[slurm-users] GPU Allocation does not limit number of available GPUs in job

Dominik Baack dominik.baack at cs.uni-dortmund.de
Thu Oct 27 17:03:57 UTC 2022


Hi,

yes ContrainDevices is set:

###
# Slurm cgroup support configuration file
###
CgroupAutomount=yes
#
#CgroupMountpoint="/sys/fs/cgroup"
ConstrainCores=yes
ConstrainDevices=yes
ConstrainRAMSpace=yes
#
#

I attached the slurm configuration file as well

Cheers
Dominik

Am 27.10.2022 um 17:57 schrieb Sean Maxwell:
> Hi Dominik,
>
> Do you have ConstrainDevices=yes set in your cgroup.conf?
>
> Best,
>
> -Sean
>
> On Thu, Oct 27, 2022 at 11:49 AM Dominik Baack 
> <dominik.baack at cs.uni-dortmund.de> wrote:
>
>     Hi,
>
>     We are in the process of setting up SLURM on some DGX A100 nodes . We
>     are experiencing the problem that all GPUs are available for
>     users, even
>     for jobs where only one should be assigned.
>
>     It seems the requirement is forwarded correctly to the node, at least
>     CUDA_VISIBLE_DEVICES is set to the correct id only discarded by
>     the rest
>     of the system.
>
>     Cheers
>     Dominik Baack
>
>     Example:
>
>     baack at gwkilab:~$ srun --gpus=1 nvidia-smi
>     Thu Oct 27 17:39:04 2022
>     +-----------------------------------------------------------------------------+
>     | NVIDIA-SMI 470.141.03   Driver Version: 470.141.03   CUDA Version:
>     11.4     |
>     |-------------------------------+----------------------+----------------------+
>     | GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile
>     Uncorr. ECC |
>     | Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util
>     Compute M. |
>     |                               | |               MIG M. |
>     |===============================+======================+======================|
>     |   0  NVIDIA A100-SXM...  On   | 00000000:07:00.0 Off
>     |                    0 |
>     | N/A   28C    P0    52W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>     |   1  NVIDIA A100-SXM...  On   | 00000000:0F:00.0 Off
>     |                    0 |
>     | N/A   28C    P0    51W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>     |   2  NVIDIA A100-SXM...  On   | 00000000:47:00.0 Off
>     |                    0 |
>     | N/A   28C    P0    52W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>     |   3  NVIDIA A100-SXM...  On   | 00000000:4E:00.0 Off
>     |                    0 |
>     | N/A   29C    P0    54W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>     |   4  NVIDIA A100-SXM...  On   | 00000000:87:00.0 Off
>     |                    0 |
>     | N/A   34C    P0    57W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>     |   5  NVIDIA A100-SXM...  On   | 00000000:90:00.0 Off
>     |                    0 |
>     | N/A   31C    P0    55W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>     |   6  NVIDIA A100-SXM...  On   | 00000000:B7:00.0 Off
>     |                    0 |
>     | N/A   31C    P0    51W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>     |   7  NVIDIA A100-SXM...  On   | 00000000:BD:00.0 Off
>     |                    0 |
>     | N/A   32C    P0    52W / 400W |      0MiB / 40536MiB | 0%     
>     Default |
>     |                               | |             Disabled |
>     +-------------------------------+----------------------+----------------------+
>
>     +-----------------------------------------------------------------------------+
>     | Processes: |
>     |  GPU   GI   CI        PID   Type   Process name GPU Memory |
>     |        ID   ID Usage      |
>     |=============================================================================|
>     |  No running processes
>     found                                                 |
>     +-----------------------------------------------------------------------------+
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20221027/807a527f/attachment.htm>
-------------- next part --------------
# slurm.conf file generated by configurator.html.
# Put this file on all nodes of your cluster.
# See the slurm.conf man page for more information.
#
ClusterName=ML2R
SlurmctldHost=gwkilab
#
#DisableRootJobs=NO
#EnforcePartLimits=NO
#Epilog=/etc/slurm/epilog.d/*
#EpilogSlurmctld=
#FirstJobId=1
#MaxJobId=67043328
GresTypes=gpu
#GroupUpdateForce=0
#GroupUpdateTime=600
#JobFileAppend=0
#JobRequeue=1
#JobSubmitPlugins=lua
#KillOnBadExit=0
#LaunchType=launch/slurm
#Licenses=foo*4,bar
#MailProg=/bin/mail
#MaxJobCount=10000
#MaxStepCount=40000
#MaxTasksPerNode=512
MpiDefault=none
#MpiParams=ports=#-#
#PluginDir=
PlugStackConfig=/etc/slurm/plugstack.conf
#PrivateData=jobs
ProctrackType=proctrack/cgroup
Prolog=/etc/slurm/prolog.d/*
#PrologFlags=
#PrologSlurmctld=
#PropagatePrioProcess=0
#PropagateResourceLimits=
#PropagateResourceLimitsExcept=
#RebootProgram=
ReturnToService=0
SlurmctldPidFile=/run/slurm/slurmctld.pid
SlurmctldPort=6817
SlurmdPidFile=/run/slurm/slurmd.pid
SlurmdPort=6818
SlurmdSpoolDir=/var/spool/slurmd
SlurmUser=slurm
#SlurmdUser=root
#SrunEpilog=
#SrunProlog=
SrunPortRange=60001-63000
StateSaveLocation=/var/spool/slurmctld
SwitchType=switch/none
#TaskEpilog=
TaskPlugin=task/affinity
#TaskProlog=
#TopologyPlugin=topology/tree
#TmpFS=/tmp
#TrackWCKey=no
#TreeWidth=
#UnkillableStepProgram=
#UsePAM=0
#
#
# TIMERS
#BatchStartTimeout=10
#CompleteWait=0
#EpilogMsgTime=2000
#GetEnvTimeout=2
#HealthCheckInterval=0
#HealthCheckProgram=
InactiveLimit=0
KillWait=30
MessageTimeout=60
#ResvOverRun=0
MinJobAge=300
#OverTimeLimit=0
SlurmctldTimeout=120
SlurmdTimeout=300
#UnkillableStepTimeout=60
#VSizeFactor=0
Waittime=0
#
#
# SCHEDULING
#DefMemPerCPU=0
#MaxMemPerCPU=0
#SchedulerTimeSlice=30
SchedulerType=sched/builtin
SelectType=select/cons_tres
SelectTypeParameters=CR_CPU_Memory
#
#
# JOB PRIORITY
#PriorityFlags=
#PriorityType=priority/basic
#PriorityDecayHalfLife=
#PriorityCalcPeriod=
#PriorityFavorSmall=
#PriorityMaxAge=
#PriorityUsageResetPeriod=
#PriorityWeightAge=
#PriorityWeightFairshare=
#PriorityWeightJobSize=
#PriorityWeightPartition=
#PriorityWeightQOS=
#
#
# LOGGING AND ACCOUNTING
#AccountingStorageEnforce=0
#AccountingStorageHost=
#AccountingStoragePass=
#AccountingStoragePort=
AccountingStorageType=accounting_storage/none
#AccountingStorageUser=
#AccountingStoreFlags=
#JobCompHost=
#JobCompLoc=
#JobCompPass=
#JobCompPort=
JobCompType=jobcomp/none
#JobCompUser=
#JobContainerType=job_container/none
JobAcctGatherFrequency=30
JobAcctGatherType=jobacct_gather/none
SlurmctldDebug=info
SlurmctldLogFile=/var/log/slurmctld.log
SlurmdDebug=info
SlurmdLogFile=/var/log/slurmd.log
#SlurmSchedLogFile=
#SlurmSchedLogLevel=
#DebugFlags=
#
#
# POWER SAVE SUPPORT FOR IDLE NODES (optional)
#SuspendProgram=
#ResumeProgram=
#SuspendTimeout=
#ResumeTimeout=
#ResumeRate=
#SuspendExcNodes=
#SuspendExcParts=
#SuspendRate=
#SuspendTime=
#
#
# COMPUTE NODES
NodeName=ml2ran01 NodeAddr=ml2ran01e0 CPUs=256 Sockets=2 RealMemory=1031875 CoresPerSocket=64 ThreadsPerCore=2 Gres=gpu:nvidia_a100-sxm4-40gb:8 State=UNKNOWN
NodeName=ml2ran03 NodeAddr=ml2ran03e0 CPUs=256 Sockets=2 RealMemory=1031875 CoresPerSocket=64 ThreadsPerCore=2 Gres=gpu:nvidia_a100-sxm4-40gb:8 State=UNKNOWN
NodeName=ml2ran05 NodeAddr=ml2ran05e0 CPUs=256 Sockets=2 RealMemory=1031875 CoresPerSocket=64 ThreadsPerCore=2 Gres=gpu:nvidia_a100-sxm4-40gb:8 State=UNKNOWN


More information about the slurm-users mailing list