hardware-based load balancers). So the probability for a closed input channel

  to represent a user hitting the "STOP" button is close to 100%, and the risk

  of being the single component to break rare but valid traffic is extremely

  low, which adds to the temptation to be able to abort a session early while

  still not served and not pollute the servers.

 

  In HAProxy, the user can choose the desired behaviour using the option

  "abortonclose". By default (without the option) the behaviour is HTTP

  compliant and aborted requests will be served. But when the option is

  specified, a session with an incoming channel closed will be aborted while

  it is still possible, either pending in the queue for a connection slot, or

  during the connection establishment if the server has not yet acknowledged

  the connection request. This considerably reduces the queue size and the load

  on saturated servers when users are tempted to click on STOP, which in turn

  reduces the response time for other users.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters

 

 

option accept-invalid-http-request

no option accept-invalid-http-request

  Enable or disable relaxing of HTTP request parsing

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  By default, HAProxy complies with RFC2616 in terms of message parsing. This

  means that invalid characters in header names are not permitted and cause an

  error to be returned to the client. This is the desired behaviour as such

  forbidden characters are essentially used to build attacks exploiting server

  weaknesses, and bypass security filtering. Sometimes, a buggy browser or

  server will emit invalid header names for whatever reason (configuration,

  implementation) and the issue will not be immediately fixed. In such a case,

  it is possible to relax HAProxy's header name parser to accept any character

  even if that does not make sense, by specifying this option.

 

  This option should never be enabled by default as it hides application bugs

  and open security breaches. It should only be deployed after a problem has

  been confirmed.

 

  When this option is enabled, erroneous header names will still be accepted in

  requests, but the complete request will be captured in order to permit later

  analysis using the "show errors" request on the UNIX stats socket. Doing this

  also helps confirming that the issue has been solved.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option accept-invalid-http-response" and "show errors" on the

             stats socket.

 

 

option accept-invalid-http-response

no option accept-invalid-http-response

  Enable or disable relaxing of HTTP response parsing

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |     no   |   yes  |   yes

  Arguments : none

 

  By default, HAProxy complies with RFC2616 in terms of message parsing. This

  means that invalid characters in header names are not permitted and cause an

  error to be returned to the client. This is the desired behaviour as such

  forbidden characters are essentially used to build attacks exploiting server

  weaknesses, and bypass security filtering. Sometimes, a buggy browser or

  server will emit invalid header names for whatever reason (configuration,

  implementation) and the issue will not be immediately fixed. In such a case,

  it is possible to relax HAProxy's header name parser to accept any character

  even if that does not make sense, by specifying this option.

 

  This option should never be enabled by default as it hides application bugs

  and open security breaches. It should only be deployed after a problem has

  been confirmed.

 

  When this option is enabled, erroneous header names will still be accepted in

  responses, but the complete response will be captured in order to permit

  later analysis using the "show errors" request on the UNIX stats socket.

  Doing this also helps confirming that the issue has been solved.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option accept-invalid-http-request" and "show errors" on the

             stats socket.

 

 

option allbackups

no option allbackups

  Use either all backup servers at a time or only the first one

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |     no   |   yes  |   yes

  Arguments : none

 

  By default, the first operational backup server gets all traffic when normal

  servers are all down. Sometimes, it may be preferred to use multiple backups

  at once, because one will not be enough. When "option allbackups" is enabled,

  the load balancing will be performed among all backup servers when all normal

  ones are unavailable. The same load balancing algorithm will be used and the

  servers' weights will be respected. Thus, there will not be any priority

  order between the backup servers anymore.

 

  This option is mostly used with static server farms dedicated to return a

  "sorry" page when an application is completely offline.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

 

option checkcache

no option checkcache

  Analyze all server responses and block requests with cacheable cookies

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |     no   |   yes  |   yes

  Arguments : none

 

  Some high-level frameworks set application cookies everywhere and do not

  always let enough control to the developer to manage how the responses should

  be cached. When a session cookie is returned on a cacheable object, there is a

  high risk of session crossing or stealing between users traversing the same

  caches. In some situations, it is better to block the response than to let

  some sensitive session information go in the wild.

 

  The option "checkcache" enables deep inspection of all server responses for

  strict compliance with HTTP specification in terms of cacheability. It

  carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server

  response to check if there's a risk of caching a cookie on a client-side

  proxy. When this option is enabled, the only responses which can be delivered

  to the client are :

    - all those without "Set-Cookie" header ;

    - all those with a return code other than 200, 203, 206, 300, 301, 410,

      provided that the server has not set a "Cache-control: public" header ;

    - all those that come from a POST request, provided that the server has not

      set a 'Cache-Control: public' header ;

    - those with a 'Pragma: no-cache' header

    - those with a 'Cache-control: private' header

    - those with a 'Cache-control: no-store' header

    - those with a 'Cache-control: max-age=0' header

    - those with a 'Cache-control: s-maxage=0' header

    - those with a 'Cache-control: no-cache' header

    - those with a 'Cache-control: no-cache="set-cookie"' header

    - those with a 'Cache-control: no-cache="set-cookie,' header

      (allowing other fields after set-cookie)

 

  If a response doesn't respect these requirements, then it will be blocked

  just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".

  The session state shows "PH--" meaning that the proxy blocked the response

  during headers processing. Additionally, an alert will be sent in the logs so

  that admins are informed that there's something to be fixed.

