Re: [HTCondor-users] Massive overinflation of RemoteUserCpu & RemoteSysCpu

[Petr Vokac <petr.vokac@xxxxxxx>]

On 9/17/24 14:29, Thomas Hartmann wrote:
> I had observed some PIDs from previous jobs in non-interuptable sleep 
> staying in a job cgroup, that got "re-used" for a new follow up job. 
> I.e., that a slot's cgroup PID list contains PID(s) from a previous 
> job, that had run on the same local slot/cgroup.
> In the end, I assume that these PIDs are no issue as they are 
> practically dead and will never awake from their sleep...
Yes, old PIDs left from previous jobs don't utilize any resources, 
because they are stuck in non-interuptable sleep, but "cpu.stat" in 
existing cgroupv2 is not cleared for new job => systime and usertime for 
jobs running in existing subgroup gets time from all previous jobs... so 
CPU account is completely wrong.
Petr