Re: [HTCondor-users] Internal cgroup error when retrieving iowait statistics

[Thomas Hartmann <thomas.hartmann@xxxxxxx>]

Hi Gianmauro,

just a guess (not sure if iowait stats are available everywhere), but 
can you check, if you have the cpu (and memory) cgroups set on your workers?

We have ours on CentOS7 under
  /sys/fs/cgroup/cpu,cpuacct/system.slice/condor.service/
  /sys/fs/cgroup/memory/system.slice/condor.service/
(but the virtual cgroup mounts might be at different paths, so in case 
you might have quick check `mount | grep cgroup` )


If you have configured the `BASE_CGROUP` ad, it should match the cgroup 
mounts, e.g.,
  BASE_CGROUP = /system.slice/condor.service
(but I am not sure, if the ad is actually still required on systemd systems)

Cheers,
  Thomas


On 28/09/2021 09.27, Gianmauro Cuccuru wrote:
> I have a lot of these messages in the executors ProcLog.
> Do you know which could be the reason?
>
> 09/28/21 08:23:22 : gathering usage data for family with root pid 307209
> 09/28/21 08:23:22 : Internal cgroup error when retrieving iowait 
> statistics: No such file or directory
> 09/28/21 08:23:22 : PROC_FAMILY_GET_USAGE
>
> HTCondor version is 8.8.15-1,
>
> Thank you,
> Gianmauro

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature