Installation instructions for HAProxy
=====================================

HAProxy 2.4 is a long-term supported version, which means that it will get
fixes for bugs as they are discovered till around Q2 2026 and will not receive
new features. This version is suitable for general deployment as it is expected
to receive less frequent updates than regular stable branches which have an odd
digit in the minor version number. New users are strongly advised encouraged to
use only such long-term supported versions such as the ones provided by their
software vendor or Linux distribution. If for any reason you would prefer a
different version than the one packaged for your system, you want to be certain
to have all the fixes or to get some commercial support, other choices are
available at http://www.haproxy.com/.


Areas covered in this document
==============================

1) Quick build & install
2) Basic principles
3) Build environment
4) Dependencies
5) Advanced build options
6) How to install HAProxy


1) Quick build & install
========================

If you've already built HAProxy and are just looking for a quick reminder, here
are a few build examples :

  - recent Linux system with all options, make and install :
    $ make clean
    $ make -j $(nproc) TARGET=linux-glibc \
                USE_OPENSSL=1 USE_LUA=1 USE_PCRE=1 USE_SYSTEMD=1
    $ sudo make install

  - FreeBSD and OpenBSD, build with all options :
    $ gmake -j 4 TARGET=freebsd USE_OPENSSL=1 USE_LUA=1 USE_PCRE=1

  - embedded Linux, build using a cross-compiler :
    $ make -j $(nproc) TARGET=linux-glibc USE_OPENSSL=1 USE_PCRE=1 \
                CC=/opt/cross/gcc730-arm/bin/gcc ADDLIB=-latomic

  - Build with static PCRE on Solaris / UltraSPARC :
    $ make TARGET=solaris CPU=ultrasparc USE_STATIC_PCRE=1

For more advanced build options or if a command above reports an error, please
read the following sections.


2) Basic principles
===================

HAProxy uses a single GNU Makefile which supports options on the command line,
so that there is no need to hack a "configure" file to work on your system. The
makefile totally supports parallel build using "make -j <jobs>" where <jobs>
matches the number of usable processors, which on some platforms is returned by
the "nproc" utility. The explanations below may occasionally refer to some
options, usually in the form "name=value", which have to be passed to the
command line. This means that the option has to be passed after the "make"
command. For example :

  $ make -j $(nproc) TARGET=generic USE_GZIP=1

One required option is TARGET, it must be set to a target platform name, which
provides a number of presets. The list of known platforms is displayed when no
target is specified. It is not strictly required to use the exact target, you
can use a relatively similar one and adjust specific variables by hand.

Most configuration variables are in fact booleans. Some options are detected and
enabled by default if available on the target platform. This is the case for all
those named "USE_<feature>". These booleans are enabled by "USE_<feature>=1"
and are disabled by "USE_<feature>=" (with no value). An exhaustive list of the
supported USE_* features is located at the top of the main Makefile. The last
occurrence of such an option on the command line overrides any previous one.
Example :

  $ make TARGET=generic USE_THREAD=

In case of error or missing TARGET, a help screen is displayed. It is also
possible to display a list of all known options using "make help".

Some optional components which may depend on third-party libraries, are used
with popular tools which are not necessarily standard implementations, or are
maintained at slower pace than the core of the project, are located in the
"addons/" directory. These ones may disappear in a future version if the
product they depend on disappears or if their maintainers do not assign enough
resources to maintain them any more. For this reason they are not built by
default, but some USE_* options are usually provided for them, and their build
is routinely tested anyway.


3) Build environment
====================

HAProxy requires a working GCC or Clang toolchain and GNU make :

  - GNU make >= 3.80. Note that neither Solaris nor OpenBSD's make work with
    the GNU Makefile. If you get many syntax errors when running "make", you
    may want to retry with "gmake" which is the name commonly used for GNU make
    on BSD systems.

  - GCC >= 3.4 (up to 11 tested). Older versions can be made to work with a
    few minor adaptations if really needed. Newer versions may sometimes break
    due to compiler regressions or behaviour changes. The version shipped with
    your operating system is very likely to work with no trouble. Clang >= 3.0
    is also known to work as an alternative solution. Recent versions may emit
    a bit more warnings that are worth reporting as they may reveal real bugs.
    TCC (https://repo.or.cz/tinycc.git) is also usable for developers but will
    not support threading and was found at least once to produce bad code in
    some rare corner cases (since fixed). But it builds extremely quickly
    (typically half a second for the whole project) and is very convenient to
    run quick tests during API changes or code refactoring.