 

  Due to the high impact on the application, the application should be tested

  in depth with the option enabled before going to production. It is also a

  good practice to always activate it during tests, even if it is not used in

  production, as it will report potentially dangerous application behaviours.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

 

option clitcpka

no option clitcpka

  Enable or disable the sending of TCP keepalive packets on the client side

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  When there is a firewall or any session-aware component between a client and

  a server, and when the protocol involves very long sessions with long idle

  periods (eg: remote desktops), there is a risk that one of the intermediate

  components decides to expire a session which has remained idle for too long.

 

  Enabling socket-level TCP keep-alives makes the system regularly send packets

  to the other end of the connection, leaving it active. The delay between

  keep-alive probes is controlled by the system only and depends both on the

  operating system and its tuning parameters.

 

  It is important to understand that keep-alive packets are neither emitted nor

  received at the application level. It is only the network stacks which sees

  them. For this reason, even if one side of the proxy already uses keep-alives

  to maintain its connection alive, those keep-alive packets will not be

  forwarded to the other side of the proxy.

 

  Please note that this has nothing to do with HTTP keep-alive.

 

  Using option "clitcpka" enables the emission of TCP keep-alive probes on the

  client side of a connection, which should help when session expirations are

  noticed between HAProxy and a client.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option srvtcpka", "option tcpka"

 

 

option contstats

  Enable continuous traffic statistics updates

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  By default, counters used for statistics calculation are incremented

  only when a session finishes. It works quite well when serving small

  objects, but with big ones (for example large images or archives) or

  with A/V streaming, a graph generated from haproxy counters looks like

  a hedgehog. With this option enabled counters get incremented continuously,

  during a whole session. Recounting touches a hotpath directly so

  it is not enabled by default, as it has small performance impact (~0.5%).

 

 

option dontlog-normal

no option dontlog-normal

  Enable or disable logging of normal, successful connections

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  There are large sites dealing with several thousand connections per second

  and for which logging is a major pain. Some of them are even forced to turn

  logs off and cannot debug production issues. Setting this option ensures that

  normal connections, those which experience no error, no timeout, no retry nor

  redispatch, will not be logged. This leaves disk space for anomalies. In HTTP

  mode, the response status code is checked and return codes 5xx will still be

  logged.

 

  It is strongly discouraged to use this option as most of the time, the key to

  complex issues is in the normal logs which will not be logged here. If you

  need to separate logs, see the "log-separate-errors" option instead.

 

  See also : "log", "dontlognull", "log-separate-errors" and section 8 about

             logging.

 

 

option dontlognull

no option dontlognull

  Enable or disable logging of null connections

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  In certain environments, there are components which will regularly connect to

  various systems to ensure that they are still alive. It can be the case from

  another load balancer as well as from monitoring systems. By default, even a

  simple port probe or scan will produce a log. If those connections pollute

  the logs too much, it is possible to enable option "dontlognull" to indicate

  that a connection on which no data has been transferred will not be logged,

  which typically corresponds to those probes.

 

  It is generally recommended not to use this option in uncontrolled

  environments (eg: internet), otherwise scans and other malicious activities

  would not be logged.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "log", "monitor-net", "monitor-uri" and section 8 about logging.

 

 

option forceclose

no option forceclose

  Enable or disable active connection closing after response is transferred.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  Some HTTP servers do not necessarily close the connections when they receive

  the "Connection: close" set by "option httpclose", and if the client does not

  close either, then the connection remains open till the timeout expires. This

  causes high number of simultaneous connections on the servers and shows high

  global session times in the logs.

 

  When this happens, it is possible to use "option forceclose". It will

  actively close the outgoing server channel as soon as the server has finished

  to respond. This option implicitly enables the "httpclose" option. Note that

  this option also enables the parsing of the full request and response, which

  means we can close the connection to the server very quickly, releasing some

  resources earlier than with httpclose.

 

  This option may also be combined with "option http-pretend-keepalive", which

  will disable sending of the "Connection: close" header, but will still cause

  the connection to be closed once the whole response is received.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option httpclose" and "option http-pretend-keepalive"

 

 

option forwardfor [ except <network> ] [ header <name> ] [ if-none ]

  Enable insertion of the X-Forwarded-For header to requests sent to servers

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    <network> is an optional argument used to disable this option for sources

              matching <network>

    <name>    an optional argument to specify a different "X-Forwarded-For"

              header name.

 

  Since HAProxy works in reverse-proxy mode, the servers see its IP address as

  their client address. This is sometimes annoying when the client's IP address

  is expected in server logs. To solve this problem, the well-known HTTP header

  "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.

  This header contains a value representing the client's IP address. Since this

  header is always appended at the end of the existing header list, the server

  must be configured to always use the last occurrence of this header only. See

  the server's manual to find how to enable use of this standard header. Note

  that only the last occurrence of the header must be used, since it is really

  possible that the client has already brought one.

 

  The keyword "header" may be used to supply a different header name to replace

  the default "X-Forwarded-For". This can be useful where you might already

  have a "X-Forwarded-For" header from a different application (eg: stunnel),

  and you need preserve it. Also if your backend server doesn't use the

  "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers

  require "X-Cluster-Client-IP").

 

  Sometimes, a same HAProxy instance may be shared between a direct client

  access and a reverse-proxy access (for instance when an SSL reverse-proxy is

  used to decrypt HTTPS traffic). It is possible to disable the addition of the

  header for a known source address or network by adding the "except" keyword

  followed by the network address. In this case, any source IP matching the

  network will not cause an addition of this header. Most common uses are with

  private networks or 127.0.0.1.

 

  Alternatively, the keyword "if-none" states that the header will only be

  added if it is not present. This should only be used in perfectly trusted

  environment, as this might cause a security issue if headers reaching haproxy

  are under the control of the end-user.

