SYSTEMD.SERVICE(5)

💖 systemd.service — Service unit configuration 💖

ok bestie, this is the systemd.service manual page from the systemd.service family, and it's literally the blueprint for every .service file on your machine. grab a snack, we're going in 🫶

---

✨ NAME ✨

systemd.service — service unit configuration, aka the file format that tells systemd how to run and babysit a process 💅

---

🦋 SYNOPSIS 🦋

service.service

---

🌸 DESCRIPTION 🌸

So. A unit config file whose name ends in .service encodes info about a process that systemd controls and supervises. That's it, that's the whole vibe. ✨

This man page lists the config options that are specific to this unit type. For the common options that all unit config files share, go read systemd.unit(5). The common stuff lives in the generic [Unit] and [Install] sections. The service-specific stuff lives in the [Service] section. 🎀

There are extra options hanging out in other pages too:

• systemd.exec(5) — defines the execution environment the commands run in
• systemd.kill(5) — defines the way the service's processes get terminated
• systemd.resource-control(5) — configures resource control settings for the service's processes

Also, the systemd-run(1) command lets you create .service and .scope units dynamically and transiently, straight from the command line. no cap, that's a whole superpower 🔥

---

🧚 SERVICE TEMPLATES 🧚

okay so systemd services can take a single argument using the service@argument.service syntax. These are called "instantiated" services, and the unit definition without the argument parameter is called a "template". 🌈

A classic example: a dhcpcd@.service service template that takes a network interface as a parameter, which forms an instantiated service when you fill it in. Inside the service file, this parameter (aka the "instance name") can be accessed with %-specifiers. Details are in systemd.unit(5). 👑

---

💜 AUTOMATIC DEPENDENCIES 💜

⭐ Implicit Dependencies

These dependencies get added implicitly, like systemd doing your homework for you (but in a helpful way, not an academic-integrity way):

• Services with Type=dbus set automatically acquire dependencies of type Requires= and After= on dbus.socket.
• Socket activated services are automatically ordered after their activating .socket units via an automatic After= dependency. Services also pull in all .socket units listed in Sockets= via automatic Wants= and After= dependencies.

Additional implicit dependencies may be added as a result of execution and resource control parameters, as documented in systemd.exec(5) and systemd.resource-control(5). 📚

🌟 Default Dependencies

These dependencies get added unless DefaultDependencies=no is set:

• Service units will have dependencies of type Requires= and After= on sysinit.target, a dependency of type After= on basic.target, and dependencies of type Conflicts= and Before= on shutdown.target. These ensure that normal service units pull in basic system initialization, and are terminated cleanly prior to system shutdown. Only services involved with early boot or late system shutdown should disable this option. 🚨
• Instanced service units (i.e. service units with an "@" in their name) are assigned by default a per-template slice unit (see systemd.slice(5)), named after the template unit, containing all instances of that specific template. This slice is normally stopped at shutdown, together with all template instances. If that's not what you want, set DefaultDependencies=no in the template unit, and either define your own per-template slice unit file that also sets DefaultDependencies=no, or set Slice=system.slice (or another suitable slice) in the template unit. Also see systemd.resource-control(5).

---

💅 OPTIONS 💅

Service unit files may include [Unit] and [Install] sections, which are described in systemd.unit(5).

Service unit files must include a [Service] section, which carries info about the service and the process it supervises. A bunch of options that may be used in this section are shared with other unit types — those live in systemd.exec(5), systemd.kill(5) and systemd.resource-control(5). The options specific to the [Service] section of service units are the following:

• Type= — Configures the mechanism via which the service notifies the manager that the service start-up has finished. One of simple, exec, forking, oneshot, dbus, notify, notify-reload, or idle:
  • If set to simple (the default if ExecStart= is specified but neither Type= nor BusName= are, and credentials are not used), the service manager will consider the unit started immediately after the main service process has been forked off (i.e. immediately after fork(), and before various process attributes have been configured and in particular before the new process has called execve() to invoke the actual service binary). Typically, Type=exec is the better choice, see below.

It is expected that the process configured with ExecStart= is the main process of the service. In this mode, if the process offers functionality to other processes on the system, its communication channels should be installed before the service is started up (e.g. sockets set up by systemd, via socket activation), as the service manager will immediately proceed starting follow-up units, right after creating the main service process, and before executing the service's binary. Note that this means systemctl start command lines for simple services will report success even if the service's binary cannot be invoked successfully (for example because the selected User= does not exist, or the service binary is missing). 😭

• The exec type is similar to simple, but the service manager will consider the unit started immediately after the main service binary has been executed. The service manager will delay starting of follow-up units until that point. (Or in other words: simple proceeds with further jobs right after fork() returns, while exec will not proceed before both fork() and execve() in the service process succeeded.) Note that this means systemctl start command lines for exec services will report failure when the service's binary cannot be invoked successfully (for example because the selected User= does not exist, or the service binary is missing). This type is implied if credentials are used (refer to LoadCredential= in systemd.exec(5) for details).
• If set to forking, the manager will consider the unit started immediately after the binary that forked off by the manager exits. The use of this type is discouraged, use notify, notify-reload, or dbus instead. 🙈

It is expected that the process configured with ExecStart= will call fork() as part of its start-up. The parent process is expected to exit when start-up is complete and all communication channels are set up. The child continues to run as the main service process, and the service manager will consider the unit started when the parent process exits. This is the behavior of traditional UNIX services. If this setting is used, it is recommended to also use the PIDFile= option, so that systemd can reliably identify the main process of the service. The manager will proceed with starting follow-up units after the parent process exits.

• Behavior of oneshot is similar to exec; however, the service manager will consider the unit up after the main process exits. It will then start follow-up units. RemainAfterExit= is particularly useful for this type of service. Type=oneshot is the implied default if neither Type= nor ExecStart= are specified. Note that if this option is used without RemainAfterExit= the service will never enter "active" unit state, but will directly transition from "activating" to "deactivating" or "dead", since no process is configured that shall run continuously. In particular this means that after a service of this type ran (and which has RemainAfterExit= not set) it will not show up as started afterwards, but as dead. 💀
• Behavior of dbus is similar to simple; however, units of this type must have the BusName= specified and the service manager will consider the unit up when the specified bus name has been acquired. This type is the default if BusName= is specified.

Service units with this option configured implicitly gain dependencies on the dbus.socket unit. A service unit of this type is considered to be in the activating state until the specified bus name is acquired. It is considered activated while the bus name is taken. Once the bus name is released the service is considered being no longer functional, which has the effect that the service manager attempts to terminate any remaining processes belonging to the service. Services that drop their bus name as part of their shutdown logic thus should be prepared to receive a SIGTERM (or whichever signal is configured in KillSignal=) as result. ⚠️

• Behavior of notify is similar to exec; however, it is expected that the service sends a "READY=1" notification message via sd_notify(3) or an equivalent call when it has finished starting up. systemd will proceed with starting follow-up units after this notification message has been sent. If this option is used, NotifyAccess= (see below) should be set to open access to the notification socket provided by systemd. If NotifyAccess= is missing or set to none, it will be forcibly set to main.

If the service supports reloading, and uses a signal to start the reload, using notify-reload instead is recommended.

• Behavior of notify-reload is similar to notify, with one difference: the SIGHUP UNIX process signal is sent to the service's main process when the service is asked to reload and the manager will wait for a notification about the reload being finished.

When initiating the reload process the service is expected to reply with a notification message via sd_notify(3) that contains the "RELOADING=1" field in combination with "MONOTONIC_USEC=" set to the current monotonic time (i.e. CLOCK_MONOTONIC in clock_gettime(2)) in μs, formatted as decimal string. Once reloading is complete another notification message must be sent, containing "READY=1". Using this service type and implementing this reload protocol is an efficient alternative to providing an ExecReload= command for reloading of the service's configuration. 🎀

The signal to send can be tweaked via ReloadSignal=, see below.

okay so for notify-reload services, systemd is not just vibing 💅 — it actually verifies that the service's main process really installed a handler for the configured reload signal (default is SIGHUP). It checks for both traditional signal handlers (via sigaction(2)) and blocked signals that are typically handled via signalfd(2). And bestie, it does this at two moments:

1. When the service first sends "READY=1" during initial startup. If no signal handler is detected at this point, the service startup is aborted 😭
2. Immediately before sending the reload signal. If the handler is no longer present, systemd logs a warning but the reload signal is still sent.