  - GNU ld (binutils package), with no particular version. Other linkers might
    work but were not tested.

On debian or Ubuntu systems and their derivatives, you may get all these tools
at once by issuing the two following commands :

  $ sudo apt-get update
  $ sudo apt-get install build-essential

On Fedora, CentOS, RHEL and derivatives, you may get the equivalent packages
with the following command :

  $ sudo yum groupinstall "Development Tools"

Please refer to your operating system's documentation for other systems.

It is also possible to build HAProxy for another system or platform using a
cross-compiler but in this case you probably already have installed these
tools.

Building HAProxy may require between 60 and 80 MB of free space in the
directory where the sources have been extracted, depending on the debugging
options involved.


4) Dependencies
===============

HAProxy in its basic form does not depend on anything beyond a working libc.
However a number of options are enabled by default, or are highly recommended,
and these options will typically involve some external components or libraries,
depending on the targeted platform.

Optional dependencies may be split into several categories :

  - memory allocation
  - regular expressions
  - multi-threading
  - password encryption
  - cryptography
  - compression
  - lua
  - device detection
  - miscellaneous


4.1) Memory allocation
----------------------
By default, HAProxy uses the standard malloc() call provided by the libc. It
may also be built to use jemalloc, which is fast and thread-safe. In order to
use it, please add "-ljemalloc" to the ADDLIB variable. You may possibly also
need to append "-lpthread" and/or "-ldl" depending on the operating system.


4.2) Regular expressions
------------------------
HAProxy may make use regular expressions (regex) to match certain patterns. The
regex engine is provided by default in the libc. On some operating systems, it
might happen that the original regex library provided by the libc is too slow,
too limited or even bogus. For example, on older Solaris versions up to 8, the
default regex used not to properly extract group references, without reporting
compilation errors. Also, some early versions of the GNU libc used to include a
regex engine which could be slow or even crash on certain patterns.

If you plan on importing a particularly heavy configuration involving a lot of
regex, you may benefit from using some alternative regex implementations such as
PCRE. HAProxy natively supports PCRE and PCRE2, both in standard and JIT
flavors (Just In Time). The following options are available depending on the
library version provided on your system :

  - "USE_PCRE=1"         : enable PCRE version 1, dynamic linking
  - "USE_STATIC_PCRE=1"  : enable PCRE version 1, static linking
  - "USE_PCRE_JIT=1"     : enable PCRE version 1 in JIT mode
  - "USE_PCRE2=1"        : enable PCRE version 2, dynamic linking
  - "USE_STATIC_PCRE2=1" : enable PCRE version 2, static linking
  - "USE_PCRE2_JIT=1"    : enable PCRE version 2 in JIT mode

Both of these libraries may be downloaded from https://www.pcre.org/.

By default, the include and library paths are figured from the "pcre-config"
and "pcre2-config" utilities. If these ones are not installed or inaccurate
(for example when cross-compiling), it is possible to force the path to include
files using "PCRE_INC" and "PCRE2_INC" respectively, and the path to library
files using "PCRE_LIB" and "PCRE2_LIB" respectively. For example :

  $ make TARGET=generic \
    USE_PCRE2_JIT=1 PCRE2_INC=/opt/cross/include PCRE2_LIB=/opt/cross/lib


4.3) Multi-threading
--------------------
On some systems for which positive feedback was reported, multi-threading will
be enabled by default. When multi-threading is used, the libpthread library
(POSIX threading) will be used. If the target system doesn't contain such a
library, it is possible to forcefully disable multi-threading by adding
"USE_THREAD=" on the command line.


4.4) Password encryption
------------------------
Many systems provide password encryption functions used for authentication. On
some systems these functions are part of the libc. On others, they're part of a
separate library called "libcrypt". The default targets are pre-configured
based on which system needs the library. It is possible to forcefully disable
the linkage against libcrypt by adding "USE_LIBCRYPT=" on the command line, or
to forcefully enable it using "USE_LIBCRYPT=1".