 

  This option may be specified either in the frontend or in the backend. If at

  least one of them uses it, the header will be added. Note that the backend's

  setting of the header subargument takes precedence over the frontend's if

  both are defined. In the case of the "if-none" argument, if at least one of

  the frontend or the backend does not specify it, it wants the addition to be

  mandatory, so it wins.

 

  It is important to note that by default, HAProxy works in tunnel mode and

  only inspects the first request of a connection, meaning that only the first

  request will have the header appended, which is certainly not what you want.

  In order to fix this, ensure that any of the "httpclose", "forceclose" or

  "http-server-close" options is set when using this option.

 

  Examples :

    # Public HTTP address also used by stunnel on the same machine

    frontend www

        mode http

        option forwardfor except 127.0.0.1  # stunnel already adds the header

 

    # Those servers want the IP Address in X-Client

    backend www

        mode http

        option forwardfor header X-Client

 

  See also : "option httpclose", "option http-server-close",

             "option forceclose"

 

 

option http-no-delay

no option http-no-delay

  Instruct the system to favor low interactive delays over performance in HTTP

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  In HTTP, each payload is unidirectional and has no notion of interactivity.

  Any agent is expected to queue data somewhat for a reasonably low delay.

  There are some very rare server-to-server applications that abuse the HTTP

  protocol and expect the payload phase to be highly interactive, with many

  interleaved data chunks in both directions within a single request. This is

  absolutely not supported by the HTTP specification and will not work across

  most proxies or servers. When such applications attempt to do this through

  haproxy, it works but they will experience high delays due to the network

  optimizations which favor performance by instructing the system to wait for

  enough data to be available in order to only send full packets. Typical

  delays are around 200 ms per round trip. Note that this only happens with

  abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not

  affected.

 

  When "option http-no-delay" is present in either the frontend or the backend

  used by a connection, all such optimizations will be disabled in order to

  make the exchanges as fast as possible. Of course this offers no guarantee on

  the functionality, as it may break at any other place. But if it works via

  HAProxy, it will work as fast as possible. This option should never be used

  by default, and should never be used at all unless such a buggy application

  is discovered. The impact of using this option is an increase of bandwidth

  usage and CPU usage, which may significantly lower performance in high

  latency environments.

 

 

option http-pretend-keepalive

no option http-pretend-keepalive

  Define whether haproxy will announce keepalive to the server or not

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  When running with "option http-server-close" or "option forceclose", haproxy

  adds a "Connection: close" header to the request forwarded to the server.

  Unfortunately, when some servers see this header, they automatically refrain

  from using the chunked encoding for responses of unknown length, while this

  is totally unrelated. The immediate effect is that this prevents haproxy from

  maintaining the client connection alive. A second effect is that a client or

  a cache could receive an incomplete response without being aware of it, and

  consider the response complete.

 

  By setting "option http-pretend-keepalive", haproxy will make the server

  believe it will keep the connection alive. The server will then not fall back

  to the abnormal undesired above. When haproxy gets the whole response, it

  will close the connection with the server just as it would do with the

  "forceclose" option. That way the client gets a normal response and the

  connection is correctly closed on the server side.

 

  It is recommended not to enable this option by default, because most servers

  will more efficiently close the connection themselves after the last packet,

  and release its buffers slightly earlier. Also, the added packet on the

  network could slightly reduce the overall peak performance. However it is

  worth noting that when this option is enabled, haproxy will have slightly

  less work to do. So if haproxy is the bottleneck on the whole architecture,

  enabling this option might save a few CPU cycles.

 

  This option may be set both in a frontend and in a backend. It is enabled if

  at least one of the frontend or backend holding a connection has it enabled.

  This option may be compbined with "option httpclose", which will cause

  keepalive to be announced to the server and close to be announced to the

  client. This practice is discouraged though.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option forceclose" and "option http-server-close"

 

 

option http-server-close

no option http-server-close

  Enable or disable HTTP connection closing on the server side

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  By default, when a client communicates with a server, HAProxy will only

  analyze, log, and process the first request of each connection. Setting

  "option http-server-close" enables HTTP connection-close mode on the server

  side while keeping the ability to support HTTP keep-alive and pipelining on

  the client side.  This provides the lowest latency on the client side (slow

  network) and the fastest session reuse on the server side to save server

  resources, similarly to "option forceclose". It also permits non-keepalive

  capable servers to be served in keep-alive mode to the clients if they

  conform to the requirements of RFC2616. Please note that some servers do not

  always conform to those requirements when they see "Connection: close" in the

  request. The effect will be that keep-alive will never be used. A workaround

  consists in enabling "option http-pretend-keepalive".

 

  At the moment, logs will not indicate whether requests came from the same

  session or not. The accept date reported in the logs corresponds to the end

  of the previous request, and the request time corresponds to the time spent

  waiting for a new request. The keep-alive request time is still bound to the

  timeout defined by "timeout http-keep-alive" or "timeout http-request" if

  not set.

 

  This option may be set both in a frontend and in a backend. It is enabled if

  at least one of the frontend or backend holding a connection has it enabled.

  It is worth noting that "option forceclose" has precedence over "option

  http-server-close" and that combining "http-server-close" with "httpclose"

  basically achieve the same result as "forceclose".

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option forceclose", "option http-pretend-keepalive",

             "option httpclose" and "1.1. The HTTP transaction model".

 

 

option http-use-proxy-header

no option http-use-proxy-header

  Make use of non-standard Proxy-Connection header instead of Connection

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  While RFC2616 explicitly states that HTTP/1.1 agents must use the

  Connection header to indicate their wish of persistent or non-persistent

  connections, both browsers and proxies ignore this header for proxied

  connections and make use of the undocumented, non-standard Proxy-Connection

  header instead. The issue begins when trying to put a load balancer between

  browsers and such proxies, because there will be a difference between what

  haproxy understands and what the client and the proxy agree on.

