Restart a service and actually verify it came back

sudo systemctl restart nginx && systemctl is-active nginx restarts the service and prints one word telling you whether it is really running. The restart itself is silent on success and equally silent when the process dies a second later, so the check has to be its own command — and is-active answers inactive for a misspelled unit name too, which is where most people fool themselves.

1. Restart, then check in the same line

$ clai restart docker and tell me if it came back→ sudo systemctl restart docker && systemctl is-active dockeractive

restart prints nothing when it succeeds, and silence is not proof of anything. is-active returns a single word with exit code 0, which is exactly what you want after &&.

2. The trap: a typo in the unit name

$ clai check whether the nosuch service is running→ systemctl is-active nosuch.serviceinactive

A unit that does not exist answers inactive with exit code 4 — not "unit not found". In a script that reads as "the service is stopped" when in fact you mistyped the name. status is honest about it: Unit nosuch.service could not be found.

3. What status is actually telling you

$ clai show the state of the docker service→ systemctl status docker● docker.service - Docker Application Container Engine     Loaded: loaded (/usr/lib/systemd/system/docker.service; enabled; preset: enabled)     Active: active (running) since Mon 2026-08-24 14:15:54 +07; 3 days ago Invocation: ca41df104d0f46f5a69c964621baf011TriggeredBy: ● docker.socket       Docs: https://docs.docker.com

Two lines carry the meaning. Loaded says the unit file was found, and enabled means it starts at boot. Active carries the uptime: if it says seconds when you restarted a minute ago, the service crashed and came back on its own.

4. A check for scripts, not for eyes

$ clai show docker state in machine readable form→ systemctl show -p ActiveState,SubState,MainPID,NRestarts dockerActiveState=activeSubState=runningMainPID=5425NRestarts=0

show doesn't format anything for humans, so it survives systemd upgrades that reshuffle the pretty output. NRestarts climbing is the fastest signal that a Restart= policy is papering over a crash loop.

5. Find everything that failed

$ clai show failed services→ systemctl --failed  UNIT              LOAD   ACTIVE SUB    DESCRIPTION● logrotate.service loaded failed failed Rotate log files

Short and to the point. For a single unit, systemctl is-failed <name> prints failed and exits 0 when the unit really is in that state.

6. Clear the failed state

$ clai clear the failed state of the logrotate service→ sudo systemctl reset-failed logrotate.service

Until the failed state is cleared, the unit stays in the --failed list and the StartLimitBurst counter can refuse new attempts with "start request repeated too quickly". reset-failed zeroes the counter and removes the unit from the list.

Gotchas

• Silence after restart proves nothing. systemctl restart prints an empty line both when the service is healthy and when it died one second in. The verification is always a separate command.
• is-active can't tell a typo from a stopped service. Both answer inactive. The exit codes differ — 4 for a unit that doesn't exist — but you won't see that by eye, so reach for status whenever the answer surprises you.
• restart does not re-read a changed unit file. After editing a .service file you need sudo systemctl daemon-reload, otherwise systemd starts the old version and you go hunting for a bug in a config you already fixed.

Related questions

What's the difference between restart and reload? reload asks the service itself to re-read its config without dropping connections; restart kills the process and starts it again. When a unit has no reload support, reload-or-restart picks whatever exists.

How do I see why a service didn't come up? journalctl -u <name> -n 50 -o cat shows the last lines of its log without the timestamp prefixes — the reason is usually right there.

How do I know it will survive a reboot? systemctl is-enabled <name> answers enabled or disabled. A service running right now can still be disabled and never come back after a restart of the machine.

See also

• journalctl: logs for one service
• kill -9 doesn't work
• Check if a port is open

CliAI turns "restart it and tell me if it came back" into both commands, and shows them before either one runs. Install it in one line.