Hi Scott,

You need to configure a DefMemPerCPU or DefMemPerNode. Later versions of 
14.11.x default to all the memory in a node if one of these are not configured.

Hope this helps,

Will

On Jun 9, 2015, at 9:52 AM, Zentz, Scott C. <[email protected]> wrote:

> Hello Everyone,
>  
>                 We recently upgraded from slurm-14.11.2 to slurm 14.11.7 and 
> now it looks like the default is to schedule 1 job per node and we would like 
> 1 job per core. Below is a copy of our config.
>  
> Thanks!
> -scz
>  
>  
> ControlMachine=xxxxxx
> ControlAddr=xxx.xxx.xxx.xxx
> AuthType=auth/munge
> CacheGroups=0
> CryptoType=crypto/munge
> DisableRootJobs=NO
> MailProg=/bin/mail
> MaxJobCount=100000
> MpiDefault=none
> ProctrackType=proctrack/pgid
> ReturnToService=1
> SlurmctldPidFile=/var/run/slurmctld.pid
> SlurmctldPort=6817
> SlurmdPidFile=/var/run/slurmd.pid
> SlurmdPort=6818
> SlurmdSpoolDir=/var/spool/slurmd
> SlurmUser=slurm
> StateSaveLocation=/site/slurm/save_state/slurmctld
> SwitchType=switch/none
> TaskPlugin=task/none
> TopologyPlugin=topology/none
> UsePAM=0
> CompleteWait=0
> InactiveLimit=0
> KillWait=30
> MinJobAge=300
> SlurmctldTimeout=120
> SlurmdTimeout=300
> Waittime=0
> DefMemPerCPU=6
> FastSchedule=1
> SchedulerType=sched/builtin
> SchedulerPort=7321
> SelectType=select/linear
> PreemptMode=SUSPEND,GANG
> PreemptType=preempt/partition_prio
> SchedulerParameters=max_job_bf=15,interval=20
> PriorityType=priority/basic
> AccountingStorageType=accounting_storage/none
> AccountingStoreJobComment=YES
> ClusterName=xxxxxxxx
> JobCompType=jobcomp/none
> JobAcctGatherFrequency=30
> JobAcctGatherType=jobacct_gather/linux
> SlurmctldDebug=3
> SlurmctldLogFile=/site/slurm/log/slurmctld.log
> SlurmdDebug=3
> SlurmdLogFile=/site/slurm/log/slurmd.log

Reply via email to