Hi all,
I stumbled over a cgroup/systemd(?) related question, when I was trying
to fix a cgroup value for the condor parent group and systemd
(presumably) took notice after some time.
I limited the walltime for all condor cgroupds to 1/4 of the overall
(logical cpu) wall time [1] (stopgap as the cpu govenor vs. board had
issues).
Which worked fine until an auto-update updated some packages and
triggered systemd, which reset the parameter to its original value (-1).
So, I wonder if somebody knows a way to fix on the fly parameters in the
condor cgroup(s) actuallybelonging to systemd (beyond modifying the unit
as I would like to avoid loosing running jobs etc. [2])?
Cheers,
Thomas
[1]
> TIMEWINDOW=1000000
> echo ${TIMEWINDOW} >
/sys/fs/cgroup/cpu,cpuacct/system.slice/condor.service/cpu.cfs_period_us
> echo $(cat /proc/cpuinfo | grep processor | wc -l)*${TIMEWINDOW}/4 |
bc > /sys/fs/cgroup/cpu,cpuacct/system.slice/condor.service/cpu.cfs_quota_us
[2]
https://www.freedesktop.org/software/systemd/man/systemd.resource-control.html
Attachment:
smime.p7s
Description: S/MIME Cryptographic Signature