That is exactly what the overcommit option is designed to do. 

On January 29, 2014 9:22:39 AM EST, Ulf Markwardt <[email protected]> 
wrote:
>Dear developers, 
>
>when I do 
>
>salloc -p sandy --overcommit --ntasks-per-node=16 --ntasks-per-core=2
>-t 10 
>
>all tasks run on a single core:
>
>root@taurusadmin1:~> scontrol -d show job 4576028
>JobId=4576028 Name=bash
>   UserId=mark(19423) GroupId=hpcsupport(50245)
>   Priority=89609 Account=hpcsupport QOS=normal
>   JobState=RUNNING Reason=None Dependency=(null)
>   Requeue=1 Restarts=0 BatchFlag=0 ExitCode=0:0
>   DerivedExitCode=0:0
>   RunTime=00:00:06 TimeLimit=00:10:00 TimeMin=N/A
>   SubmitTime=2014-01-29T15:15:40 EligibleTime=2014-01-29T15:15:40
>   StartTime=2014-01-29T15:15:40 EndTime=2014-01-29T15:25:40
>   PreemptTime=None SuspendTime=None SecsPreSuspend=0
>   Partition=sandy AllocNode:Sid=tauruslogin1:31534
>   ReqNodeList=(null) ExcNodeList=(null)
>   NodeList=taurusi1108
>   BatchHost=taurusi1108
>   NumNodes=1 NumCPUs=1 CPUs/Task=1 ReqS:C:T=*:*:*
>     Nodes=taurusi1108 CPU_IDs=15 Mem=0
>   MinCPUsNode=1 MinMemoryNode=0 MinTmpDiskNode=0
>   Features=(null) Gres=(null) Reservation=(null)
>   Shared=OK Contiguous=0 Licenses=(null) Network=(null)
>   Command=(null)
>   WorkDir=/home/h3/mark
>
>(We have even disabled our own plugins - it did not help.)
>
>slurm.conf:
>SelectType=select/cons_res
>SelectTypeParameters=CR_Core_Memory,CR_CORE_DEFAULT_DIST_BLOCK,CR_ONE_TASK_PER_CORE,CR_ALLOCATE_FULL_SOCKET
>
>Thank you for your help,
>Ulf
>-- 
>___________________________________________________________________
>Dr. Ulf Markwardt
>
>Technische Universität Dresden
>Center for Information Services and High Performance Computing (ZIH)
>01062 Dresden, Germany
>
>Phone: (+49) 351/463-33640      WWW:  http://www.tu-dresden.de/zih

-- 
Sent from my Android phone with K-9 Mail. Please excuse my brevity.

Reply via email to