 

  By setting this option in a frontend, haproxy can automatically switch to use

  that non-standard header if it sees proxied requests. A proxied request is

  defined here as one where the URI begins with neither a '/' nor a '*'. The

  choice of header only affects requests passing through proxies making use of

  one of the "httpclose", "forceclose" and "http-server-close" options. Note

  that this option can only be specified in a frontend and will affect the

  request along its whole life.

 

  Also, when this option is set, a request which requires authentication will

  automatically switch to use proxy authentication headers if it is itself a

  proxied request. That makes it possible to check or enforce authentication in

  front of an existing proxy.

 

  This option should normally never be used, except in front of a proxy.

 

  See also : "option httpclose", "option forceclose" and "option

             http-server-close".

 

 

option httpchk

option httpchk <uri>

option httpchk <method> <uri>

option httpchk <method> <uri> <version>

  Enable HTTP protocol to check on the servers health

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <method>  is the optional HTTP method used with the requests. When not set,

              the "OPTIONS" method is used, as it generally requires low server

              processing and is easy to filter out from the logs. Any method

              may be used, though it is not recommended to invent non-standard

              ones.

 

    <uri>     is the URI referenced in the HTTP requests. It defaults to " / "

              which is accessible by default on almost any server, but may be

              changed to any other URI. Query strings are permitted.

 

    <version> is the optional HTTP version string. It defaults to "HTTP/1.0"

              but some servers might behave incorrectly in HTTP 1.0, so turning

              it to HTTP/1.1 may sometimes help. Note that the Host field is

              mandatory in HTTP/1.1, and as a trick, it is possible to pass it

              after "\r\n" following the version string.

 

  By default, server health checks only consist in trying to establish a TCP

  connection. When "option httpchk" is specified, a complete HTTP request is

  sent once the TCP connection is established, and responses 2xx and 3xx are

  considered valid, while all other ones indicate a server failure, including

  the lack of any response.

 

  The port and interval are specified in the server configuration.

 

  This option does not necessarily require an HTTP backend, it also works with

  plain TCP backends. This is particularly useful to check simple scripts bound

  to some dedicated ports using the inetd daemon.

 

  Examples :

      # Relay HTTPS traffic to Apache instance and check service availability

      # using HTTP request "OPTIONS * HTTP/1.1" on port 80.

      backend https_relay

          mode tcp

          option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www

          server apache1 192.168.1.1:443 check port 80

 

  See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",

             "http-check" and the "check", "port" and "inter" server options.

 

 

option httpclose

no option httpclose

  Enable or disable passive HTTP connection closing

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  By default, when a client communicates with a server, HAProxy will only

  analyze, log, and process the first request of each connection. If "option

  httpclose" is set, it will check if a "Connection: close" header is already

  set in each direction, and will add one if missing. Each end should react to

  this by actively closing the TCP connection after each transfer, thus

  resulting in a switch to the HTTP close mode. Any "Connection" header

  different from "close" will also be removed.

 

  It seldom happens that some servers incorrectly ignore this header and do not

  close the connection eventhough they reply "Connection: close". For this

  reason, they are not compatible with older HTTP 1.0 browsers. If this happens

  it is possible to use the "option forceclose" which actively closes the

  request connection once the server responds. Option "forceclose" also

  releases the server connection earlier because it does not have to wait for

  the client to acknowledge it.

 

  This option may be set both in a frontend and in a backend. It is enabled if

  at least one of the frontend or backend holding a connection has it enabled.

  If "option forceclose" is specified too, it has precedence over "httpclose".

  If "option http-server-close" is enabled at the same time as "httpclose", it

  basically achieves the same result as "option forceclose".

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option forceclose", "option http-server-close" and

             "1.1. The HTTP transaction model".

 

 

option httplog [ clf ]

  Enable logging of HTTP request, session state and timers

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    clf       if the "clf" argument is added, then the output format will be

              the CLF format instead of HAProxy's default HTTP format. You can

              use this when you need to feed HAProxy's logs through a specific

              log analyser which only support the CLF format and which is not

              extensible.

 

  By default, the log output format is very poor, as it only contains the

  source and destination addresses, and the instance name. By specifying

  "option httplog", each log line turns into a much richer format including,

  but not limited to, the HTTP request, the connection timers, the session

  status, the connections numbers, the captured headers and cookies, the

  frontend, backend and server name, and of course the source address and

  ports.

 

  This option may be set either in the frontend or the backend.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it. Specifying

  only "option httplog" will automatically clear the 'clf' mode if it was set

  by default.

 

  See also :  section 8 about logging.

 

 

option http_proxy

no option http_proxy

  Enable or disable plain HTTP proxy mode

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  It sometimes happens that people need a pure HTTP proxy which understands

  basic proxy requests without caching nor any fancy feature. In this case,

  it may be worth setting up an HAProxy instance with the "option http_proxy"

  set. In this mode, no server is declared, and the connection is forwarded to

  the IP address and port found in the URL after the "http://" scheme.

 

  No host address resolution is performed, so this only works when pure IP

  addresses are passed. Since this option's usage perimeter is rather limited,

  it will probably be used only by experts who know they need exactly it. Last,

  if the clients are susceptible of sending keep-alive requests, it will be

  needed to add "option httpclose" to ensure that all requests will correctly

  be analyzed.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  Example :

    # this backend understands HTTP proxy requests and forwards them directly.

    backend direct_forward

        option httpclose

        option http_proxy

 

  See also : "option httpclose"

 

 

option independant-streams

no option independant-streams

  Enable or disable independant timeout processing for both directions

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |  yes

  Arguments : none

 

  By default, when data is sent over a socket, both the write timeout and the

  read timeout for that socket are refreshed, because we consider that there is

  activity on that socket, and we have no other means of guessing if we should

  receive data or not.