This prevents accidentally starting services that lack the required signal handler. So if you're on notify-reload, you must install your reload signal handler before you send the initial "READY=1" notification. periodt.

• idle — behavior is very similar to simple, but here's the tea: actual execution of the service program is delayed until all active jobs are dispatched. It's cute for avoiding the interleaving of output of shell services with the status output on the console 🎀 But ngl, this type is useful only to improve console output — it is not useful as a general unit ordering tool, and the effect of this service type is subject to a 5s timeout, after which the service program is invoked anyway. 💀

Now the recommendation, and this is important: use Type=exec for long-running services, because it ensures that process setup errors (e.g. errors such as a missing service executable, or missing user) are properly tracked. iconic. However — this service type will not propagate the failures in the service's own startup code (as opposed to failures in the preparatory steps the service manager executes before execve()), and it does not allow ordering of other units against completion of initialization of the service code itself. Which, for the record, is useful if clients need to connect to the service through some form of IPC, and the IPC channel is only established by the service itself — in contrast to doing this ahead of time through socket or bus activation or similar. So yeah, it might not be sufficient for many cases 🙈

If that's you, then notify, notify-reload, or dbus (the last one only if the service provides a D-Bus interface) are the preferred options, because they let the service program code precisely schedule when to consider the service started up successfully and when to proceed with follow-up units ✨ The notify/notify-reload service types require explicit support in the service codebase — the service has to invoke sd_notify() or an equivalent API at the appropriate time. If that's not supported, then forking is an alternative: it supports the traditional heavy-weight UNIX service start-up protocol. Just know that using any type other than simple can possibly delay the boot process, since the service manager needs to wait for at least some service initialization to complete. (Also note it is generally not recommended to use idle or oneshot for long-running services.) 🌸

One more heads up: various service settings (e.g. User=, Group= through libc NSS) might result in "hidden" blocking IPC calls to other services when used. Sometimes it might be advisable to use the simple service type to ensure that the service manager's transaction logic is not affected by such potentially slow operations and hidden dependencies, since this is the only service type where the service manager will not wait for such service execution setup operations to complete before proceeding. 👑

---

🌸 ExitType= 🌸

Specifies when the manager should consider the service to be finished. One of main or cgroup:

• If set to main (the default), the service manager will consider the unit stopped when the main process — which is determined according to the Type= — exits.
• If set to cgroup, the service will be considered running as long as at least one process in the cgroup has not exited. This option cannot be used with Type=oneshot.

It's generally recommended to use ExitType=main when a service has a known forking model and a main process can reliably be determined. ExitType=cgroup is meant for applications whose forking model is not known ahead of time and which might not have a specific main process. It's well suited for transient or automatically generated services, such as graphical applications inside of a desktop environment. 🦋

Added in version 250.

💜 RemainAfterExit= 💜

Takes a boolean value that specifies whether the service shall be considered active even when all its processes exited. Defaults to no.

✨ GuessMainPID= ✨

Takes a boolean value that specifies whether systemd should try to guess the main PID of a service if it cannot be determined reliably. Heads up: this option is ignored unless Type=forking is set and PIDFile= is unset, because for the other types or with an explicitly configured PID file, the main PID is always known. The guessing algorithm might come to incorrect conclusions if a daemon consists of more than one process 😬 And if the main PID cannot be determined, failure detection and automatic restarting of a service will not work reliably. Defaults to yes.

🎀 PIDFile= 🎀

Takes a path referring to the PID file of the service. Usage of this option is recommended for services where Type= is set to forking. The path specified typically points to a file below /run/. If a relative path is specified for system service, then it is hence prefixed with /run/, and prefixed with $XDG_RUNTIME_DIR if specified in a user service. The service manager will read the PID of the main process of the service from this file after start-up of the service. The service manager will not write to the file configured here, although it will remove the file after the service has shut down if it still exists. The PID file does not need to be owned by a privileged user, but if it is owned by an unprivileged user, additional safety restrictions are enforced: the file may not be a symlink to a file owned by a different user (neither directly nor indirectly), and the PID file must refer to a process already belonging to the service. 🔒

Note that PID files should be avoided in modern projects. Use Type=notify, Type=notify-reload or Type=simple where possible, which does not require use of PID files to determine the main process of a service and avoids needless forking. 💅

🌈 BusName= 🌈

Takes a D-Bus destination name that this service shall use. This option is mandatory for services where Type= is set to dbus. It's recommended to always set this property if known, to make it easy to map the service name to the D-Bus destination. In particular, systemctl service-log-level/service-log-target verbs make use of this. ⭐

🔥 ExecStart= 🔥

Commands that are executed when this service is started.

Unless Type= is oneshot, exactly one command must be given. When Type=oneshot is used, this setting may be used multiple times to define multiple commands to execute. If the empty string is assigned to this option, the list of commands to start is reset, and prior assignments of this option will have no effect. If no ExecStart= is specified, then the service must have RemainAfterExit=yes and at least one ExecStop= line set. (Services lacking both ExecStart= and ExecStop= are not valid.) 💀

If more than one command is configured, the commands are invoked sequentially in the order they appear in the unit file. If one of the commands fails (and is not prefixed with -), other lines are not executed, and the unit is considered failed.

Unless Type=forking is set, the process started via this command line will be considered the main process of the daemon.

🌸 ExecStartPre=, ExecStartPost= 🌸

Additional commands that are executed before or after the command in ExecStart=, respectively. Syntax is the same as for ExecStart=. Multiple command lines are allowed, regardless of the service type (i.e. Type=), and the commands are executed one after the other, serially.

If any of those commands (not prefixed with -) fail, the rest are not executed and the unit is considered failed. 😭

ExecStart= commands are only run after all ExecStartPre= commands that were not prefixed with a - exit successfully.

ExecStartPost= commands are only run after the commands specified in ExecStart= have been invoked successfully, as determined by Type= (i.e. the process has been started for Type=simple or Type=idle, the last ExecStart= process exited successfully for Type=oneshot, the initial process exited successfully for Type=forking, "READY=1" is sent for Type=notify/Type=notify-reload, or the BusName= has been taken for Type=dbus). ✨

Note that ExecStartPre= may not be used to start long-running processes. All processes forked off by processes invoked via ExecStartPre= will be killed before the next service process is run.

Note that if any of the commands specified in ExecStartPre=, ExecStart=, or ExecStartPost= fail (and are not prefixed with -, see above) or time out before the service is fully up, execution continues with commands specified in ExecStopPost=, and the commands in ExecStop= are skipped.

Note that the execution of ExecStartPost= is taken into account for the purpose of Before=/After= ordering constraints.

🦋 ExecCondition= 🦋

Optional commands that are executed before the commands in ExecStartPre=. Syntax is the same as for ExecStart=. Multiple command lines are allowed, regardless of the service type (i.e. Type=), and the commands are executed one after the other, serially.

The behavior is like an ExecStartPre= and condition check hybrid: when an ExecCondition= command exits with exit code 1 through 254 (inclusive), the remaining commands are skipped and the unit is not marked as failed. However, if an ExecCondition= command exits with 255 or abnormally (e.g. timeout, killed by a signal, etc.), the unit will be considered failed (and remaining commands will be skipped). Exit code of 0 or those matching SuccessExitStatus= will continue execution to the next commands. 👑

bestie we're still in the [Service] section and there's so much left. strap in. 💅

---

okay so the tea on ExecCondition= first: a skip there is NOT the same as a unit-level Condition...= or Assert...= check failing. why? because ExecCondition= runs as part of the activation transition, so a skip makes the unit slide from active → inactive, and consequently SuccessAction= (peep systemd.unit(5)) will be honored. meanwhile, Condition...= directives in the "[Unit]" section block activation entirely, which means they never trigger SuccessAction=. two different vibes, fr fr. 🌸

and the same whole recommendation about not running long-running processes in ExecStartPre= also applies to ExecCondition=. also, ExecCondition= will run the commands in ExecStopPost=, as part of stopping the service, whenever there's a non-zero or abnormal exit, like the ones described above.

