[systemd-devel] When service consumption cpu time message will be at the journal?

u34 at net9.cf u34 at net9.cf
Sun Jun 25 01:39:30 UTC 2023


Lennart Poettering <lennart at poettering.net> wrote:

> On Di, 20.06.23 18:13, u34 at net9.cf (u34 at net9.cf) wrote:
> 
> > Hello,
> > I have timers to run user, and system, services.
> > Usually, but not always, the journal has a
> >
> >     systemd[1]: NAME.service: Consumed DURATION CPU time.
> >
> > as the last message for each activation of a service. I wonder whether it
> > is expected that sometimes there is no such message. I can not predict,
> > or characterise, when such message will not be given, other than saying
> > I suspect the user service is more likely to not have such a
> > message.
> 
> This should be shown for all sevices that go down fully (i.e. into
> dead/failed states)


What about stopped, as in 

    $ journalctl --user --since ...
    ... Stopped
    ... Failed with result 'signal'.
    ... Main process exited, code=killed, status=15/TERM
    .
    .
    .

Once again, it is activated periodically by a timer.

?

--
u34

>                     and for which CPUAccounting=1 is one (which can
> happen by means of global DefaultCPUAccount=1).
> 
> Otherwise file a github issue, and include a relevant debug log output
> (i.e. issue systemd-analyze log-level debug first) of an occasion
> where the message wasn't written.
> 
> Without such logs this is not actionable to us.
> 
> Lennart
> 
> --
> Lennart Poettering, Berlin


More information about the systemd-devel mailing list