 

  While this default behaviour is desirable for almost all applications, there

  exists a situation where it is desirable to disable it, and only refresh the

  read timeout if there are incoming data. This happens on sessions with large

  timeouts and low amounts of exchanged data such as telnet session. If the

  server suddenly disappears, the output data accumulates in the system's

  socket buffers, both timeouts are correctly refreshed, and there is no way

  to know the server does not receive them, so we don't timeout. However, when

  the underlying protocol always echoes sent data, it would be enough by itself

  to detect the issue using the read timeout. Note that this problem does not

  happen with more verbose protocols because data won't accumulate long in the

  socket buffers.

 

  When this option is set on the frontend, it will disable read timeout updates

  on data sent to the client. There probably is little use of this case. When

  the option is set on the backend, it will disable read timeout updates on

  data sent to the server. Doing so will typically break large HTTP posts from

  slow lines, so use it with caution.

 

  See also : "timeout client" and "timeout server"

 

 

option ldap-check

  Use LDAPv3 health checks for server testing

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  It is possible to test that the server correctly talks LDAPv3 instead of just

  testing that it accepts the TCP connection. When this option is set, an

  LDAPv3 anonymous simple bind message is sent to the server, and the response

  is analyzed to find an LDAPv3 bind response message.

 

  The server is considered valid only when the LDAP response contains success

  resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).

 

  Logging of bind requests is server dependent see your documentation how to

  configure it.

 

  Example :

        option ldap-check

 

  See also : "option httpchk"

 

 

option log-health-checks

no option log-health-checks

  Enable or disable logging of health checks

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |  yes

  Arguments : none

 

  Enable health checks logging so it possible to check for example what

  was happening before a server crash. Failed health check are logged if

  server is UP and succeeded health checks if server is DOWN, so the amount

  of additional information is limited.

 

  If health check logging is enabled no health check status is printed

  when servers is set up UP/DOWN/ENABLED/DISABLED.

 

  See also: "log" and section 8 about logging.

 

 

option log-separate-errors

no option log-separate-errors

  Change log level for non-completely successful connections

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  Sometimes looking for errors in logs is not easy. This option makes haproxy

  raise the level of logs containing potentially interesting information such

  as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The

  level changes from "info" to "err". This makes it possible to log them

  separately to a different file with most syslog daemons. Be careful not to

  remove them from the original file, otherwise you would lose ordering which

  provides very important information.

 

  Using this option, large sites dealing with several thousand connections per

  second may log normal traffic to a rotating buffer and only archive smaller

  error logs.

 

  See also : "log", "dontlognull", "dontlog-normal" and section 8 about

             logging.

 

 

option logasap

no option logasap

  Enable or disable early logging of HTTP requests

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

  Arguments : none

 

  By default, HTTP requests are logged upon termination so that the total

  transfer time and the number of bytes appear in the logs. When large objects

  are being transferred, it may take a while before the request appears in the

  logs. Using "option logasap", the request gets logged as soon as the server

  sends the complete headers. The only missing information in the logs will be

  the total number of bytes which will indicate everything except the amount

  of data transferred, and the total time which will not take the transfer

  time into account. In such a situation, it's a good practice to capture the

  "Content-Length" response header so that the logs at least indicate how many

  bytes are expected to be transferred.

 

  Examples :

      listen http_proxy 0.0.0.0:80

          mode http

          option httplog

          option logasap

          log 192.168.2.200 local3

 

    >>> Feb  6 12:14:14 localhost \

          haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \

          static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \

          "GET /image.iso HTTP/1.0"

 

  See also : "option httplog", "capture response header", and section 8 about

             logging.

 

 

option mysql-check [ user <username> ]

  Use MySQL health checks for server testing

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <username> This is the username which will be used when connecting to MySQL

               server.

 

  If you specify a username, the check consists of sending two MySQL packet,

  one Client Authentication packet, and one QUIT packet, to correctly close

  MySQL session. We then parse the MySQL Handshake Initialisation packet and/or

  Error packet. It is a basic but useful test which does not produce error nor

  aborted connect on the server. However, it requires adding an authorization

  in the MySQL table, like this :

 

      USE mysql;

      INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');

      FLUSH PRIVILEGES;

 

  If you don't specify a username (it is deprecated and not recommended), the

  check only consists in parsing the Mysql Handshake Initialisation packet or

  Error packet, we don't send anything in this mode. It was reported that it

  can generate lockout if check is too frequent and/or if there is not enough

  traffic. In fact, you need in this case to check MySQL "max_connect_errors"

  value as if a connection is established successfully within fewer than MySQL

  "max_connect_errors" attempts after a previous connection was interrupted,

  the error count for the host is cleared to zero. If HAProxy's server get

  blocked, the "FLUSH HOSTS" statement is the only way to unblock it.

 

  Remember that this does not check database presence nor database consistency.

  To do this, you can use an external check with xinetd for example.

 

  The check requires MySQL >=3.22, for older version, please use TCP check.

 

  Most often, an incoming MySQL server needs to see the client's IP address for

  various purposes, including IP privilege matching and connection logging.

  When possible, it is often wise to masquerade the client's IP address when

  connecting to the server using the "usesrc" argument of the "source" keyword,

  which requires the cttproxy feature to be compiled in, and the MySQL server

  to route the client via the machine hosting haproxy.