Added in version 243.

---

ExecReload= 🔄

Commands to execute to trigger a configuration reload in the service. this setting may take multiple command lines, following the same scheme as described for ExecStart= above. using it is optional. specifier and environment variable substitution is supported here, same scheme as for ExecStart=. ✨

one additional, special environment variable gets set: if it's known, $MAINPID is set to the main process of the daemon, and you can use it for command lines like this:

ExecReload=kill -HUP $MAINPID

but bestie. listen. reloading a daemon by enqueuing a signal without completion notification (which is exactly what that example line above does) is usually not a good choice, because it's an asynchronous operation and therefore not suitable when you're ordering reloads of multiple services against each other. 😭

so it is thus strongly recommended to either use Type=notify-reload, or to set ExecReload= to a command that not only triggers a configuration reload of the daemon, but also synchronously waits for it to complete. for example, dbus-broker(1) uses this:

ExecReload=busctl call org.freedesktop.DBus \
        /org/freedesktop/DBus org.freedesktop.DBus \
        ReloadConfig

this setting can be combined with Type=notify-reload, in which case the service main process is signaled after all specified command lines finish execution. specially, if a "RELOADING=1" notification is received before ExecReload= completes, the signaling is skipped and the service manager immediately starts listening for "READY=1". iconic behavior tbh 👑

---

ExecReloadPost= 🎀

Commands to execute after a successful reload operation. syntax for this setting is exactly the same as ExecReload=. no notes, slay.

Added in version 259.

---

ExecStop= 🛑

Commands to execute to stop the service that was started via ExecStart=. this argument takes multiple command lines, following the same scheme as described for ExecStart= above. using it is optional.

after the commands configured in this option are run, it is implied that the service is stopped, and any processes remaining for it are terminated according to the KillMode= setting (see systemd.kill(5)). if this option is not specified, the process is terminated by sending the signal specified in KillSignal= or RestartKillSignal= when a service stop is requested. specifier and environment variable substitution is supported (including $MAINPID, see above). 💖

warning bestie: it is usually not sufficient to specify a command here that only asks the service to terminate (like, sending some form of termination signal to it) but doesn't wait for it to actually do so. because the remaining processes of the service are killed according to KillMode= and KillSignal= or RestartKillSignal= as described above immediately after the command exited, this may not result in a clean stop. so the command you specify should be a synchronous operation, not an asynchronous one. periodt.

also note that the commands specified in ExecStop= are only executed when the service started successfully first. they are not invoked if the service was never started at all, or if its start-up failed — for example because any of the commands specified in ExecStart=, ExecStartPre= or ExecStartPost= failed (and weren't prefixed with "-", see above) or timed out. use ExecStopPost= to invoke commands when a service failed to start up correctly and is shut down again. 🙈

and another thing: the stop operation is always performed if the service started successfully, even if the processes in the service terminated on their own or were killed. so your stop commands must be prepared to deal with that case. $MAINPID will be unset if systemd knows that the main process exited by the time the stop commands are called. 😭

service restart requests are implemented as a stop operation followed by a start operation. which means ExecStop= and ExecStopPost= are executed during a service restart operation. iykyk.

it's recommended to use this setting for commands that communicate with the service requesting clean termination. for post-mortem clean-up steps, use ExecStopPost= instead. 🧹

---

ExecStopPost= 🧼

Additional commands that are executed after the service is stopped. this includes cases where the commands configured in ExecStop= were used, where the service doesn't have any ExecStop= defined at all, or where the service exited unexpectedly. this argument takes multiple command lines, following the same scheme as described for ExecStart=. using these settings is optional. specifier and environment variable substitution is supported.

and note that — unlike ExecStop= — commands specified with this setting are invoked when a service failed to start up correctly and is shut down again. that's the whole point bestie 💅

it's recommended to use this setting for clean-up operations that shall be executed even when the service failed to start up correctly. commands configured here need to be able to operate even if the service failed starting up half-way and left incompletely initialized data lying around. and since the service's processes have likely exited already by the time these commands run, they should not attempt to communicate with them. they're gone, babe.

note that all commands configured with this setting are invoked with the result code of the service, as well as the main process' exit code and status, set in the $SERVICE_RESULT, $EXIT_CODE and $EXIT_STATUS environment variables — peep systemd.exec(5) for the details. 🔍

also: the execution of ExecStopPost= is taken into account for the purpose of Before=/After= ordering constraints. don't forget that part.

---

RestartSec= ⏱️

Configures the time to sleep before restarting a service (as configured with Restart=). takes a unit-less value in seconds, or a time span value such as "5min 20s". defaults to 100ms. cute little nap before the comeback ✨

---

RestartSteps= 📈

Configures the number of exponential steps to take to increase the interval of auto-restarts from RestartSec= to RestartMaxDelaySec=. takes a positive integer or 0 to disable it. defaults to 0.

hint: values between 3 and 5 are good choices when exponential backoff is desired. 👑

example:

RestartSec=10s
RestartSteps=4
RestartMaxDelaySec=160s

this will produce the following restart intervals: 10s, 20s, 40s, 80s, 160s, 160s, 160s, etc. notice the geometric interpolation and the resulting constant ratio between intervals; here it is 2. the formula for the ratio is (RestartMaxDelaySec / RestartSec)^(1 / RestartSteps). a (repeating) delay equal to RestartMaxDelaySec= is always reached after RestartSteps + 1 steps. math ate, no cap 🧮

this setting is effective only if RestartMaxDelaySec= is also set and RestartSec= is not zero.

Added in version 254.

---

RestartMaxDelaySec= 🛑

Configures the longest time to sleep before restarting a service as the interval goes up with RestartSteps=. takes a value in the same format as RestartSec=, or "infinity" to disable the setting. defaults to "infinity".

this setting is effective only if RestartSteps= is also set and RestartSec= is not zero.

Added in version 254.

---

RestartRandomizedDelaySec= 🎲

Delay automatic restarts by a randomly selected, evenly distributed amount of time between 0 and the specified time value, added on top of the delay otherwise configured via RestartSec= (and RestartSteps=/RestartMaxDelaySec=, if used). takes a value in the same format as RestartSec=. defaults to 0, which means no randomized delay shall be applied.

this setting is useful to stretch out the restarts of similarly configured service instances that fail at the same time, so they don't all restart simultaneously and possibly cause resource congestion. it's the restart-side analogue of RandomizedDelaySec= in systemd.timer(5). 💜

Added in version 262.

---

TimeoutStartSec= ⏳

Configures the time to wait for start-up. if a daemon service does not signal start-up completion within the configured time, the service will be considered failed and will be shut down again. the precise action depends on the TimeoutStartFailureMode= option. takes a unit-less value in seconds, or a time span value such as "5min 20s". pass "infinity" to disable the timeout logic entirely. defaults to DefaultTimeoutStartSec= set in the manager — except when Type=oneshot is used, in which case the timeout is disabled by default (see systemd-system.conf(5)). 🌸

if a service of Type=notify/Type=notify-reload sends "EXTEND_TIMEOUT_USEC=...", this may cause the start time to be extended beyond TimeoutStartSec=. the first receipt of this message must occur before TimeoutStartSec= is exceeded, and once the start time has extended beyond TimeoutStartSec=, the service manager will allow the service to continue to start — provided the service repeats "EXTEND_TIMEOUT_USEC=..." within the interval specified until the service startup status is finished by "READY=1". (see sd_notify(3)). 🦋

note that the start timeout is also applied to service reloads, regardless of whether they're implemented through ExecReload= or via the reload logic enabled via Type=notify-reload. if the reload does not complete within the configured time, the reload will be considered failed and the service will continue running with the old configuration. this will not affect the running service, but it will be logged and will cause e.g. systemctl reload to fail. sad girl hours 😭

Added in version 188.

