DOKK / manpages / debian 11 / openrc / openrc-run.8.en
openrc-run(8) System Manager's Manual (smm) openrc-run(8)

openrc-runa means of hooking shell commands into a service

openrc-run [-D, --nodeps] [-d, --debug] [-s, --ifstarted] [-S, --ifstopped] [-Z, --dry-run] [command ...]

openrc-run is basically an interpreter for shell scripts which provides an easy interface to the often complex system commands and daemons. When a service runs a command it first loads its multiplexed configuration file, then its master configuration file, then /etc/rc.conf and finally the script itself. At this point openrc-run then runs the command given.

Commands are defined as shell functions within the script. Here is a list of some functions that all scripts have by default:

describe
Describes what the service does and each command the service defines.
start
First we ensure that any services we depend on are started. If any needed services fail to start then we exit with a suitable error, otherwise call the supplied start function if it exists.
stop
First we ensure that any services that depend on us are stopped. If any services that need us fail to stop then we exit with a suitable error, otherwise call the supplied stop function if it exists.
restart
Stops and starts the service, including dependencies. This cannot be overridden. See the description of the RC_CMD variable below for the method to make your service behave differently when restart is being executed.
status
Shows the status of the service. The return code matches the status, with the exception of "started" returning 0 to match standard command behaviour.
zap
Resets the service state to stopped and removes all saved data about the service.

The following options affect how the service is run:

, --debug
Set xtrace on in the shell to assist in debugging.
, --nodeps
Ignore all dependency information the service supplies.
, --ifstarted
Only run the command if the service has been started.
, --ifstopped
Only run the command if the service has been stopped.
, --quiet
Turns off all informational output the service generates. Output from any non OpenRC commands is not affected.
, --verbose
Turns on any extra informational output the service generates.
, --dry-run
Shows which services would be stopped and/or started without actually stopping or starting them.

The following variables affect the service script:

extra_commands
Space separated list of extra commands the service defines. These should not depend on the service being stopped or started.
extra_started_commands
Space separated list of extra commands the service defines. These only work if the service has already been started.
extra_stopped_commands
Space separated list of extra commands the service defines. These only work if the service has already been stopped.
description
String describing the service.
description_$command
String describing the extra command.
supervisor
Supervisor to use to monitor this daemon. If this is unset or invalid, start-stop-daemon will be used. Currently, we support s6 from skarnet software, and supervise-daemon which is a light-weight supervisor internal to OpenRC. To use s6, set supervisor=s6. or set supervisor=supervise-daemon to use supervise-daemon. Note that supervise-daemon is still in early development, so it is considered experimental.
s6_service_path
The path to the s6 service directory if you are monitoring this service with S6. The default is /var/svc.d/${RC_SVCNAME}.
s6_svwait_options_start
The options to pass to s6-svwait when starting the service via s6.
s6_force_kill
Should we force-kill this service if s6_service_timeout_stop expires but the service doesn't go down during shutdown? The default is yes.
s6_service_timeout_stop
The amount of time, in milliseconds, s6-svc should wait for the service to go down when stopping the service. The default is 60000.
start_stop_daemon_args
List of arguments passed to start-stop-daemon when starting the daemon.
supervise_daemon_args
List of arguments passed to supervise-daemon when starting the daemon. If undefined, start_stop_daemon_args is used as a fallback.
command
Daemon to start or stop via start-stop-daemon or supervise-daemon if no start or stop function is defined by the service.
command_args
List of arguments to pass to the daemon when starting via start-stop-daemon.
command_args_background
This variable should be used if the daemon you are starting with start-stop-daemon(8) runs in the foreground by default but has its own command line options to request that it background and write a pid file. It should be set to those options. It should not be used at the same time as command_background, because command_background requests that start-stop-daemon(8) go into the background before executing the daemon.
command_args_foreground
List of arguments to pass to the daemon when starting via supervise-daemon. to force the daemon to stay in the foreground
command_background
Set this to "true", "yes" or "1" (case-insensitive) if you want start-stop-daemon(8) to force the daemon into the background. This forces the "--make-pidfile" and "--pidfile" options, so the pidfile variable must be set.
command_progress
Set this to "true", "yes" or "1" (case-insensitive) if you want start-stop-daemon(8) to display a progress meter when waiting for a daemon to stop.
command_user
If the daemon does not support changing to a different user id, you can use this to change the user id, and optionally group id, before start-stop-daemon(8) or supervise-daemon(8) launches the daemon.
output_log
This is the path to a file or named pipe where the standard output from the service will be redirected. If you are starting this service with start-stop-daemon(8), , you must set command_background to true. Keep in mind that this path will be inside the chroot if the chroot variable is set.
error_log
The same thing as output_log but for the standard error output.
output_logger
This is a process which will be used to log the standard output from the service. If you are starting this service with start-stop-daemon(8), , you must set command_background to true. Keep in mind that this command must be executable as a shell command inside the chroot if the chroot variable is set. Keep in mind also that this command works by accepting the stdout of the service on stdin. An example of a command that can be run this way is logger if you want your service output to go to syslog.
error_logger
The same thing as output_logger but for the standard error output.
directory
start-stop-daemon(8) and supervise-daemon(8) will chdir to this directory before starting the daemon.
chroot
start-stop-daemon(8) and supervise-daemon(8) will chroot into this path before writing the pid file or starting the daemon.
pidfile
Pidfile to use for the above defined command.
name
Display name used for the above defined command.
procname
Process name to match when signaling the daemon.
stopsig
Signal to send when stopping the daemon.
respawn_delay
Respawn delay supervise-daemon(8) will use for this daemon. See supervise-daemon(8) for more information about this setting.
respawn_max
Respawn max supervise-daemon(8) will use for this daemon. See supervise-daemon(8) for more information about this setting.
respawn_period
Respawn period supervise-daemon(8) will use for this daemon. See supervise-daemon(8) for more information about this setting.
retry
Retry schedule to use when stopping the daemon. It can either be a timeout in seconds or multiple signal/timeout pairs (like SIGTERM/5).
required_dirs
A list of directories which must exist for the service to start.
required_files
A list of files which must exist for the service to start.
start_inactive
Set to yes to have the service marked inactive when it starts. This is used along with in_background_fake to support re-entrant services.
in_background_fake
Space separated list of commands which should always succeed when in_background is yes.
umask
Set the umask of the daemon.