4.5) Cryptography
-----------------
For SSL/TLS, it is necessary to use a cryptography library. HAProxy currently
supports the OpenSSL library, and is known to build ant work with branches
0.9.8, 1.0.0, 1.0.1, 1.0.2, 1.1.0 and 1.1.1. OpenSSL follows a long-term
support cycle similar to HAProxy's, and each of the branches above receives its
own fixes, without forcing you to upgrade to another branch. There is no excuse
for staying vulnerable by not applying a fix available for your version. There
is always a small risk of regression when jumping from one branch to another
one, especially when it's very new, so it's preferable to observe for a while
if you use a different version than your system's defaults.

Two OpenSSL derivatives called LibreSSL and BoringSSL are reported to work as
well. While there are some efforts from the community to ensure they work well,
OpenSSL remains the primary target and this means that in case of conflicting
choices, OpenSSL support will be favored over other options.

In order to enable SSL/TLS support, simply pass "USE_OPENSSL=1" on the command
line and the default library present on your system will be used :

  $ make TARGET=generic USE_OPENSSL=1

If you want to use a different version from the one provided by your system
(which is not recommended due to the risk of missing security fixes), it is
possible to indicate the path to the SSL include files using SSL_INC, and the
SSL library files using SSL_LIB. Example :

  $ make TARGET=generic \
    USE_OPENSSL=1 SSL_INC=/opt/ssl-1.1.1/include SSL_LIB=/opt/ssl-1.1.1/lib

In order to link OpenSSL statically against HAProxy, first download OpenSSL
from https://www.openssl.org/ then build it with the "no-shared" keyword and
install it to a local directory, so your system is not affected :

  $ export STATICLIBSSL=/tmp/staticlibssl
  $ ./config --prefix=$STATICLIBSSL no-shared
  $ make && make install_sw

Then when building haproxy, pass that path via SSL_INC and SSL_LIB :

  $ make TARGET=generic \
    USE_OPENSSL=1 SSL_INC=$STATICLIBSSL/include SSL_LIB=$STATICLIBSSL/lib

When building with OpenSSL on some systems, you may also need to enable support
for the "libz" library, which is visible if the linker complains about function
"deflateInit()" not being found. In this case, simply append "ADDLIB=-lz" to
the command line.

It is worth mentioning that asynchronous cryptography engines are supported on
OpenSSL 1.1.0 and above. Such engines are used to access hardware cryptography
acceleration that might be present on your system.


4.6) Compression
----------------
HAProxy can compress HTTP responses before delivering them to clients, in order
to save network bandwidth. Two compression options are available. The first one
relies on the libslz library (http://libslz.org) that is embedded in haproxy.
It is enabled by default as it is very fast and does not keep a copy of the
contents in memory. It is possible to disable it, for example for very small
systems, by passing "USE_SLZ=" to the "make" command.

Please note that SLZ will benefit from some CPU-specific instructions like the
availability of the CRC32 extension on some ARM processors. Thus it can further
improve its performance to build with "CPU=native" on the target system, or
"CPU=armv81" (modern systems such as Graviton2 or A55/A75 and beyond),
"CPU=a72" (e.g. for RPi4, or AWS Graviton), "CPU=a53" (e.g. for RPi3), or
"CPU=armv8-auto" (automatic detection with minor runtime penalty).

A second option involves the widely known zlib library, which is very likely
installed on your system. In order to use zlib, simply pass "USE_ZLIB=1" to the
"make" command line, which will also automatically disable SLZ. If the library
is not installed in your default system's path, it is possible to specify the
path to the include files using ZLIB_INC, and the path to the library files
using ZLIB_LIB :

  $ make TARGET=generic \
    USE_ZLIB=1 ZLIB_INC=/opt/zlib-1.2.11/include ZLIB_LIB=/opt/zlib-1.2.11/lib

Zlib is commonly found on most systems, otherwise updates can be retrieved from
http://www.zlib.net/. It is easy and fast to build, and new versions sometimes
provide better performance so it might be worth using an up-to-date one.

