service

control a systemd unit

With running: false, a missing unit file is ok, with the note <unit> does not exist. A service step with confirm_access: true runs the access check after the step, and rolls back when it fails. Run winch doc step.confirm_access. In a handler, restarted and reloaded check as ok when a service step of the run started the unit after the handler's last notify, no earlier step of the handler changed anything, and the step has no confirm_access.

A unit of type simple, exec, forking, or idle does not report readiness. After its start or restart, apply watches it for 2 seconds. A reload of such a unit without ExecReload is a restart, and gets the watch. The step fails when the unit is not running then, or when systemd restarted it in that time. A unit that crashes after the 2 seconds passes. A reload ends when ExecReload ends: one that only sends a signal is confirmed as sent, and the unit can still end after it. A health gate checks that a service works. Run winch doc pipeline.health.

Keys

service.name

string, required

Unit, with or without .service, such as nginx or certbot.timer.

service.running

boolean

Start or stop; unset leaves it. Without running, the step does not start or stop the unit.

service.enabled

boolean

Start at boot; unset leaves it. Without enabled, the step does not change it.

service.reloaded

boolean, default false

Reload, in handlers and winch run only. A unit that cannot reload restarts.

service.restarted

boolean, default false

Restart, in handlers and winch run only. A stopped unit stays stopped.

Example

- service:
    name: nginx
    running: true
    enabled: true