Hi all, i have used scontrol to make a reservation for all cluster nodes for a scheduled system maintenance (cluster shutdown) on 2013-08-05T18:00:00. My problem is, that in one partition, jobs are not eligable to run (ReqNodeNotAvail) even though nodes are idle and the job time (7-00:00:00) still fits before the maintenance window. I have the problem only on a partition which is marked "Shared=YES", the other partition with "Shared=FORCE:1" is working fine. Another interesting fact is, that jobs are still accepted when they complete until 2013-07-29T00:00:00.
Is this a bug or a configuration problem? Markus bagheera2 slurm # scontrol show reserv ReservationName=root_3 StartTime=2013-08-05T18:00:00 EndTime=2013-08-08T18:00:00 Duration=3-00:00:00 Nodes=kaa-[1-17,26-33,45-48,50-106] NodeCnt=86 CoreCnt=1420 Features=(null) PartitionName=(null) Flags=MAINT,IGNORE_JOBS,SPEC_NODES Users=root Accounts=(null) Licenses=(null) State=INACTIVE bagheera2 slurm # srun -p parallel -w kaa-104 --qos=long --time=2-13:15:00 hostname kaa-104 bagheera2 slurm # srun -p parallel -w kaa-104 --qos=long --time=2-13:20:00 hostname srun: Required node not available (down or drained) srun: job 12899 queued and waiting for resources PartitionName=serial AllocNodes=ALL AllowGroups=ALL Default=NO DefaultTime=NONE DisableRootJobs=NO GraceTime=0 Hidden=NO MaxNodes=UNLIMITED MaxTime=UNLIMITED MinNodes=1 MaxCPUsPerNode=UNLIMITED Nodes=kaa-[1-17,26-33,45-48,50-84] Priority=1 RootOnly=NO ReqResv=NO Shared=FORCE:1 PreemptMode=GANG,SUSPEND State=UP TotalCPUs=492 TotalNodes=64 SelectTypeParameters=N/A DefMemPerNode=UNLIMITED MaxMemPerNode=UNLIMITED PartitionName=parallel AllocNodes=ALL AllowGroups=ALL Default=NO DefaultTime=NONE DisableRootJobs=NO GraceTime=0 Hidden=NO MaxNodes=UNLIMITED MaxTime=UNLIMITED MinNodes=1 MaxCPUsPerNode=UNLIMITED Nodes=kaa-[95-106] Priority=1 RootOnly=NO ReqResv=NO Shared=YES:4 PreemptMode=GANG,SUSPEND State=UP TotalCPUs=768 TotalNodes=12 SelectTypeParameters=N/A DefMemPerNode=UNLIMITED MaxMemPerNode=UNLIMITED
slurmconfig
Description: Binary data