---

• TimeoutStopSec= — bestie, this one is doing double duty and honestly? multitasking queen 💅. Two whole jobs:
  1. It sets how long we wait for each ExecStop= command to do its thing. If any one of them times out, all the later ExecStop= commands get skipped and the service gets yeeted with SIGTERM. And if there are no ExecStop= commands at all? SIGTERM shows up immediately. (You can change this default behaviour with TimeoutStopFailureMode=.)
  2. It also sets how long we wait for the service itself to actually stop. If it doesn't wrap up in that window, it gets forcibly terminated by SIGKILL — yes, the unignorable one (peep KillMode= in systemd.kill(5)).

Takes a unit-less value in seconds, or a time span value like "5min 20s". Pass "infinity" to disable the timeout logic entirely. Defaults to DefaultTimeoutStopSec= from the manager configuration file (see systemd-system.conf(5)).

Oh and — if a service of Type=notify / Type=notify-reload sends "EXTEND_TIMEOUT_USEC=...", it can push the stop time past TimeoutStopSec=. The first receipt of that message has to land before TimeoutStopSec= is exceeded, and once the stop time has extended beyond TimeoutStopSec=, the service manager will let the service keep stopping, as long as it keeps repeating "EXTEND_TIMEOUT_USEC=..." within the interval specified, or just terminates itself (see sd_notify(3)).

Added in version 188.

• TimeoutAbortSec= — this one configures how long we wait for the service to terminate when it got aborted because of a watchdog timeout (see WatchdogSec=). If the service has a short TimeoutStopSec=, this option is your girl for giving the system more time to write a core dump of the service 🫶. When it expires, the service gets forcibly terminated by SIGKILL (see KillMode= in systemd.kill(5)). And the core file will be truncated in that case — sad but true. So use TimeoutAbortSec= to set a sensible per-service core-dumping timeout: big enough to write all the expected data, short enough to still handle the service failure in due time.

Takes a unit-less value in seconds, or a time span value such as "5min 20s". Pass an empty value to skip the dedicated watchdog abort timeout handling and fall back to TimeoutStopSec=. Pass "infinity" to disable the timeout logic. Defaults to DefaultTimeoutAbortSec= from the manager configuration file (see systemd-system.conf(5)).

And if a service of Type=notify / Type=notify-reload handles SIGABRT itself (instead of letting the kernel write the core dump), it can send "EXTEND_TIMEOUT_USEC=..." to extend the abort time beyond TimeoutAbortSec=. First receipt of that message must happen before TimeoutAbortSec= is exceeded, and once the abort time has extended beyond TimeoutAbortSec=, the manager lets the service keep aborting, provided it repeats "EXTEND_TIMEOUT_USEC=..." within the interval specified, or terminates itself (see sd_notify(3)).

Added in version 243.

• TimeoutSec= — the shorthand bestie ✨. Configure TimeoutStartSec= and TimeoutStopSec= both to the value you give, in one go.
• TimeoutStartFailureMode=, TimeoutStopFailureMode= — these configure what action gets taken when a daemon service doesn't signal start-up within its configured TimeoutStartSec=, or respectively doesn't stop within TimeoutStopSec=. Takes one of terminate, abort, or kill. Both default to terminate.
  • If terminate is set, the service gets gracefully terminated by sending the signal specified in KillSignal= (defaults to SIGTERM, see systemd.kill(5)). If it still doesn't terminate, FinalKillSignal= is sent after TimeoutStopSec=.
  • If abort is set, WatchdogSignal= is sent instead and TimeoutAbortSec= applies before FinalKillSignal= goes out. Use this to analyze services that fail to start-up or shut-down intermittently 👀.
  • If kill is set, the service is immediately terminated by sending FinalKillSignal= with no further timeout at all. Use this to expedite the shutdown of failing services — no cap, it's over.

Added in version 246.

• RuntimeMaxSec= — configures a maximum time for the service to run. If this is set and the service has been active for longer than the specified time, it gets terminated and put into a failure state 💀. Note: this setting has no effect on Type=oneshot services, because those terminate immediately after activation completes (use TimeoutStartSec= to limit their activation instead). Pass "infinity" — which is the default — to configure no runtime limit.

If a service of Type=notify / Type=notify-reload sends "EXTEND_TIMEOUT_USEC=...", that can extend the runtime past RuntimeMaxSec=. First receipt of that message must happen before RuntimeMaxSec= is exceeded, and once the runtime has extended beyond RuntimeMaxSec=, the manager lets the service keep running as long as it repeats "EXTEND_TIMEOUT_USEC=..." within the interval specified until the service shutdown is achieved by "STOPPING=1" (or termination). See sd_notify(3).

Added in version 229.

• RuntimeRandomizedExtraSec= — this one modifies RuntimeMaxSec= by increasing the maximum runtime by an evenly distributed duration between 0 and the value you specify (in seconds). If RuntimeMaxSec= is unspecified, this feature is disabled. (Spreading out the chaos, love that for us 🌈.)

Added in version 250.

• WatchdogSec= — configures the watchdog timeout for a service. The watchdog activates once start-up is completed. From then on the service must call sd_notify(3) regularly with "WATCHDOG=1" — that's the keep-alive ping, babes 🫶. If the time between two such calls is larger than the configured time, the service is placed in a failed state and terminated with SIGABRT (or whatever signal WatchdogSignal= says). Set Restart= to on-failure, on-watchdog, on-abnormal, or always and the service will automatically restart itself.

The time configured here gets passed to the executed service process in the WATCHDOG_USEC= environment variable — which lets daemons automatically switch on their keep-alive pinging logic when watchdog support is enabled for the service. If you use this option, NotifyAccess= (see below) should be set to open access to the notification socket provided by systemd. If NotifyAccess= isn't set, it gets implicitly set to main. Defaults to 0, which disables this feature. The service can check whether the manager expects watchdog keep-alive notifications — see sd_watchdog_enabled(3) for details. And sd_event_set_watchdog(3) may be used to enable automatic watchdog notification support.

• Restart= — configures whether the service shall be restarted when the service process exits, is killed, or a timeout is reached. The service process may be the main service process, but it can also be one of the processes specified with ExecStartPre=, ExecStartPost=, ExecStop=, ExecStopPost=, or ExecReload=. Now listen — when the death of the process is a result of systemd operation (like a service stop or restart), the service will not be restarted. And timeouts here include missing the watchdog keep-alive ping deadline plus a service start, reload, or stop operation timeout.

Takes one of no, on-success, on-failure, on-abnormal, on-watchdog, on-abort, or always.

• no — the default. Service will not be restarted.
• on-success — restarted only when the service process exits cleanly. And in this context, a clean exit means any of the following:
  • exit code of 0;
  • for types other than Type=oneshot, one of the signals SIGHUP, SIGINT, SIGTERM, or SIGPIPE;
  • exit statuses and signals specified in SuccessExitStatus=.
• on-failure — restarted when the process exits with a non-zero exit code, is terminated by a signal (including on core dump, but excluding those four signals above), when an operation (like a service reload) times out, and when the configured watchdog timeout is triggered.
• on-abnormal — restarted when the process is terminated by a signal (including on core dump, excluding those same four signals), when an operation times out, or when the watchdog timeout is triggered.
• on-abort — restarted only if the service process exits due to an uncaught signal that isn't specified as a clean exit status.
• on-watchdog — restarted only if the watchdog timeout for the service expires.
• always — restarted regardless of whether it exited cleanly or not, got terminated abnormally by a signal, or hit a timeout. (main character energy honestly 👑.)

Heads up bestie: Type=oneshot services will never be restarted on a clean exit status — meaning always and on-success are rejected for them.

Table 1. Exit causes and the effect of the Restart= settings ✨

okay bestie this is THE chart, the one you screenshot and keep in your notes app. Left side is why the process died, and the right side shows which Restart= value will bring it back from the dead 🧟

• Clean exit code or signal — restarts under: always, on-success
• Unclean exit code — restarts under: always, on-failure
• Unclean signal — restarts under: always, on-failure, on-abnormal, on-abort
• Timeout — restarts under: always, on-failure, on-abnormal
• Watchdog — restarts under: always, on-failure, on-abnormal, on-watchdog
• Termination due to OOM — restarts under: always, on-failure, on-abnormal

(and no never restarts anything, obviously, it's giving "I'm not doing that" 💅)

As exceptions to the setting above, the service will not be restarted if the exit code or signal is specified in RestartPreventExitStatus= (see below) or the service is stopped with systemctl stop or an equivalent operation. Also, the services will always be restarted if the exit code or signal is specified in RestartForceExitStatus= (see below).

lowkey important: service restart is subject to unit start rate limiting configured with StartLimitIntervalSec= and StartLimitBurst=, see systemd.unit(5) for details.

