Re: s390 private runner CI job timing out

[Peter Maydell]

On Thu, 6 Apr 2023 at 13:13, Christian Borntraeger
<borntraeger@linux.ibm.com> wrote:
>
>
>
> Am 06.04.23 um 14:05 schrieb Peter Maydell:
> > On Thu, 6 Apr 2023 at 12:17, Christian Borntraeger
> > <borntraeger@linux.ibm.com> wrote:
> >>
> >> Am 06.04.23 um 12:44 schrieb Peter Maydell:
> >>> On Thu, 6 Apr 2023 at 11:40, Christian Borntraeger
> >>> <borntraeger@linux.ibm.com> wrote:
> >>>> Am 06.04.23 um 11:21 schrieb Peter Maydell:
> >>>>> Christian, does our S390X machine get a guaranteed amount of CPU,
> >>>>> or does it depend on what else is running on the hardware?
> >>>>
> >>>> I think its a shared system with shared CPUs. Can you check the steal
> >>>> time in top or proc? If this is far too high we could ask to give you
> >>>> more weight for that VM.
> >>>
> >>> It's idle at the moment and steal time seems to be low (0.0 .. 0.3);
> >>> I'll try to remember to check next time it's running a job.
> >>>
> >>
> >> Do you have /proc/stat ?
> >
> > Yes; hopefully it means more to you than it does to me :-)
> >
> > linux1@qemu01:~$ cat /proc/stat
> > cpu  60904459 604975 15052194 1435958176 17128179 351949 758578 22218760 0 0
> > cpu0 15022535 146734 3786909 358774818 4283172 98313 237156 5894809 0 0
> > cpu1 15306890 151164 3746024 358968957 4378864 85629 172492 5434255 0 0
> > cpu2 15307709 157180 3762691 359141276 4138714 85736 176367 5474594 0 0
> > cpu3 15267324 149895 3756569 359073124 4327428 82269 172562 5415101 0 0
>
> This is
> user,nice,system,idle,iowait,irq,softirq,steal,guest,guest_nice
> So overall there is some (20-30% since the last reboot) steal going on.
> Not sure if this is the real problem since it is only Ubuntu 20.04.
> Does a higher timeout make the problem go away?

I would expect that raising the timeout would improve this (it seems
to be on the borderline where some jobs make it in in time and some don't),
but on the other hand 75 mins is already a pretty high timeout.
I would rather we looked at whether we could cut down the amount
we are compiling/testing on this box instead.

thanks
-- PMM