[slurm-users] can't use GRES to increase Raw Usage

[Erik Bryer]

I have 2 partitions:
PartitionName=scavenge Nodes=saga-test01,saga-test02 MaxTime=72:00:00 State=UP PriorityTier=0 PreemptMode=REQUEUE AllowQos=scavenge AllowAccounts=borrowed,gaia default=yes TRESBillingWeights="CPU=1.0,Mem=0.25G,GRES/foolsgold=200000"
PartitionName=scavtres Nodes=saga-test01,saga-test02 MaxTime=72:00:00 State=UP PriorityTier=0 PreemptMode=REQUEUE AllowQos=scavenge AllowAccounts=borrowed,gaia default=yes TRESBillingWeights="CPU=1.0,Mem=0.25G,GRES/foolsgold=200.0"

I run 2 jobs in each partition, each job using 2 gpus. The job in the first partition should bill at a higher rate, but it doesn't. After about 1 hour of running the jobs, I see little increase in RawUsage shown by sshare -a.

$ sshare -a | egrep "(Raw|\-\-\-|borrowed.*sagatest01)"
             Account       User  RawShares  NormShares    RawUsage  EffectvUsage  FairShare
-------------------- ---------- ---------- ----------- ----------- ------------- ----------
  borrowed           sagatest01       1582    0.333333       25911      0.000103   0.166667

RawUsage has increased by 540 after 30 minutes. If I change in the partition definition CPU=1.0 to CPU=200000, the RawUsage numbers leap up. How can I get TRESBillingWeights to acknowledge GRES?

Thanks,
Erik
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20210107/8cd8f666/attachment.htm>