[slurm-users] Half of cpus used despite CR_CPU

[Adrian Sevcenco]

Hi! I have a weird situation with a cluster that i switched from CR_Core to CR_CPU
select/cons_res, TaskPlugin=task/affinity,task/cgroup TaskPluginParam=autobind=threads

despite reporting in the jobs that only 1 CPU is needed:

NumNodes=1 NumCPUs=1 NumTasks=1 CPUs/Task=1 ReqB:S:C:T=0:0:*:*
     TRES=cpu=1,node=1,billing=1
     Socks/Node=* NtasksPerN:B:S:C=0:0:*:* CoreSpec=*
     JOB_GRES=(null)
       Nodes=issaf-0-0 CPU_IDs=106-107 Mem=0 GRES=
     MinCPUsNode=1 MinMemoryNode=0 MinTmpDiskNode=0
     Features=(null) DelayBoot=00:00:00
     OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null)

only half of the job slots are used

then, sinfo reports that all cpus are used

root at issaf: ~ # sinfo -o "%10R %.16N %.6a %.14F %.14C %.12L %.12l"
PARTITION          NODELIST  AVAIL NODES(A/I/O/T)  CPUS(A/I/O/T)  DEFAULTTIME    TIMELIMIT
CLUSTER       issaf-0-[0-2]     up        3/0/0/3    384/0/0/384   2-00:00:00  20-00:00:00

but
root at issaf: ~ # squeue -h -t R | wc -l
192


Does anyone have any idea/experience why not all 384 cores are used as 384 job slots?

Thanks a lot!
Adrian