Zlib compresses a bit better than libslz but at the expense of more CPU usage
(about 3.5 times more minimum), and a huge memory usage (~260 kB per compressed
stream). The only valid reason for uzing Zlib instead of SLZ here usually is to
deal with a very limited internet bandwidth while CPU and RAM are abundant so
that the last few percent of compression ratio are worth the invested hardware.


4.7) Lua
--------
Lua is an embedded programming language supported by HAProxy to provide more
advanced scripting capabilities. Only versions 5.3 and above are supported.
In order to enable Lua support, please specify "USE_LUA=1" on the command line.
Some systems provide this library under various names to avoid conflicts with
previous versions. By default, HAProxy looks for "lua5.3", "lua53", "lua". If
your system uses a different naming, you may need to set the library name in
the "LUA_LIB_NAME" variable.

If Lua is not provided on your system, it can be very simply built locally. It
can be downloaded from https://www.lua.org/, extracted and built, for example :

  $ cd /opt/lua-5.3.5
  $ make linux

The path to the include files and library files may be set using "LUA_INC" and
"LUA_LIB" respectively. For example :

  $ make TARGET=generic \
    USE_LUA=1 LUA_INC=/opt/lua-5.3.5/src LUA_LIB=/opt/lua-5.3.5/src


4.8) Device detection
---------------------
HAProxy supports several device detection modules relying on third party
products. Some of them may provide free code, others free libs, others free
evaluation licenses. Please read about their respective details in the
following files :

    doc/DeviceAtlas-device-detection.txt   for DeviceAtlas
    doc/51Degrees-device-detection.txt     for 51Degrees
    doc/WURFL-device-detection.txt         for Scientiamobile WURFL


4.9) Miscellaneous
------------------
Some systems have specificities. Usually these specificities are known and/or
detected and properly set for you. If you need to adjust the behaviour, here
are the extra libraries that may be referenced at build time :

  - USE_RT=1      build with librt, which is sometimes needed on some systems
                  when using threads. It is set by default on Linux platforms,
                  and may be disabled using "USE_RT=" if your system doesn't
                  have one. You may have to set it as well if you face an error
                  indicating that clock_gettime() was not found.

  - USE_DL=1      build with libdl, which is usually needed for Lua and OpenSSL
                  on Linux. It is automatically detected and may be disabled
                  using "USE_DL=", though it should never harm.

  - USE_SYSTEMD=1 enables support for the sdnotify features of systemd,
                  allowing better integration with systemd on Linux systems
                  which come with it. It is never enabled by default so there
                  is no need to disable it.


4.10) Common errors
-------------------
Some build errors may happen depending on the options combinations or the
selected target. When facing build errors, if you know that your system is a
bit special or particularly old, start from TARGET=generic, it is easier to
start from there and fix the remaining issues than trying to degrade another
target. Common issues may include:

  - clock_gettime() not found
    => your system needs USE_RT=1

  - __sync_sub_and_fetch undefined in cache.o
    => your system needs either USE_PTHREAD_PSHARED=1 or USE_PRIVATE_CACHE=1

  - many __sync_<something> errors in many files
    => your gcc is too old, build without threads and with private cache.

  - many openssl errors
    => your OpenSSL version really is too old, do not enable OpenSSL


5) How to build HAProxy
=======================

This section assumes that you have already read section 2 (basic principles)
and section 3 (build environment). It often refers to section 4 (dependencies).

To build haproxy, you have to choose your target OS amongst the following ones
and assign it to the TARGET variable :

  - linux-glibc         for Linux kernel 2.6.28 and above
  - linux-glibc-legacy  for Linux kernel 2.6.28 and above without new features
  - linux-musl          for Linux kernel 2.6.28 and above with musl libc
  - solaris             for Solaris 10 and above
  - freebsd             for FreeBSD 10 and above
  - dragonfly           for DragonFlyBSD 4.3 and above
  - netbsd              for NetBSD 8 and above
  - osx                 for Mac OS/X
  - openbsd             for OpenBSD 6.3 and above
  - aix51               for AIX 5.1
  - aix52               for AIX 5.2
  - aix72-gcc           for AIX 7.2 (using gcc)
  - cygwin              for Cygwin
  - haiku               for Haiku
  - generic             for any other OS or version.
  - custom              to manually adjust every setting

You may also choose your CPU to benefit from some optimizations. This is
particularly important on UltraSparc machines. For this, you can assign
one of the following choices to the CPU variable :