 

  See also: "option httpchk"

 

 

option nolinger

no option nolinger

  Enable or disable immediate session resource cleaning after close

  May be used in sections:    defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  When clients or servers abort connections in a dirty way (eg: they are

  physically disconnected), the session timeouts triggers and the session is

  closed. But it will remain in FIN_WAIT1 state for some time in the system,

  using some resources and possibly limiting the ability to establish newer

  connections.

 

  When this happens, it is possible to activate "option nolinger" which forces

  the system to immediately remove any socket's pending data on close. Thus,

  the session is instantly purged from the system's tables. This usually has

  side effects such as increased number of TCP resets due to old retransmits

  getting immediately rejected. Some firewalls may sometimes complain about

  this too.

 

  For this reason, it is not recommended to use this option when not absolutely

  needed. You know that you need it when you have thousands of FIN_WAIT1

  sessions on your system (TIME_WAIT ones do not count).

 

  This option may be used both on frontends and backends, depending on the side

  where it is required. Use it on the frontend for clients, and on the backend

  for servers.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

 

option originalto [ except <network> ] [ header <name> ]

  Enable insertion of the X-Original-To header to requests sent to servers

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments :

    <network> is an optional argument used to disable this option for sources

              matching <network>

    <name>    an optional argument to specify a different "X-Original-To"

              header name.

 

  Since HAProxy can work in transparent mode, every request from a client can

  be redirected to the proxy and HAProxy itself can proxy every request to a

  complex SQUID environment and the destination host from SO_ORIGINAL_DST will

  be lost. This is annoying when you want access rules based on destination ip

  addresses. To solve this problem, a new HTTP header "X-Original-To" may be

  added by HAProxy to all requests sent to the server. This header contains a

  value representing the original destination IP address. Since this must be

  configured to always use the last occurrence of this header only. Note that

  only the last occurrence of the header must be used, since it is really

  possible that the client has already brought one.

 

  The keyword "header" may be used to supply a different header name to replace

  the default "X-Original-To". This can be useful where you might already

  have a "X-Original-To" header from a different application, and you need

  preserve it. Also if your backend server doesn't use the "X-Original-To"

  header and requires different one.

 

  Sometimes, a same HAProxy instance may be shared between a direct client

  access and a reverse-proxy access (for instance when an SSL reverse-proxy is

  used to decrypt HTTPS traffic). It is possible to disable the addition of the

  header for a known source address or network by adding the "except" keyword

  followed by the network address. In this case, any source IP matching the

  network will not cause an addition of this header. Most common uses are with

  private networks or 127.0.0.1.

 

  This option may be specified either in the frontend or in the backend. If at

  least one of them uses it, the header will be added. Note that the backend's

  setting of the header subargument takes precedence over the frontend's if

  both are defined.

 

  It is important to note that by default, HAProxy works in tunnel mode and

  only inspects the first request of a connection, meaning that only the first

  request will have the header appended, which is certainly not what you want.

  In order to fix this, ensure that any of the "httpclose", "forceclose" or

  "http-server-close" options is set when using this option.

 

  Examples :

    # Original Destination address

    frontend www

        mode http

        option originalto except 127.0.0.1

 

    # Those servers want the IP Address in X-Client-Dst

    backend www

        mode http

        option originalto header X-Client-Dst

 

  See also : "option httpclose", "option http-server-close",

             "option forceclose"

 

 

option persist

no option persist

  Enable or disable forced persistence on down servers

  May be used in sections:    defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  When an HTTP request reaches a backend with a cookie which references a dead

  server, by default it is redispatched to another server. It is possible to

  force the request to be sent to the dead server first using "option persist"

  if absolutely needed. A common use case is when servers are under extreme

  load and spend their time flapping. In this case, the users would still be

  directed to the server they opened the session on, in the hope they would be

  correctly served. It is recommended to use "option redispatch" in conjunction

  with this option so that in the event it would not be possible to connect to

  the server at all (server definitely dead), the client would finally be

  redirected to another valid server.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option redispatch", "retries", "force-persist"

 

 

option redispatch

no option redispatch

  Enable or disable session redistribution in case of connection failure

  May be used in sections:    defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  In HTTP mode, if a server designated by a cookie is down, clients may

  definitely stick to it because they cannot flush the cookie, so they will not

  be able to access the service anymore.

 

  Specifying "option redispatch" will allow the proxy to break their

  persistence and redistribute them to a working server.

 

  It also allows to retry last connection to another server in case of multiple

  connection failures. Of course, it requires having "retries" set to a nonzero

  value.

 

  This form is the preferred form, which replaces both the "redispatch" and

  "redisp" keywords.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "redispatch", "retries", "force-persist"

 

 

option smtpchk

option smtpchk <hello> <domain>

  Use SMTP health checks for server testing

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <hello>   is an optional argument. It is the "hello" command to use. It can

              be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other

              values will be turned into the default command ("HELO").

 

    <domain>  is the domain name to present to the server. It may only be

              specified (and is mandatory) if the hello command has been

              specified. By default, "localhost" is used.

 

  When "option smtpchk" is set, the health checks will consist in TCP

  connections followed by an SMTP command. By default, this command is

  "HELO localhost". The server's return code is analyzed and only return codes

  starting with a "2" will be considered as valid. All other responses,

  including a lack of response will constitute an error and will indicate a

  dead server.

 

  This test is meant to be used with SMTP servers or relays. Depending on the

  request, it is possible that some servers do not log each connection attempt,

  so you may want to experiment to improve the behaviour. Using telnet on port

  25 is often easier than adjusting the configuration.