Setting this to on-failure is the recommended choice for long-running services, in order to increase reliability by attempting automatic recovery from errors. For services that shall be able to terminate on their own choice (and avoid immediate restarting), on-abnormal is an alternative choice. 👑

---

💫 RestartMode= 💫

Takes a string value that specifies how a service should restart:

• If set to normal (the default), the service restarts by going through a failed/inactive state.

Added in version 254.

• If set to direct, the service transitions to the activating state directly during auto-restart, skipping failed/inactive state. ExecStopPost= is still invoked. OnSuccess= and OnFailure= are skipped.

This option is useful in cases where a dependency can fail temporarily but we do not want these temporary failures to make the dependent units fail. Dependent units are not notified of these temporary failures.

Added in version 254.

• If set to debug, the service manager will log messages that are related to this unit at debug level while automated restarts are attempted, until either the service hits the rate limit or it succeeds, and the $DEBUG_INVOCATION=1 environment variable will be set for the unit. This is useful to be able to get additional information when a service fails to start, without needing to proactively or permanently enable debug level logging in systemd, which is very verbose. This is otherwise equivalent to normal mode.

Added in version 254.

Added in version 254.

---

🌈 SuccessExitStatus= 🌈

Takes a list of exit status definitions that, when returned by the main service process, will be considered successful termination — in addition to the normal successful exit status 0 and, except for Type=oneshot, the signals SIGHUP, SIGINT, SIGTERM, and SIGPIPE. Exit status definitions can be numeric termination statuses, termination status names, or termination signal names, separated by spaces. See the Process Exit Codes section in systemd.exec(5) for a list of termination status names (for this setting only the part without the "EXIT_" or "EX_" prefix should be used). See signal(7) for a list of signal names.

Note that this setting does not change the mapping between numeric exit statuses and their names, i.e. regardless how this setting is used 0 will still be mapped to "SUCCESS" (and thus typically shown as "0/SUCCESS" in tool outputs) and 1 to "FAILURE" (and thus typically shown as "1/FAILURE"), and so on. It only controls what happens as effect of these exit statuses, and how it propagates to the state of the service as a whole. ✨

This option may appear more than once, in which case the list of successful exit statuses is merged. If the empty string is assigned to this option, the list is reset, all prior assignments of this option will have no effect.

Example 1. A service with the SuccessExitStatus= setting 🎀

SuccessExitStatus=TEMPFAIL 250 SIGKILL

Exit status 75 (TEMPFAIL), 250, and the termination signal SIGKILL are considered clean service terminations.

Note: systemd-analyze exit-status may be used to list exit statuses and translate between numerical status values and names.

Added in version 189.

---

🦋 RestartPreventExitStatus= 🦋

Takes a list of exit status definitions that, when returned by the main service process, will prevent automatic service restarts, regardless of the restart setting configured with Restart=. Exit status definitions can be numeric termination statuses, termination status names, or termination signal names, separated by spaces. Defaults to the empty list, so that, by default, no exit status is excluded from the configured restart logic.

This option may appear more than once, in which case the list of restart-preventing statuses is merged. If the empty string is assigned to this option, the list is reset and all prior assignments of this option will have no effect.

Note that this setting has no effect on processes configured via ExecStartPre=, ExecStartPost=, ExecStop=, ExecStopPost= or ExecReload=, but only on the main service process, i.e. either the one invoked by ExecStart= or (depending on Type=, PIDFile=, ...) the otherwise configured main process. 💅

Added in version 189.

---

🌸 RestartForceExitStatus= 🌸

Takes a list of exit status definitions that, when returned by the main service process, will force automatic service restarts, regardless of the restart setting configured with Restart=. The argument format is similar to RestartPreventExitStatus=.

Note that for Type=oneshot services, a success exit status will prevent them from auto-restarting, no matter whether the corresponding exit statuses are listed in this option or not.

Added in version 215.

---

🎀 RootDirectoryStartOnly= 🎀

Takes a boolean argument. If true, the root directory, as configured with the RootDirectory= option (see systemd.exec(5) for more information), is only applied to the process started with ExecStart=, and not to the various other ExecStartPre=, ExecStartPost=, ExecReload=, ExecReloadPost=, ExecStop=, and ExecStopPost= commands. If false, the setting is applied to all configured commands the same way. Defaults to false.

---

💜 NonBlocking= 💜

Set the O_NONBLOCK flag for all file descriptors passed via socket-based activation. If true, all file descriptors >= 3 (i.e. all except stdin, stdout, stderr), excluding those passed in via the file descriptor storage logic (see FileDescriptorStoreMax= for details), will have the O_NONBLOCK flag set and hence are in non-blocking mode. This option is only useful in conjunction with a socket unit, as described in systemd.socket(5) and has no effect on file descriptors which were previously saved in the file-descriptor store for example. Defaults to false.

okay so real talk, bestie, this one matters 💅

Note that if the same socket unit is configured to be passed to multiple service units (that's the Sockets= setting, we're getting to it 👀), and those services have different NonBlocking= configurations, then the precise state of O_NONBLOCK is literally up in the air — it depends on the order in which those services are invoked, and it will possibly change after your service code already took possession of the socket file descriptor, simply because the O_NONBLOCK state of a socket is shared by all file descriptors referencing it. Yeah. Chaos.

Hence it is essential that all services sharing the same socket use the same NonBlocking= configuration, and that they do not change the flag in service code either. no cap, do not freestyle here 🙅‍♀️