  - i686 for intel PentiumPro, Pentium 2 and above, AMD Athlon (32 bits)
  - i586 for intel Pentium, AMD K6, VIA C3.
  - ultrasparc : Sun UltraSparc I/II/III/IV processor
  - power8 : IBM POWER8 processor
  - power9 : IBM POWER9 processor
  - armv81 : modern ARM cores (Cortex A55/A75/A76/A78/X1, Neoverse, Graviton2)
  - a72    : ARM Cortex-A72 or A73 (e.g. RPi4, Odroid N2, AWS Graviton)
  - a53    : ARM Cortex-A53 or any of its successors in 64-bit mode (e.g. RPi3)
  - armv8-auto : support both older and newer armv8 cores with a minor penalty,
                 thanks to gcc 10's outline atomics (default with gcc 10.2).
  - native : use the build machine's specific processor optimizations. Use with
    extreme care, and never in virtualized environments (known to break).
  - generic : any other processor or no CPU-specific optimization. (default)

Alternatively, you may just set the CPU_CFLAGS value to the optimal GCC options
for your platform. A second variable named SMALL_OPTS also supports passing a
number of defines and compiler options usually for small systems. For better
clarity it's recommended to pass the options which result in a smaller binary
(like memory limits or -Os) into this variable.

If you are building for a different system than the one you're building on,
this is called "cross-compiling". HAProxy supports cross-compilation pretty
well and tries to ease it by letting you adjust paths to all libraries (please
read section 4 on dependencies for more details). When cross-compiling, you
just need to pass the path to your compiler in the "CC" variable, and the path
to the linker in the "LD" variable. Most of the time, setting the CC variable
is enough since LD points to it by default.

By default the build process runs in quiet mode and hide the details of the
commands that are executed. This allows to more easily catch build warnings
and see what is happening. However it is not convenient at all to observe what
flags are passed to the compiler nor what compiler is involved. Simply append
"V=1" to the "make" command line to switch to verbose mode and display the
details again. It is recommended to use this option when cross-compiling to
verify that the paths are correct and that /usr/include is never invovled.

You may want to build specific target binaries which do not match your native
compiler's target. This is particularly true on 64-bit systems when you want
to build a 32-bit binary. Use the ARCH variable for this purpose. Right now
it only knows about a few x86 variants (i386,i486,i586,i686,x86_64), two
generic ones (32,64) and sets -m32/-m64 as well as -march=<arch> accordingly.
This variable is only used to set ARCH_FLAGS to preset values, so if you know
the arch-specific flags that your system needs, you may prefer to set
ARCH_FLAGS instead. Note that these flags are passed both to the compiler and
to the linker. For example, in order to build a 32-bit binary on an x86_64
Linux system with SSL support without support for compression but when OpenSSL
requires ZLIB anyway :

    $ make TARGET=linux-glibc ARCH=i386 USE_OPENSSL=1 ADDLIB=-lz

Recent systems can resolve IPv6 host names using getaddrinfo(). This primitive
is not present in all libcs and does not work in all of them either. Support in
glibc was broken before 2.3. Some embedded libs may not properly work either,
thus, support is disabled by default, meaning that some host names which only
resolve as IPv6 addresses will not resolve and configs might emit an error
during parsing. If you know that your OS libc has reliable support for
getaddrinfo(), you can add USE_GETADDRINFO=1 on the make command line to enable
it. This is the recommended option for most Linux distro packagers since it's
working fine on all recent mainstream distros. It is automatically enabled on
Solaris 8 and above, as it's known to work.

If your system supports PCRE (Perl Compatible Regular Expressions), then you
really should build with libpcre which is between 2 and 10 times faster than
other libc implementations. Regex are used for header processing (deletion,
rewriting, allow, deny). Please see section 4 about dependencies to figure
how to build with PCRE support.

It is possible to add native support for SSL, by passing "USE_OPENSSL=1" on the
make command line. The libssl and libcrypto will automatically be linked with
HAProxy. Some systems also require libz, so if the build fails due to missing
symbols such as deflateInit(), then try again with "ADDLIB=-lz". Please check
section 4 about dependencies for more information on how to build with OpenSSL.