 

  Most often, an incoming SMTP server needs to see the client's IP address for

  various purposes, including spam filtering, anti-spoofing and logging. When

  possible, it is often wise to masquerade the client's IP address when

  connecting to the server using the "usesrc" argument of the "source" keyword,

  which requires the cttproxy feature to be compiled in.

 

  Example :

        option smtpchk HELO mydomain.org

 

  See also : "option httpchk", "source"

 

 

option socket-stats

no option socket-stats

 

  Enable or disable collecting & providing separate statistics for each socket.

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   no

 

  Arguments : none

 

 

option splice-auto

no option splice-auto

  Enable or disable automatic kernel acceleration on sockets in both directions

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  When this option is enabled either on a frontend or on a backend, haproxy

  will automatically evaluate the opportunity to use kernel tcp splicing to

  forward data between the client and the server, in either direction. Haproxy

  uses heuristics to estimate if kernel splicing might improve performance or

  not. Both directions are handled independently. Note that the heuristics used

  are not much aggressive in order to limit excessive use of splicing. This

  option requires splicing to be enabled at compile time, and may be globally

  disabled with the global option "nosplice". Since splice uses pipes, using it

  requires that there are enough spare pipes.

 

  Important note: kernel-based TCP splicing is a Linux-specific feature which

  first appeared in kernel 2.6.25. It offers kernel-based acceleration to

  transfer data between sockets without copying these data to user-space, thus

  providing noticeable performance gains and CPU cycles savings. Since many

  early implementations are buggy, corrupt data and/or are inefficient, this

  feature is not enabled by default, and it should be used with extreme care.

  While it is not possible to detect the correctness of an implementation,

  2.6.29 is the first version offering a properly working implementation. In

  case of doubt, splicing may be globally disabled using the global "nosplice"

  keyword.

 

  Example :

        option splice-auto

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option splice-request", "option splice-response", and global

             options "nosplice" and "maxpipes"

 

 

option splice-request

no option splice-request

  Enable or disable automatic kernel acceleration on sockets for requests

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  When this option is enabled either on a frontend or on a backend, haproxy

  will user kernel tcp splicing whenever possible to forward data going from

  the client to the server. It might still use the recv/send scheme if there

  are no spare pipes left. This option requires splicing to be enabled at

  compile time, and may be globally disabled with the global option "nosplice".

  Since splice uses pipes, using it requires that there are enough spare pipes.

 

  Important note: see "option splice-auto" for usage limitations.

 

  Example :

        option splice-request

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option splice-auto", "option splice-response", and global options

             "nosplice" and "maxpipes"

 

 

option splice-response

no option splice-response

  Enable or disable automatic kernel acceleration on sockets for responses

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  When this option is enabled either on a frontend or on a backend, haproxy

  will user kernel tcp splicing whenever possible to forward data going from

  the server to the client. It might still use the recv/send scheme if there

  are no spare pipes left. This option requires splicing to be enabled at

  compile time, and may be globally disabled with the global option "nosplice".

  Since splice uses pipes, using it requires that there are enough spare pipes.

 

  Important note: see "option splice-auto" for usage limitations.

 

  Example :

        option splice-response

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option splice-auto", "option splice-request", and global options

             "nosplice" and "maxpipes"

 

 

option srvtcpka

no option srvtcpka

  Enable or disable the sending of TCP keepalive packets on the server side

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  When there is a firewall or any session-aware component between a client and

  a server, and when the protocol involves very long sessions with long idle

  periods (eg: remote desktops), there is a risk that one of the intermediate

  components decides to expire a session which has remained idle for too long.

 

  Enabling socket-level TCP keep-alives makes the system regularly send packets

  to the other end of the connection, leaving it active. The delay between

  keep-alive probes is controlled by the system only and depends both on the

  operating system and its tuning parameters.

 

  It is important to understand that keep-alive packets are neither emitted nor

  received at the application level. It is only the network stacks which sees

  them. For this reason, even if one side of the proxy already uses keep-alives

  to maintain its connection alive, those keep-alive packets will not be

  forwarded to the other side of the proxy.

 

  Please note that this has nothing to do with HTTP keep-alive.

 

  Using option "srvtcpka" enables the emission of TCP keep-alive probes on the

  server side of a connection, which should help when session expirations are

  noticed between HAProxy and a server.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option clitcpka", "option tcpka"

 

 

option ssl-hello-chk

  Use SSLv3 client hello health checks for server testing

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  When some SSL-based protocols are relayed in TCP mode through HAProxy, it is

  possible to test that the server correctly talks SSL instead of just testing

  that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure

  SSLv3 client hello messages are sent once the connection is established to

  the server, and the response is analyzed to find an SSL server hello message.

  The server is considered valid only when the response contains this server

  hello message.

 

  All servers tested till there correctly reply to SSLv3 client hello messages,

  and most servers tested do not even log the requests containing only hello

  messages, which is appreciable.

 

  See also: "option httpchk"

 

 

option tcp-smart-accept

no option tcp-smart-accept

  Enable or disable the saving of one ACK packet during the accept sequence

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |    no

  Arguments : none

 

  When an HTTP connection request comes in, the system acknowledges it on

  behalf of HAProxy, then the client immediately sends its request, and the

  system acknowledges it too while it is notifying HAProxy about the new

  connection. HAProxy then reads the request and responds. This means that we

  have one TCP ACK sent by the system for nothing, because the request could

  very well be acknowledged by HAProxy when it sends its response.

 

  For this reason, in HTTP mode, HAProxy automatically asks the system to avoid

  sending this useless ACK on platforms which support it (currently at least

  Linux). It must not cause any problem, because the system will send it anyway

  after 40 ms if the response takes more time than expected to come.

