3

I have found systemd-run which allows one to run processes in the background once off ("transient services"). I always specify the service unit name with --unit $NAME. But if I have already run the systemd-run command and my process is running, then systemd-run will fail, with non-zero exit code. Is there anyway to tell systemd-run to be more idempotent and not fail in this case?

Currently I'm doing:

 systemctl is-active $NAME || systemd-run --unit $NAME $COMMAND

Is there a better way?

This is on Ubuntu 18.04, with the current systemd for that (version 237?)

Improve this question

3 Answers 3

Reset to default
5

-G, --collect

Unload the transient unit after it completed, even if it failed. Normally, without this option, all units that ran and failed are kept in memory until the user explicitly resets their failure state with systemctl reset-failed or an equivalent command. On the other hand, units that ran successfully are unloaded immediately. If this option is turned on the "garbage collection" of units is more aggressive, and unloads units regardless if they exited successfully or failed. This option is a shortcut for --property=CollectMode=inactive-or-failed, see the explanation for CollectMode= in systemd.unit(5) for further information.

systemd-run --collect --unit $NAME $COMMAND
Improve this answer
1
  • 1
    Note that after unloading the unit (with --collect or systemctl reset-failed or through rebooting) it cannot be inspected using systemctl status, but its logs are still in journal, i.e. a failing transient unit tells us why with journalctl -u $NAME -e. Commented Apr 6, 2022 at 7:22
1

I don't think there is a way not to fail.

The problem here is that systemd unit names must be unique. If they aren't, then manipulating the units could become ambiguous. For this reason systemd perceives an error if you try to duplicate unit names, and I don't think there is a way to make it think otherwise.

If you could convince systemd-run to run the second service with the same name, what would happen if you issue a systemctl stop command? Which of the two identically named units should stop? The one you ran first? The last? Both?

Improve this answer
1
  • That's good points. I don't want to run the service twice. An option like --no-start-if-active or --reset-failed-unit-name (to run systemctl reset-failed $NAME) would address this issue without starting a second unit. Commented May 6, 2019 at 8:19
0

You will need --collect for the unit to be deleted (see answer https://serverfault.com/a/1056478/161855).

$ systemctl is-active foo || systemd-run --collect --unit foo sleep 1000 inactive Running as unit: foo.service $ systemctl is-active foo || systemd-run --collect --unit foo sleep 1000 active

This introduces a slight TOCTOU, but is better than nothing, and will not cause systemd-run to print an error if the unit is already running.

Improve this answer

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.