• NotifyAccess= — this one controls access to the service status notification socket, the one you poke through the sd_notify(3) call. It takes one of none (the default), main, exec or all.
  • If none: no daemon status updates are accepted from the service processes, and all status update messages are just… ignored. 💀
  • If main: only service updates sent from the main process of the service are accepted.
  • If exec: only service updates sent from any of the main or control processes originating from one of the Exec*= commands are accepted.
  • If all: all service updates from all members of the service's control group are accepted.
  • You should set this option to open access to the notification socket when you're using Type=notify/Type=notify-reload or WatchdogSec= (see above ⬆️). And if you use those options but don't configure NotifyAccess=, it gets implicitly set to main. sneaky but okay ✨
  • Note that sd_notify() notifications may be attributed to units correctly only if either the sending process is still around at the time PID 1 processes the message, or if the sending process is explicitly runtime-tracked by the service manager. The latter is the case if the service manager originally forked off the process, i.e. on all processes that match main or exec. Conversely, if an auxiliary process of the unit sends an sd_notify() message and immediately exits, the service manager might not be able to properly attribute the message to the unit, and thus will ignore it — even if NotifyAccess=all is set for it. 😭
  • Hence, to eliminate all race conditions involving lookup of the client's unit and attribution of notifications to units correctly, sd_notify_barrier() may be used. This call acts as a synchronization point and ensures all notifications sent before this call have been picked up by the service manager when it returns successfully. Use of sd_notify_barrier() is needed for clients which are not invoked by the service manager — otherwise this synchronization mechanism is unnecessary for attribution of notifications to the unit.
• Sockets= — specifies the name of the socket units this service shall inherit socket file descriptors from when the service is started. Normally you shouldn't even need this setting, because all socket file descriptors whose unit shares the same name as the service (just with a different unit name suffix, obviously 🎀) get passed to the spawned process automatically.
  • Note that the same socket file descriptors may be passed to multiple processes simultaneously. Also note that a different service may be activated on incoming socket traffic than the one which is ultimately configured to inherit the socket file descriptors. Or, in other words: the Service= setting of .socket units does not have to match the inverse of the Sockets= setting of the .service it refers to. yes it's giving unhinged, it's fine 🦋
  • This option may appear more than once, in which case the list of socket units is merged. Note that once set, clearing the list of sockets again (for example, by assigning the empty string to this option) is not supported. 🚫
• FileDescriptorStoreMax= — configures how many file descriptors may be stored in the service manager for the service using sd_pid_notify_with_fds(3)'s "FDSTORE=1" messages. This is useful for implementing services that can restart after an explicit request or a crash without losing state. Any open sockets and other file descriptors which should not be closed during the restart may be stored this way. Application state can either be serialized to a file in RuntimeDirectory=, or stored in a memfd_create(2) memory file descriptor. Defaults to 0, i.e. no file descriptors may be stored in the service manager.
  • All file descriptors passed to the service manager from a specific service are passed back to the service's main process on the next service restart (see sd_listen_fds(3) for details about the precise protocol used and the order in which the file descriptors are passed).
  • Any file descriptors passed to the service manager are automatically closed when POLLHUP or POLLERR is seen on them, or when the service is fully stopped and no job is queued or being executed for it (the latter can be tweaked with FileDescriptorStorePreserve=, see below).
  • If this option is used, NotifyAccess= (see above) should be set to open access to the notification socket provided by systemd. If NotifyAccess= is not set, it will be implicitly set to main.
  • The fdstore command of systemd-analyze(1) may be used to list the current contents of a service's file descriptor store. handiest lil thing 🫶
  • Note that the service manager will only pass file descriptors contained in the file descriptor store to the service's own processes, never to other clients via IPC or similar. However, it does allow unprivileged clients to query the list of currently open file descriptors of a service. Sensitive data may hence be safely placed inside the referenced files, but should not be attached to the metadata (e.g. included in filenames) of the stored file descriptors. 🫣
  • If this option is set to a non-zero value the $FDSTORE environment variable will be set for processes invoked for this service. See systemd.exec(5) for details.
  • For further information on the file descriptor store see the File Descriptor Store[1] overview.
  • Added in version 219.
• FileDescriptorStorePreserve= — takes one of no, yes, restart, on-success and controls when to release the service's file descriptor store (i.e. when to close the contained file descriptors, if any).
  • If set to no the file descriptor store is automatically released when the service is stopped.
  • If restart (the default) it is kept around as long as the unit is neither inactive nor failed, or a job is queued for the service, or the service is expected to be restarted.
  • If yes the file descriptor store is kept around and garbage collection of the unit is disabled. The latter is useful to keep entries in the file descriptor store pinned until the unit is removed, the service manager exits, or the file descriptors get EPOLLHUP or EPOLLERR.
  • If on-success the behaviour is identical to yes, except that the file descriptor store is discarded if the unit enters the permanent "failed" state (i.e. once all automated restart attempts driven by Restart= have been exhausted). The store is preserved across the transitionary failed states that precede each individual auto-restart attempt. 💀 but make it polite
  • When set to yes or on-success, and the service is itself running under another service manager (e.g. a service of user@.service, or a payload inside systemd-nspawn(1)), file descriptors pushed into the store are also forwarded one level up via the enveloping manager's $NOTIFY_SOCKET, tagged with the originating unit id, so that they are preserved across restarts of the inner manager and handed back to the originating unit when it is started again. For this to take effect, the enveloping unit must itself enable FileDescriptorStoreMax= and a non-no/restart value for FileDescriptorStorePreserve=. See the File Descriptor Store[1] overview for details. ✨
  • Setting this to yes or on-success also ensures the file descriptor store is kept loaded across a "kexec"-based reboot on kernels supporting the Live Update Orchestrator[2], so that compatible file descriptors (such as memfd_create(2)) are preserved and handed back to the service on the other side. See the File Descriptor Store[1] overview for details.
  • Use systemctl clean --what=fdstore ... to release the file descriptor store explicitly.
  • Added in version 254.
• LUOSession= — takes a whitespace-separated list of names. For each name, when the service is started the service manager creates a LUO session via the Live Update Orchestrator[2] (LUO, i.e. /dev/liveupdate) and hands the resulting session file descriptor to the service through the file descriptor store, using the configured name as its "FDNAME=". The session is hence passed to the service's processes via sd_listen_fds(3), like any other file descriptor store entry. The service may then preserve additional file descriptors (such as memfd_create(2) memory file descriptors) in the session, which survive a "kexec"-based reboot and can be retrieved again on the other side. 🌈
  • A session is only handed out if one with the same name is not already present in the service's file descriptor store. This way a session handed out on first start, or restored across a "kexec", is reused on the next start rather than replaced. The kernel-level session name is derived deterministically from the unit name and the configured name, so that it remains stable and unique, and it fits within the length limit imposed by the kernel. iconic behaviour, honestly 👑
  • This setting implies FileDescriptorStoreMax= is set to at least the number of configured sessions. To ensure the sessions are kept pinned and preserved across a "kexec", combine this with FileDescriptorStorePreserve=yes (see above). If /dev/liveupdate is not available no session is handed out, and the service is started normally. This setting may be specified more than once, in which case the lists are combined. If the empty string is assigned the list is reset and all prior assignments have no effect.
  • For further information on the file descriptor store see the File Descriptor Store[1] overview.
  • This option is only available for system services and is not supported for services running in per-user instances of the service manager. 🙅‍♀️
  • Added in version 262.
• USBFunctionDescriptors= — points at a file stuffed with USB FunctionFS[3] descriptors, for when you're out here implementing USB gadget functions 💅 Heads up bestie: this one only does anything when it's paired up with a socket unit that has ListenUSBFunction= configured. Once ep0 is opened, the contents of that file get written right into it. That's the whole vibe.

Added in version 227. 🎀

• USBFunctionStrings= — same energy as USBFunctionDescriptors= up there ⬆️ but for the file holding your USB FunctionFS strings instead.

Added in version 227.

• OOMPolicy= — okay so this one sets the out-of-memory (OOM) killing policy for both the kernel and the userspace OOM killer, a.k.a. systemd-oomd.service(8). 🫠 On Linux, when memory gets so scarce that the kernel can't even allocate memory for itself anymore, it might just decide to unalive a running process to free stuff up and drop the memory pressure. No cap. And note this: systemd-oomd.service is the more flexible glow-up here — it's trying to prevent OOM situations for userspace too, not just the kernel, by terminating services earlier, before the kernel has to step in and do it the dramatic way. ✨

This setting takes one of continue, stop or kill. Let's break it down:

• continue — if a process in the unit gets killed by the OOM killer, it just gets logged 📝 and the unit keeps running. She's fine, she's unbothered.
• stop — the event gets logged and the unit's processes get terminated cleanly by the service manager. Rude but polite about it.
• kill — if one of the unit's processes gets killed by the OOM killer, the kernel is instructed to kill all the remaining processes of the unit too, by setting the memory.oom.group attribute to 1. Also peep the kernel page Control Group v2[4] for more on that.

In the case of both stop and kill, the service ultimately ends up in the oom-kill failed state, after which Restart= may apply. So like... she can come back. 🙈

Defaults to whatever the setting DefaultOOMPolicy= in systemd-system.conf(5) is set to — except for units where Delegate= is turned on, where it defaults to continue instead. Little special case, love that for them.

Use the OOMScoreAdjust= setting if you wanna configure whether the unit's processes are considered preferred or less-preferred candidates for termination by the Linux OOM killer logic. See systemd.exec(5) for the details, bestie.

This setting also applies to systemd-oomd.service(8). Same as with the kernel OOM kills: this setting determines the state of the unit after systemd-oomd kills a cgroup associated with it. 💀

Added in version 243.

• OpenFile= — takes an argument shaped like this:

"path[:fd-name:options]"

and here's what each piece means 🧚:

• "path" — a path to a file, or to an AF_UNIX socket in the file system.
• "fd-name" — a name that'll get associated with the file descriptor. It can contain any ASCII character, but must not contain control characters or ":", and must be at most 255 characters long. It's optional — if you don't give one, it defaults to the file name.
• "options" — a comma-separated list of access options. Possible values are "read-only", "append", "truncate", "graceful". If you don't specify anything, files get opened in rw mode. And if "graceful" is specified, errors during file/socket opening are just ignored. 🫶 Also: specifying the same option several times is treated as an error. Don't do that.

