This section covers the basics of building, installing and setting up FRR.
The project publishes packages for Red Hat, Centos, Debian and Ubuntu on the GitHub releases. page. External contributors offer packages for many other platforms including *BSD, Alpine, Gentoo, Docker, and others. There is currently no documentation on how to use those but we hope to add it soon.
In addition to traditional packages the project also builds and publishes universal Snap images, available at https://snapcraft.io/frr.
Building FRR from source is the best way to ensure you have the latest features and bug fixes. Details for each supported platform, including dependency package listings, permissions, and other gotchas, are in the developer’s documentation. This section provides a brief overview on the process.
Getting the Source¶
FRR’s source is available on the project GitHub page.
git clone https://github.com/FRRouting/frr.git
When building from Git there are several branches to choose from. The
master branch is the primary development branch. It should be considered
unstable. Each release has its own branch named
is the release version.
In addition, release tarballs are published on the GitHub releases page here.
FRR has an excellent configure script which automatically detects most host configurations. There are several additional configure options to customize the build to include or exclude specific features and dependencies.
First, update the build system. Change into your FRR source directory and issue:
This will install any missing build scripts and update the Autotools configuration. Once this is done you can move on to choosing your configuration options from the list below.
Enable the alternate malloc library. In some cases this is faster and more efficient, in some cases it is not.
Do not build any documentation, including this one.
From the documentation build html docs as well in addition to the normal output.
Do not build zebra daemon. This generally only be useful in a scenario where you are building bgp as a standalone server.
Do not build ripd.
Do not build ripngd.
Do not build ospfd.
Do not build ospf6d.
Do not build bgpd.
Do not build ldpd.
Do not build nhrpd.
Do not build eigrpd.
Do not build babeld.
Do not build watchfrr. Watchfrr is used to integrate daemons into startup/shutdown software available on your machine. This is needed for systemd integration, if you disable watchfrr you cannot have any systemd integration.
Build with all warnings converted to errors as a compile option. This is recommended for developers only.
Turn off building of pimd. On some BSD platforms pimd will not build properly due to lack of kernel support.
Turn off building of vrrpd. Linux is required for vrrpd support; other platforms are not supported.
Turn off building of pbrd. This daemon currently requires linux in order to function properly.
Turn on building of sharpd. This daemon facilitates testing of FRR and can also be used as a quick and easy route generator.
Do not build staticd. This daemon is necessary if you want static routes.
Do not build bfdd.
Make bgpd which does not make bgp announcements at all. This feature is good for using bgpd as a BGP announcement listener.
Turn off bgpd’s ability to use VNC.
Turn off BGP BMP support
This option is deprecated as it is superseded by the -F (profile) command line option which allows adjusting the setting at startup rather than compile time.
Enable system defaults to work as if in a Data Center. See defaults.h for what is changed by this configure option.
Enable SNMP support. By default, SNMP support is disabled.
Disable support for OSPF-API, an API to interface directly with ospfd. OSPF-API is enabled if –enable-opaque-lsa is set.
Disable building of the example OSPF-API client.
Do not build isisd.
Do not build fabricd.
Enable IS-IS topology generator.
Enable the support of Linux Realms. Convert tag values from 1-255 into a realm value when inserting into the Linux kernel. Then routing policy can be assigned to the realm. See the tc man page.
Disable IRDP server support. This is enabled by default if we have both struct in_pktinfo and struct icmphdr available to us.
Disable support IPV6 router advertisement in zebra.
-rdynamicoption to the linker driver. This is in most cases necessary for getting usable backtraces. This option defaults to on if the compiler is detected as gcc, but giving an explicit enable/disable is suggested.
Controls backtrace support for the crash handlers. This is autodetected by default. Using the switch will enforce the requested behaviour, failing with an error if support is requested but not available. On BSD systems, this needs libexecinfo, while on glibc support for this is part of libc itself.
Turn on some options for compiling FRR within a development environment in mind. Specifically turn on -g3 -O0 for compiling options and add inclusion of grammar sandbox.
Build without SNMP support.
Build without VTYSH.
Build with FPM module support.
Alpine Linux does not allow non-numeric characters in the version string. With this option, we provide a way to strip out these characters for APK dev package builds.
Remove the “configuerd with” field that has all of the build configuration arguments when reporting the version string in show version command.
Add git information to MOTD and build version string
Compile FRR with up to X way ECMP supported. This number can be from 0-999. For backwards compatibility with older configure options when setting X = 0, we will build FRR with 64 way ECMP. This is needed because there are hardcoded arrays that FRR builds towards, so we need to know how big to make these arrays at build time. Additionally if this parameter is not passed in FRR will default to 16 ECMP.
Turn on the ability of FRR to access some shell options( telnet/ssh/bash/etc. ) from vtysh itself. This option is considered extremely unsecure and should only be considered for usage if you really really know what you are doing.
Code coverage reports from gcov require adjustments to the C and LD flags. With this option, gcov instrumentation is added to the build and coverage reports are created during execution. The check-coverage make target is also created to ease report uploading to codecov.io. The upload requires the COMMIT (git hash) and TOKEN (codecov upload token) environment variables be set.
Build with configuration rollback support. Requires SQLite3.
Build the ConfD northbound plugin. Look for the libconfd libs and headers in dir.
Build the Sysrepo northbound plugin.
Enable the gRPC northbound plugin.
Enable the ZeroMQ handler.
Use libpam for PAM support in vtysh.
- --enable-time-check XXX¶
This option is deprecated as it was replaced by the
service cputime-statsCLI command, which may be adjusted at runtime rather than being a compile-time setting. See there for further detail.