HAProxy can compress HTTP responses to save bandwidth. Please see section 4
about dependencies to see the available libraries and associated options.

By default, the DEBUG variable is set to '-g' to enable debug symbols. It is
not wise to disable it on uncommon systems, because it's often the only way to
get a usable core when you need one. Otherwise, you can set DEBUG to '-s' to
strip the binary.

If the ERR variable is set to any non-empty value, then -Werror will be added
to the compiler so that any build warning will trigger an error. This is the
recommended way to build when developing, and it is expected that contributed
patches were tested with ERR=1.

The SSL stack supports session cache synchronization between all running
processes. This involves some atomic operations and synchronization operations
which come in multiple flavors depending on the system and architecture :

  Atomic operations :
    - internal assembler versions for x86/x86_64 architectures

    - gcc builtins for other architectures. Some architectures might not
      be fully supported or might require a more recent version of gcc.
      If your architecture is not supported, you willy have to either use
      pthread if supported, or to disable the shared cache.

    - pthread (posix threads). Pthreads are very common but inter-process
      support is not that common, and some older operating systems did not
      report an error when enabling multi-process mode, so they used to
      silently fail, possibly causing crashes. Linux's implementation is
      fine. OpenBSD doesn't support them and doesn't build. FreeBSD 9 builds
      and reports an error at runtime, while certain older versions might
      silently fail. Pthreads are enabled using USE_PTHREAD_PSHARED=1.

  Synchronization operations :
    - internal spinlock : this mode is OS-independent, light but will not
      scale well to many processes. However, accesses to the session cache
      are rare enough that this mode could certainly always be used. This
      is the default mode.

    - Futexes, which are Linux-specific highly scalable light weight mutexes
      implemented in user-space with some limited assistance from the kernel.
      This is the default on Linux 2.6 and above and is enabled by passing
      USE_FUTEX=1

    - pthread (posix threads). See above.

If none of these mechanisms is supported by your platform, you may need to
build with USE_PRIVATE_CACHE=1 to totally disable SSL cache sharing. Then it
is better not to run SSL on multiple processes. Note that you don't need these
features if you only intend to use multi-threading and never multi-process.

If you need to pass other defines, includes, libraries, etc... then please
check the Makefile to see which ones will be available in your case, and
use/override the USE_* variables from the Makefile.

AIX 5.3 is known to work with the generic target. However, for the binary to
also run on 5.2 or earlier, you need to build with DEFINE="-D_MSGQSUPPORT",
otherwise __fd_select() will be used while not being present in the libc, but
this is easily addressed using the "aix52" target. If you get build errors
because of strange symbols or section mismatches, simply remove -g from
DEBUG_CFLAGS.

Building on AIX 7.2 works fine using the "aix72-gcc" TARGET. It adds two
special CFLAGS to prevent the loading of AIX's xmem.h and var.h. This is done
by defining the corresponding include-guards _H_XMEM and _H_VAR. Without
excluding those header-files the build fails because of redefinition errors.
Furthermore, the atomic library is added to the LDFLAGS to allow for
multithreading via USE_THREAD.

You can easily define your own target with the GNU Makefile. Unknown targets
are processed with no default option except USE_POLL=default. So you can very
well use that property to define your own set of options. USE_POLL and USE_SLZ
can even be disabled by setting them to an empty string. For example :

  $ gmake TARGET=tiny USE_POLL="" USE_SLZ="" TARGET_CFLAGS=-fomit-frame-pointer

If you need to pass some defines to the preprocessor or compiler, you may pass
them all in the DEFINE variable. Example:

  $ make TARGET=generic DEFINE="-DDEBUG_DONT_SHARE_POOLS -DDEBUG_MEMORY_POOLS"

The ADDINC variable may be used to add some extra include paths; this is
sometimes needed when cross-compiling. Similarly the ADDLIB variable may be
used to specifify extra paths to library files. Example :

  $ make TARGET=generic ADDINC=-I/opt/cross/include ADDLIB=-L/opt/cross/lib64


6) How to install HAProxy
=========================

To install haproxy, you can either copy the single resulting binary to the
place you want, or run :

    $ sudo make install

If you're packaging it for another system, you can specify its root directory
in the usual DESTDIR variable.

-- end