So the service manager opens the file or socket, and the file descriptor gets passed to the service. If the path is a socket, we call connect() on it. See sd_listen_fds(3) for more deets on how to actually retrieve these file descriptors.

This setting is useful for letting services access files/sockets they can't access themselves — because they're running in a separate mount namespace, or they just don't have the privileges, you know how it is. 👑

You can specify this setting multiple times, and then all the specified paths get opened and all the file descriptors get passed to the service. And if you assign the empty string, the entire list of open files defined before that point gets reset. Clean slate, periodt.

Added in version 253.

• ReloadSignal= — configures the UNIX process signal to send to the service's main process when it's asked to reload the service's configuration. Defaults to SIGHUP. 🎀 Heads up though: this option has no effect unless Type=notify-reload is used, see above.

Added in version 253.

• RefreshOnReload= — takes a boolean argument, or a list of resources defined in systemd.exec(5). Possible values are extensions and credentials, separated by a space. Prepend the list with a single tilde character ("~") to invert the effect. Defaults to extensions. An empty assignment resets the list back to default. If enabled, the corresponding resources get refreshed on service reload — that's ExtensionImages=/ExtensionDirectories= for extensions, and LoadCredential=/ImportCredential=/SetCredential= (along with their Encrypted counterparts) for credentials. If yes, all the resources listed above that the service uses shall be refreshed. ✨

And specially: if this option is set explicitly, and the respective resources are in use, the service may be reloaded without any actual reload mechanism (ExecReload= or Type=notify-reload) for notifying the main process — in which case the reload is considered complete immediately after refreshing. Iconic, honestly. 💅

Added in version 260.

---

Check systemd.unit(5), systemd.exec(5), and systemd.kill(5) for more settings, bestie — there's a whole other world of options waiting for you out there 🌈

✨ COMMAND LINES ✨

okay bestie this section is about how the command lines actually get parsed and how variables and specifiers get substituted, for ExecStart=, ExecStartPre=, ExecStartPost=, ExecReload=, ExecStop=, ExecStopPost=, and ExecCondition=. yes that's a lot of Execs. we love a family 💖

You can have multiple command lines — just use the relevant setting multiple times. stack 'em up, queen 👑

Each command line gets unquoted using the rules in the "Quoting" section of systemd.syntax(7). The first item becomes the command to execute, and everything after it is arguments. simple as that fr fr

So this syntax is inspired by shell syntax but it's giving "shell's cooler cousin who doesn't do everything shell does" — only the meta-characters and expansions described in the next paragraphs are understood, and the variable expansion works differently. Specifically: redirection with <, <<, >, and >>? nope. pipes with |? nope. running programs in the background with &? nope. other shell syntax elements? also nope. it's not a shell, bestie, don't treat it like one 😭

The command to execute may contain spaces, but control characters are not allowed. periodt.

Each command may be prefixed with a bunch of special characters, and here's the whole lineup:

• @ — If the executable path is prefixed with @, the second specified token will be passed as argv[0] to the executed process (instead of the actual filename), followed by the further arguments specified — unless | is also specified, in which case it enables login shell semantics for the shell spawned by prefixing - to argv[0]. iconic behavior tbh
• - — If the executable path is prefixed with -, an exit code of the command normally considered a failure (i.e. non-zero exit status or abnormal exit due to signal) is recorded, but has no further effect and is considered equivalent to success. basically "we saw the L but we're pretending it's a W" 💅
• : — If the executable path is prefixed with :, environment variable substitution (as described below) is not applied. no substitution, no notes.
• + — If the executable path is prefixed with + then the process is executed with full privileges. In this mode, privilege restrictions configured with User=, Group=, CapabilityBoundingSet= or the various file system namespacing options (such as PrivateDevices=, PrivateTmp=) are not applied to the invoked command line (but still affect any other ExecStart=, ExecStop=, ... lines). Resource limits configured with LimitCPU=, LimitCORE=, LimitNOFILE= and the other resource limit settings described in systemd.exec(5) (including the defaults set with DefaultLimitCPU=, DefaultLimitCORE=, ... in systemd-system.conf(5)) are not applied either, and the invoked command inherits the resource limits of the service manager itself. However, note that this will not bypass options that apply to the whole control group, such as DevicePolicy= — see systemd.resource-control(5) for the full list. so like, big main character energy but there's still a group chat you can't leave 🫶
• ! — Similar to the + character discussed above, this permits invoking command lines with elevated privileges. But unlike +, the ! character exclusively alters the effect of User=, Group= and SupplementaryGroups=, i.e. only the stanzas that affect user and group credentials. Note that this setting may be combined with DynamicUser=, in which case a dynamic user/group pair is allocated before the command is invoked, but credential changing is left to the executed process itself. it's the "I'll get myself ready, you just hand me the keys" option ✨
• | — If | is specified standalone as executable path, invoke the default shell of User=. If specified as a prefix, use the shell (-c) to spawn the executable. When @ is used in conjunction, argv[0] of shell will be prefixed with - to enable login shell semantics. shell girlies rise 🌸

So @, |, -, :, and one of +/! may be used together, and they can appear in any order. The one rule: + and ! may not be specified at the same time. pick a lane babes 😤

---

For each command, the first argument must be either an absolute path to an executable or a simple file name without any slashes. If the command is not a full (absolute) path, it will be resolved to a full path using a fixed search path determined at compilation time. Searched directories include /usr/local/bin/, /usr/bin/, and their sbin/ counterparts (only on systems using split bin/ and sbin/). So it's totally safe to just use the executable name if it lives in any of the "standard" directories — otherwise you must use an absolute path. Hint: this search path may be queried using systemd-path search-binaries-default. hot tip, screenshot it 📸

The command line accepts % specifiers as described in systemd.unit(5).

An argument solely consisting of ; must be escaped, i.e. specified as \;. don't forget the backslash or it'll get eaten alive 💀

---

🦋 Variable substitution 🦋

Basic environment variable substitution is supported. Use ${FOO} as part of a word, or as a word of its own, on the command line — in which case it will be erased and replaced by the exact value of the environment variable (if any) including all whitespace it contains, always resulting in exactly a single argument.

Use $FOO as a separate word on the command line, in which case it will be replaced by the value of the environment variable split at whitespace, resulting in zero or more arguments. For this type of expansion, quotes are respected when splitting into words, and afterwards removed. no cap, quotes get their moment and then leave the group chat 🙈

Example:

Environment="ONE=one" 'TWO=two two'
ExecStart=echo $ONE $TWO ${TWO}

This will execute /bin/echo with four arguments: "one", "two", "two", and "two two".

Example:

Environment=ONE='one' "TWO='two two' too" THREE=
ExecStart=/bin/echo ${ONE} ${TWO} ${THREE}
ExecStart=/bin/echo $ONE $TWO $THREE

This results in /bin/echo being called twice — the first time with arguments "one", "'two two' too", "", and the second time with arguments "one", "two two", "too".

Unless for commands with the special executable prefix :, to pass a literal dollar sign, use $$. Variables whose value is not known at expansion time are treated as empty strings. Note that the first argument (i.e. the program to execute) may not be a variable. rude but true.

Variables to be used in this fashion may be defined through Environment= and EnvironmentFile=. In addition, variables listed in the section "Environment variables in spawned processes" in systemd.exec(5), which are considered "static configuration", may be used (this includes e.g. $USER, but not $TERM).

Note that shell command lines are not directly supported, and | invokes the user's default shell which isn't deterministic. It's recommended to specify a shell implementation explicitly if portability is desired. bestie please just be specific, we're begging 🙏

Example:

ExecStart=sh -c 'dmesg | tac'

Example:

ExecStart=echo one
ExecStart=echo "two two"

This will execute echo two times, each time with one argument: "one" and "two two", respectively. Because two commands are specified, Type=oneshot must be used. (yeah, oneshot, don't fight it.)