 

  During complex network debugging sessions, it may be desirable to disable

  this optimization because delayed ACKs can make troubleshooting more complex

  when trying to identify where packets are delayed. It is then possible to

  fall back to normal behaviour by specifying "no option tcp-smart-accept".

 

  It is also possible to force it for non-HTTP proxies by simply specifying

  "option tcp-smart-accept". For instance, it can make sense with some services

  such as SMTP where the server speaks first.

 

  It is recommended to avoid forcing this option in a defaults section. In case

  of doubt, consider setting it back to automatic values by prepending the

  "default" keyword before it, or disabling it using the "no" keyword.

 

  See also : "option tcp-smart-connect"

 

 

option tcp-smart-connect

no option tcp-smart-connect

  Enable or disable the saving of one ACK packet during the connect sequence

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  On certain systems (at least Linux), HAProxy can ask the kernel not to

  immediately send an empty ACK upon a connection request, but to directly

  send the buffer request instead. This saves one packet on the network and

  thus boosts performance. It can also be useful for some servers, because they

  immediately get the request along with the incoming connection.

 

  This feature is enabled when "option tcp-smart-connect" is set in a backend.

  It is not enabled by default because it makes network troubleshooting more

  complex.

 

  It only makes sense to enable it with protocols where the client speaks first

  such as HTTP. In other situations, if there is no data to send in place of

  the ACK, a normal ACK is sent.

 

  If this option has been enabled in a "defaults" section, it can be disabled

  in a specific instance by prepending the "no" keyword before it.

 

  See also : "option tcp-smart-accept"

 

 

option tcpka

  Enable or disable the sending of TCP keepalive packets on both sides

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  When there is a firewall or any session-aware component between a client and

  a server, and when the protocol involves very long sessions with long idle

  periods (eg: remote desktops), there is a risk that one of the intermediate

  components decides to expire a session which has remained idle for too long.

 

  Enabling socket-level TCP keep-alives makes the system regularly send packets

  to the other end of the connection, leaving it active. The delay between

  keep-alive probes is controlled by the system only and depends both on the

  operating system and its tuning parameters.

 

  It is important to understand that keep-alive packets are neither emitted nor

  received at the application level. It is only the network stacks which sees

  them. For this reason, even if one side of the proxy already uses keep-alives

  to maintain its connection alive, those keep-alive packets will not be

  forwarded to the other side of the proxy.

 

  Please note that this has nothing to do with HTTP keep-alive.

 

  Using option "tcpka" enables the emission of TCP keep-alive probes on both

  the client and server sides of a connection. Note that this is meaningful

  only in "defaults" or "listen" sections. If this option is used in a

  frontend, only the client side will get keep-alives, and if this option is

  used in a backend, only the server side will get keep-alives. For this

  reason, it is strongly recommended to explicitly use "option clitcpka" and

  "option srvtcpka" when the configuration is split between frontends and

  backends.

 

  See also : "option clitcpka", "option srvtcpka"

 

 

option tcplog

  Enable advanced logging of TCP connections with session state and timers

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    yes   |   yes  |   yes

  Arguments : none

 

  By default, the log output format is very poor, as it only contains the

  source and destination addresses, and the instance name. By specifying

  "option tcplog", each log line turns into a much richer format including, but

  not limited to, the connection timers, the session status, the connections

  numbers, the frontend, backend and server name, and of course the source

  address and ports. This option is useful for pure TCP proxies in order to

  find which of the client or server disconnects or times out. For normal HTTP

  proxies, it's better to use "option httplog" which is even more complete.

 

  This option may be set either in the frontend or the backend.

 

  See also :  "option httplog", and section 8 about logging.

 

 

option transparent

no option transparent

  Enable client-side transparent proxying

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments : none

 

  This option was introduced in order to provide layer 7 persistence to layer 3

  load balancers. The idea is to use the OS's ability to redirect an incoming

  connection for a remote address to a local process (here HAProxy), and let

  this process know what address was initially requested. When this option is

  used, sessions without cookies will be forwarded to the original destination

  IP address of the incoming request (which should match that of another

  equipment), while requests with cookies will still be forwarded to the

  appropriate server.

 

  Note that contrary to a common belief, this option does NOT make HAProxy

  present the client's IP to the server when establishing the connection.

 

  See also: the "usesrc" argument of the "source" keyword, and the

            "transparent" option of the "bind" keyword.

 

 

persist rdp-cookie

persist rdp-cookie(name)

  Enable RDP cookie-based persistence

  May be used in sections :   defaults | frontend | listen | backend

                                 yes   |    no    |   yes  |   yes

  Arguments :

    <name>    is the optional name of the RDP cookie to check. If omitted, the

              default cookie name "msts" will be used. There currently is no

              valid reason to change this name.

 

  This statement enables persistence based on an RDP cookie. The RDP cookie

  contains all information required to find the server in the list of known

  servers. So when this option is set in the backend, the request is analysed

  and if an RDP cookie is found, it is decoded. If it matches a known server

  which is still UP (or if "option persist" is set), then the connection is

  forwarded to this server.

 

  Note that this only makes sense in a TCP backend, but for this to work, the

  frontend must have waited long enough to ensure that an RDP cookie is present

  in the request buffer. This is the same requirement as with the "rdp-cookie"

  load-balancing method. Thus it is highly recommended to put all statements in

  a single "listen" section.

 

  Also, it is important to understand that the terminal server will emit this

  RDP cookie only if it is configured for "token redirection mode", which means

  that the "IP address redirection" option is disabled.