This option is deprecated as it was replaced by the
service cputime-warning NNNCLI command, which may be adjusted at runtime rather than being a compile-time setting. See there for further detail.
Turn on the usage of PCRE Posix libs for regex functionality.
Set hardcoded rpaths in the executable [default=yes].
Enable Lua scripting [default=no].
You may specify any combination of the above options to the configure
script. By default, the executables are placed in
and the configuration files in
installation prefix and other directories may be changed using the following
options to the configuration script.
- --prefix <prefix>¶
Install architecture-independent files in prefix [/usr/local].
- --sysconfdir <dir>¶
Look for configuration files in dir [prefix/etc]. Note that sample configuration files will be installed here.
- --localstatedir <dir>¶
Configure zebra to use dir for local state files, such as pid files and unix sockets.
- --with-scriptdir <dir>¶
Look for Lua scripts in
- --with-yangmodelsdir <dir>¶
Look for YANG modules in dir [prefix/share/yang]. Note that the FRR YANG modules will be installed here.
- --with-vici-socket <path>¶
Set StrongSWAN vici interface socket path [/var/run/charon.vici].
--enable-systemd option does not exist anymore. Support for
systemd is now always available through built-in functions, without
depending on libsystemd.
Python dependency, documentation and tests¶
FRR’s documentation and basic unit tests heavily use code written in Python. Additionally, FRR ships Python extensions written in C which are used during its build process.
To this extent, FRR needs the following:
an installation of CPython, preferably version 3.2 or newer (2.7 works but is end of life and will stop working at some point.)
development files (mostly headers) for that version of CPython
an installation of sphinx for that version of CPython, to build the documentation
an installation of pytest for that version of CPython, to run the unit tests
The sphinx and pytest dependencies can be avoided by not building
documentation / not running
make check, but the CPython dependency is a
hard dependency of the FRR build process (for the clippy tool.)
Additionally, you may configure zebra to drop its elevated privileges shortly after startup and switch to another user. The configure script will automatically try to configure this support. There are three configure options to control the behaviour of FRR daemons.
- --enable-user <user>¶
Switch to user user shortly after startup, and run as user `user in normal operation.
- --enable-group <user>¶
Switch real and effective group to group shortly after startup.
- --enable-vty-group <group>¶
Create Unix Vty sockets (for use with vtysh) with group ownership set to group. This allows one to create a separate group which is restricted to accessing only the vty sockets, hence allowing one to delegate this group to individual users, or to run vtysh setgid to this group.
The default user and group which will be configured is ‘frr’ if no user or
group is specified. Note that this user or group requires write access to the
local state directory (see
--localstatedir) and requires at least
read access, and write access if you wish to allow daemons to write out their
configuration, to the configuration directory (see
On systems which have the ‘libcap’ capabilities manipulation library (currently only Linux), FRR will retain only minimal capabilities required and will only raise these capabilities for brief periods. On systems without libcap, FRR will run as the user specified and only raise its UID to 0 for brief periods.
There are several options available only to GNU/Linux systems. If you use GNU/Linux, make sure that the current kernel configuration is what you want. FRR will run with any kernel configuration but some recommendations do exist.
Kernel/User Netlink socket. This enables an advanced interface between the Linux kernel and zebra (Kernel Interface).
This makes it possible to receive Netlink routing messages. If you specify this option, zebra can detect routing information updates directly from the kernel (Kernel Interface).
This option enables IP multicast and should be specified when you use ripd (RIP) or ospfd (OSPFv2) because these protocols use multicast.
Linux sysctl settings and kernel modules¶
There are several kernel parameters that impact overall operation of FRR when
using Linux as a router. Generally these parameters should be set in a
sysctl related configuration file, e.g.,
Ubuntu based systems and a new file
/etc/sysctl.d/90-routing-sysctl.conf on Centos based systems.
Additional kernel modules are also needed to support MPLS forwarding.
- IPv4 and IPv6 forwarding
The following are set to enable IP forwarding in the kernel:
- MPLS forwarding
Basic MPLS support was introduced in the kernel in version 4.1 and additional capability was introduced in 4.3 and 4.5. For some general information on Linux MPLS support, see https://www.netdevconf.org/1.1/proceedings/slides/prabhu-mpls-tutorial.pdf. The following modules should be loaded to support MPLS forwarding, and are generally added to a configuration file such as
# Load MPLS Kernel Modules mpls_router mpls_iptunnel
The following is an example to enable MPLS forwarding in the kernel, typically by editing
# Enable MPLS Label processing on all interfaces net.mpls.conf.eth0.input=1 net.mpls.conf.eth1.input=1 net.mpls.conf.eth2.input=1 net.mpls.platform_labels=100000
Make sure to add a line equal to
net.mpls.conf.<if>.inputfor each interface ‘<if>’ used with MPLS and to set labels to an appropriate value.
- VRF forwarding
General information on Linux VRF support can be found in https://www.kernel.org/doc/Documentation/networking/vrf.txt.
Kernel support for VRFs was introduced in 4.3, but there are known issues in versions up to 4.15 (for IPv4) and 5.0 (for IPv6). The FRR CI system doesn’t perform VRF tests on older kernel versions, and VRFs may not work on them. If you experience issues with VRF support, you should upgrade your kernel version.
Once you have chosen your configure options, run the configure script and pass the options you chose:
./configure \ --prefix=/usr \ --localstatedir=/var/run/frr \ --sbindir=/usr/lib/frr \ --sysconfdir=/etc/frr \ --enable-pimd \ --enable-watchfrr \ ...
After configuring the software, you are ready to build and install it in your system.
make && sudo make install
If everything finishes successfully, FRR should be installed. You should now skip to the section on Basic Setup.