Example:

Type=oneshot
ExecStart=:echo $USER
ExecStart=-false
ExecStart=+:@true $TEST

This will execute /usr/bin/echo with the literal argument "$USER" (: suppresses variable expansion), and then /usr/bin/false (the return value will be ignored because - suppresses checking of the return value), and /usr/bin/true (with elevated privileges, with "$TEST" as argv[0]). three commands, three vibes, all valid ✨

Example:

ExecStart=echo / >/dev/null & \; \
ls

This will execute echo with five arguments: "/", ">/dev/null", "&", ";", and "ls". yeah it just... passes them through. no shell magic happening here, it's giving "look what I can't do" 💅

✨ EXAMPLES ✨

---

🍓 Example 3. Simple service

okay bestie, this one is the classic. The unit file below makes a service that runs /usr/sbin/foo-daemon. Since there's no Type= written anywhere, the default Type=simple gets assumed, periodt. systemd just assumes the unit is started the second the program starts executing. no waiting, no drama.

[Unit]
Description=Foo

[Service]
ExecStart=/usr/sbin/foo-daemon

[Install]
WantedBy=multi-user.target

so heads up: systemd is assuming here that the process it started will just keep running until the service terminates. That's the whole vibe. But if your program daemonizes itself (i.e. it forks 🙈), then please use Type=forking instead, okay?

and since there's no ExecStop= specified, systemd will send SIGTERM to all processes started from this service, and after a timeout it also sends SIGKILL. 💀 You can modify that behavior though — see systemd.kill(5) for the details.

also lowkey important: this unit type does not include any type of notification for when a service has completed initialization. None. Zero. So if you want that, you should use other unit types, like Type=notify/Type=notify-reload if the service understands systemd's notification protocol, Type=forking if the service can background itself, or Type=dbus if the unit acquires a DBus name once initialization is complete. See below, bestie. 👇

---

🧚 Example 4. Oneshot service

sometimes units should just execute an action without keeping active processes around, like a filesystem check or a cleanup action on boot. For that, Type=oneshot exists. 💅 Units of this type will wait until the process specified terminates and then fall back to being inactive. The following unit will perform a cleanup action:

[Unit]
Description=Cleanup old Foo data

[Service]
Type=oneshot
ExecStart=/usr/sbin/foo-cleanup

[Install]
WantedBy=multi-user.target

note that systemd will consider the unit to be in the state "starting" until the program has terminated, so ordered dependencies will wait for the program to finish before starting themselves. Then the unit reverts to the "inactive" state after the execution is done — it never reaches the "active" state. Which means another request to start the unit will perform the action again. 🔁

Type=oneshot are the only service units that may have more than one ExecStart= specified. And for units with multiple commands (Type=oneshot), all commands will be run again.

🚫 For Type=oneshot, Restart=always and Restart=on-success are not allowed. Not up for debate, bestie.

---

🌸 Example 5. Stoppable oneshot service

similarly to the oneshot services, there are sometimes units that need to execute a program to set up something and then execute another one to shut it down, but no process remains active while they're considered "started". Network configuration can sometimes fall into this category. Another use case is if a oneshot service shall not be executed each time when they are pulled in as a dependency, but only the first time. 👑

For this, systemd knows the setting RemainAfterExit=yes, which causes systemd to consider the unit to be active if the start action exited successfully. This directive can be used with all types, but is most useful with Type=oneshot and Type=simple. With Type=oneshot, systemd waits until the start action has completed before it considers the unit to be active, so dependencies start only after the start action has succeeded. With Type=simple, dependencies will start immediately after the start action has been dispatched. The following unit provides an example for a simple static firewall:

[Unit]
Description=Simple firewall

[Service]
Type=oneshot
RemainAfterExit=yes
ExecStart=/usr/local/sbin/simple-firewall-start
ExecStop=/usr/local/sbin/simple-firewall-stop

[Install]
WantedBy=multi-user.target

and since the unit is considered to be running after the start action has exited, invoking systemctl start on that unit again will cause no action to be taken. it's giving "already handled it, bye" ✨

---

🦋 Example 6. Traditional forking services

many traditional daemons/services background (i.e. fork, daemonize) themselves when starting. Set Type=forking in the service's unit file to support this mode of operation. systemd will consider the service to be in the process of initialization while the original program is still running. Once it exits successfully and at least a process remains (and RemainAfterExit=no), the service is considered started. 🎀

often, a traditional daemon only consists of one process. Therefore, if only one process is left after the original process terminates, systemd will consider that process the main process of the service. In that case, the $MAINPID variable will be available in ExecReload=, ExecStop=, etc.

in case more than one process remains, systemd will be unable to determine the main process, so it will not assume there is one. In that case, $MAINPID will not expand to anything. 😭 However, if the process decides to write a traditional PID file, systemd will be able to read the main PID from there. Please set PIDFile= accordingly. And note that the daemon should write that file before finishing with its initialization — otherwise systemd might try to read the file before it exists. race condition vibes, we don't love that.

The following example shows a simple daemon that forks and just starts one process in the background:

[Unit]
Description=My Simple Daemon

[Service]
Type=forking
ExecStart=/usr/sbin/my-simple-daemon -d

[Install]
WantedBy=multi-user.target

Please see systemd.kill(5) for details on how you can influence the way systemd terminates the service. 🔥

---

💜 Example 7. DBus services

for services that acquire a name on the DBus system bus, use Type=dbus and set BusName= accordingly. The service should not fork (daemonize). systemd will consider the service to be initialized once the name has been acquired on the system bus. The following example shows a typical DBus service:

[Unit]
Description=Simple DBus Service

[Service]
Type=dbus
BusName=org.example.simple-dbus-service
ExecStart=/usr/sbin/simple-dbus-service

[Install]
WantedBy=multi-user.target

for bus-activatable services, do not include a [Install] section in the systemd service file, but use the SystemdService= option in the corresponding DBus service file, for example (/usr/share/dbus-1/system-services/org.example.simple-dbus-service.service):

[D-BUS Service]
Name=org.example.simple-dbus-service
Exec=/usr/sbin/simple-dbus-service
User=root
SystemdService=simple-dbus-service.service

Please see systemd.kill(5) for details on how you can influence the way systemd terminates the service. 💖

---

🌈 Example 8. Services that notify systemd about their initialization

Type=simple services are really easy to write, but they have the major disadvantage of systemd not being able to tell when initialization of the given service is complete. 😩 For this reason, systemd supports a simple notification protocol that allows daemons to make systemd aware that they are done initializing. Use Type=notify or Type=notify-reload for this. A typical service file for such a daemon would look like this:

[Unit]
Description=Simple Notifying Service

[Service]
Type=notify-reload
ExecStart=/usr/sbin/simple-notifying-service

[Install]
WantedBy=multi-user.target

⚠️ Note that the daemon has to support systemd's notification protocol — else systemd will think the service has not started yet and kill it after a timeout. no cap. 💀

for an example of how to update daemons to support this protocol transparently, take a look at sd_notify(3). systemd will consider the unit to be in the 'starting' state until a readiness notification has arrived. ⏳

Please see systemd.kill(5) for details on how you can influence the way systemd terminates the service.

and to avoid code duplication, it is preferable to use sd_notify(3) when possible, especially when other APIs provided by libsystemd(3) are also used. But note that the notification protocol is very simple and guaranteed to be stable as per the Interface Portability and Stability Promise[5], so it can be reimplemented by services with no external dependencies. For a self-contained example, see sd_notify(3). 🫶

---

👑 SEE ALSO 👑

systemd(1), systemctl(1), systemd-system.conf(5), systemd.unit(5), systemd.exec(5), systemd.resource-control(5), systemd.kill(5), systemd.directives(7), systemd-run(1)

---

📎 NOTES 📎

1. File Descriptor Store https://systemd.io/FILE_DESCRIPTOR_STORE
2. Live Update Orchestrator https://docs.kernel.org/userspace-api/liveupdate.html
3. USB FunctionFS https://docs.kernel.org/usb/functionfs.html
4. Control Group v2 https://docs.kernel.org/admin-guide/cgroup-v2.html
5. Interface Portability and Stability Promise https://systemd.io/PORTABILITY_AND_STABILITY/

---

and that's a wrap bestie!! 🎀✨ this whole thing is systemd 262.1, page SYSTEMD.SERVICE(5) — you read the entire manual page, which is lowkey iconic of you, no notes, ate and left no crumbs. go start some services (responsibly), and remember: Type= is not a suggestion, it's a lifestyle. 💅🦋 mwah!