Keep in mind that eval is used to process chroot, command, command_args_*, command_user, pidfile and procname. This may affect how they are evaluated depending on how they are quoted.

You should define a depend function for the service so that openrc-run will start and stop it in the right order in relation to other services. As it's a function it can be very flexible, see the example below. Here is a list of the functions you can use in a depend function. You simply pass the names of the services you want to add to that dependency type to the function, or prefix the names with ! to remove them from the dependencies.

The service will attempt to start any services it needs regardless of whether they have been added to the runlevel. It will refuse to start until all services it needs have started, and it will refuse to stop until all services that need it have stopped.
The service will attempt to start any services it uses that have been added to the runlevel.
The service will attempt to start any services it wants, regardless of whether they have been added to the runlevel.
The service will start after these services and stop before these services.
The service will start before these services and stop after these services.
The service provides this virtual service. For example, named provides dns. Note that it is not legal to have a virtual and real service with the same name. If you do this, you will receive an error message, and you must rename either the real or virtual service.
We should recalculate our dependencies if the listed files have changed.
Tags a service with a keyword. These are the keywords we currently understand:
Don't stop this service when shutting the system down. This is normally quite safe as remaining daemons will be sent a SIGTERM just before final shutdown. Network related services such as the network and dhcpcd init scripts normally have this keyword.
Don't stop this service when changing runlevels, even if not present. This includes shutting the system down.
Other services should wait indefinitely for this service to start. Use this keyword if your service may take longer than 60 seconds to start.
When in a jail, exclude this service from any dependencies. The service can still be run directly. Set via rc_sys in /etc/rc.conf
Same as -jail, but for Linux Resource Containers (LXC).
Same as -jail, but for OpenVZ systems.
Same as -jail, but for Prefix systems.
Same as -jail, but for RKT systems.
Same as -jail, but for UML systems.
Same as -jail, but for VServer systems.
Same as -jail, but for Xen DOM0 systems.
Same as -jail, but for Xen DOMU systems.
Same as -jail, but for docker systems.
Same as -jail, but for all relevant container types on the operating system.

To see how to influence dependencies in configuration files, see the FILES section below.

Any command defined in extra_commands, extra_started_commands or extra_stopped_commands can have _pre and _post functions in the service script. If the command function is called foo, the_pre and _post functions for it should be called foo_pre and foo_post.

These functions should be used to perform preparation before the command is run and cleanup after the command completes. In order for openrc-run to record the command as being run successfully, the _pre function, command function itself and the _post function should all exit with a zero return code.

openrc-run defines some builtin functions that you can use inside your service scripts:

[string]
Output a green asterisk followed by the string.
[string]
Output a yellow asterisk followed by the string.
[string]
Output a red asterisk followed by the string to stderr.
[string]
Same as einfo, but append 3 dots to the end.
retval [string]
If retval does not equal 0 then output the string using eerror and !! in square brackets at the end of the line. Otherwise output ok in square brackets at the end of the line. The value of retval is returned.
retval [string]
Same as eend, but use ewarn instead of eerror.

You can prefix the above commands with the letter v, which means they only output when the environment variable EINFO_VERBOSE is true.

timeout file1 file2 ...
Wait for timeout seconds until all files exist. Returns 0 if all files exist, otherwise non zero. If timeout is less than 1 then we wait indefinitely.
file1 file2 ...
If file1 is newer than file2 return 0, otherwise 1. If file2 is a directory, then check all its contents too.
file1 file2 ...
If file1 is newer than file2 return 0, otherwise 1. If file2 is a directory, then check all its contents too.
name value
Saves the name value for later retrieval. Saved values are lost when the service stops.
name
Returns the saved value called name.
[service]
If the service is started, return 0 otherwise 1.
[service]
If the service is starting, return 0 otherwise 1.
[service]
If the service is inactive, return 0 otherwise 1.
[service]
If the service is stopping, return 0 otherwise 1.
[service]
If the service is stopped, return 0 otherwise 1.
[service]
If the service is coldplugged, return 0 otherwise 1.
[service]
If the service was inactive, return 0 otherwise 1.
[service] daemon [index]
If the service has started the daemon using start-stop-daemon, return 0 otherwise 1. If an index is specified, it has to be the nth daemon started by the service.
[service]
Mark the service as started.
[service]
Mark the service as starting.
[service]
Mark the service as inactive.
[service]
Mark the service as stopping.
[service]
Mark the service as stopped.
[service]
Mark the service as coldplugged.
[service]
Mark the service as inactive.
[-D, --directory-truncate] [-d, --directory] [-F, --file-truncate] [-f, --file] [-p, --pipe] [-m, --mode mode] [-o, --owner owner] [-W, --writable] [-q, --quiet] path ...
If -d, -f or -p is specified, checkpath checks to see if the path exists, is the right type and has the correct owner and access modes. If any of these tests fail, the path is created and set up as specified. If more than one of -d, -f or -p are specified, the last one will be used.

The argument to -m is a three or four digit octal number. If this option is not provided, the value defaults to 0644 for files and 0775 for directories.

The argument to -o is a representation of the user and/or group which should own the path. The user and group can be represented numerically or with names, and are separated by a colon.

The truncate options (-D and -F) cause the directory or file to be cleared of all contents.

If -W is specified, checkpath checks to see if the first path given on the command line is writable. This is different from how the test command in the shell works, because it also checks to make sure the file system is not read only.

Also, the -d, -f or -p options should not be specified along with this option.

The -q option suppresses all informational output. If it is specified twice, all error messages are suppressed as well.

[-M, --mount] [-R, --remount] [-b, --blockdevice] [-m, --mountargs] [-o, --options] [-p, --passno passno] [-t, --type fstype] path
If -b, -m, -o, -p or -t is specified,the appropriate information is extracted from fstab. If -M or -R are given, file systems are mounted or remounted.

The -q option suppresses all informational output. If it is specified twice, all error messages are suppressed as well.

[-f, --fstype-regex regex] [-F, --skip-fstype-regex regex] [-n, --node-regex regex] [-N, --skip-node-regex regex] [-o, --options-regex regex] [-O, --skip-options-regex regex] [-p, --point-regex regex] [-P, --skip-point-regex regex] [-e, --netdev] [-E, --nonetdev] [-i, --options] [-s, --fstype] [-t, --node] mount1 mount2 ...
The f, F, n, N, o, O, p, P, e and E options specify what you want to search for or skip in the mounted file systems. The i, s and t options specify what you want to display. If no mount points are given, all mount points will be considered.
value
If value matches YES, TRUE, ON or 1 regardless of case then we return 0, otherwise 1.

openrc-run sets the following environment variables for use in the service scripts:

RC_SVCNAME
Name of the service.
RC_SERVICE
Full path to the service.
RC_RUNLEVEL
Current runlevel that OpenRC is in. Note that, in OpenRC, the reboot runlevel is mapped to the shutdown runlevel. This was done because most services do not need to know if a system is shutting down or rebooting. If you are writing a service that does need to know this, see the RC_REBOOT variable.
RC_REBOOT
This variable contains YES if the system is rebooting. If your service needs to know the system is rebooting, you should test this variable.
RC_BOOTLEVEL
Boot runlevel chosen. Default is boot.
RC_DEFAULTLEVEL
Default runlevel chosen. Default is default.
RC_SYS
A special variable to describe the system more. Possible values are OPENVZ, XENU, XEN0, UML and VSERVER.
RC_PREFIX
In a Gentoo Prefix installation, this variable contains the prefix offset. Otherwise it is undefined.
RC_UNAME
The result of `uname -s`.
RC_CMD
This contains the name of the command the service script is executing, such as start, stop, restart etc. One example of using this is to make a service script behave differently when restart is being executed.
RC_GOINGDOWN
This variable contains YES if the system is going into single user mode or shutting down.
RC_LIBEXECDIR
The value of libexecdir which OpenRC was configured with during build time.
RC_NO_UMOUNTS
This variable is used by plugins to contain a list of directories which should not be unmounted.

Configuration files, relative to the location of the service. If a file ending with .${RC_RUNLEVEL} exists then we use that instead.

../conf.d/${RC_SVCNAME%%.*}
multiplexed configuration file. Example: if ${RC_SVCNAME} is net.eth1 then look for ../conf.d/net.
../conf.d/${RC_SVCNAME}
service configuration file.
/etc/rc.conf
host configuration file.

With the exception of /etc/rc.conf, the configuration files can also influence the dependencies of the service through variables. Simply prefix the name of the dependency with rc_. Examples:

# Whilst most services don't bind to a specific interface, our
# openvpn configuration requires a specific interface, namely bge0.
rc_need="net.bge0"
# To put it in /etc/rc.conf you would do it like this
rc_openvpn_need="net.bge0"

# Services should not depend on the tap1 interface for network,
# but we need to add net.tap1 to the default runlevel to start it.
rc_provide="!net"
# To put it in /etc/conf.d/net you would do it like this
rc_provide_tap1="!net"
# To put in in /etc/rc.conf you would do it like this
rc_net_tap1_provide="!net"

# It's also possible to negate keywords. This is mainly useful for prefix
# users testing OpenRC.
rc_keyword="!-prefix"
# This can also be used to block a script from running in all
# containers except one or two
rc_keyword="!-containers !-docker"

An example service script for foo.

#!/sbin/openrc-run
command=/usr/bin/foo
command_args="${foo_args} --bar"
pidfile=/var/run/foo.pid
name="FooBar Daemon"

description="FooBar is a daemon that eats and drinks"
extra_commands="show"
extra_started_commands="drink eat"
description_drink="Opens mouth and reflexively swallows"
description_eat="Chews food in mouth"
description_show="Shows what's in the tummy"

_need_dbus()
{
    grep -q dbus /etc/foo/plugins
}

depend()
{
    # We write a pidfile and to /var/cache, so we need localmount.
    need localmount
    # We can optionally use the network, but it's not essential.
    use net
    # We should be after bootmisc so that /var/run is cleaned before
    # we put our pidfile there.
    after bootmisc

    # Foo may use a dbus plugin.
    # However, if we add the dbus plugin whilst foo is running and
    # stop dbus, we don't need to stop foo as foo didn't use dbus.
    config /etc/foo/plugins
    local _need=
    if service_started; then
	_need=`service_get_value need`
    else
	if _need_dbus; then
	   _need="${_need} dbus"
	fi
    fi
    need ${_need}
}

# This function does any pre-start setup. If it fails, the service will
# not be started.
# If you need this function to behave differently for a restart command,
# you should check the value of RC_CMD for "restart".
# This also applies to start_post, stop_pre and stop_post.
start_pre()
{
	if [ "$RC_CMD" = restart ]; then
		# This block will only execute for a restart command. Use a
		# structure like this if you need special processing for a
		# restart which you do not need for a normal start.
		# The function can also fail from here, which will mean that a
		# restart can fail.
		# This logic can also be used in start_post, stop_pre and
		# stop_post.
	fi
    # Ensure that our dirs are correct
    checkpath --directory --owner foo:foo --mode 0775 \
	/var/run/foo /var/cache/foo
}

start_post()
{
    # Save our need
    if _need_dbus; then
	service_set_value need dbus
    fi
}

stop_post() {
    # Clean any spills
    rm -rf /var/cache/foo/*
}

drink()
{
    ebegin "Starting to drink"
    ${command} --drink beer
    eend $? "Failed to drink any beer :("
}

eat()
{
    local result=0 retval= ate= food=
    ebegin "Starting to eat"

    if yesno "${foo_diet}"; then
    	eend 1 "We are on a diet!"
	return 1
    fi

    for food in /usr/share/food/*; do
	veinfo "Eating `basename ${food}`"
	${command} --eat ${food}
	retval=$?
	: $(( result += retval ))
	[ ${retval} = 0 ] && ate="${ate} `basename ${food}`"
    done

    if eend ${result} "Failed to eat all the food"; then
	service_set_value ate "${ate}"
    fi
}

show()
{
    einfo "Foo has eaten: `service_get_value ate`"
}

Because of the way we load our configuration files and the need to handle more than one service directory, you can only use symlinks in service directories to other services in the same directory. You cannot symlink to a service in a different directory even if it is another service directory.

is_older_than should return 0 on success. Instead we return 1 to be compliant with Gentoo baselayout. Users are encouraged to use the is_newer_than function which returns correctly.

einfo(3), openrc(8), rc-status(8), rc-update(8), rc_plugin_hook(3), sh(1p), start-stop-daemon(8), supervise-daemon(8), uname(1)

Roy Marples <roy@marples.name>
William Hubbs <w.d.hubbs@gmail.com>

November 30, 2017 OpenRC