Re: [slurm-users] Job not running with Resource Reason even though resources appear to be available
While doing more investigation I found an interesting situation. I have a 32 core (2 x 16 core Xeon) node with the 10 RTX cards where all 10 cards have affinity to just one socket (cores 0-15 as shown by 'nvidia-smi topo -m'). The current running jobs on it are using 5 GPUS and 15 cores # scontrol show node=rtx-04 | grep gres CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=15,mem=220G,gres/gpu=5 Checking /sys/fs/cgroup I see these jobs are using cores 0-14 # grep . /sys/fs/cgroup/cpuset/slurm/uid_*/job_*/cpuset.cpus /sys/fs/cgroup/cpuset/slurm/uid_4181545/job_38409/cpuset.cpus:12-14 /sys/fs/cgroup/cpuset/slurm/uid_4181545/job_38670/cpuset.cpus:0-2 /sys/fs/cgroup/cpuset/slurm/uid_4181545/job_38673/cpuset.cpus:3-5 /sys/fs/cgroup/cpuset/slurm/uid_5829/job_49088/cpuset.cpus:9-11 /sys/fs/cgroup/cpuset/slurm/uid_8285/job_49048/cpuset.cpus:6-8 If I submit a job to rtx-04 asking for 1 core and 1 GPU the job runs no problem and it uses core 15. And then if I submit more jobs asking for a GPU they run fine on core 16 and up. Now if I cancel my jobs so I am back to the jobs using 5 GPUS and 15 cores and then submit a job asking for 2 cores and 1 GPU, the job stays in Pending state and refused to run on rtx-04. So before submitting any bug report I decided to upgrade to the latest SLURM version. I upgraded from 20.02.03 to 20.11.3 (with those jobs still running on rtx-04) and now the problem has gone away. I can submit a 2 core and 1 GPU job and it runs immediately. So my problem seems fixed, but in the update I noticed a wierd thing happen. Now SLURM insistes that the Cores in gres.conf must be set to Cores=0-31 even though 'nvidia-smi topo -m' still says 0-15. I decided to just remove the Cores= setting from /etc/slurm/gres.conf So before the update slurmd.log has: [2021-01-26T03:07:45.673] Gres Name=gpu Type=quadro_rtx_8000 Count=1 Index=0 ID=7696487 File=/dev/nvidia0 Cores=0-15 CoreCnt=32 Links=-1,0,0,0,0,0,2,0,0,0 and after the update [2021-01-26T14:31:47.282] Gres Name=gpu Type=quadro_rtx_8000 Count=1 Index=0 ID=7696487 File=/dev/nvidia0 Cores=0-31 CoreCnt=32 Links=-1,0,0,0,0,0,2,0,0,0 This is fine with me as I want SLURM to ignore GPU affinity on these nodes but it is curious. -- Paul Raines (http://help.nmr.mgh.harvard.edu) On Mon, 25 Jan 2021 10:07am, Paul Raines wrote: I tried submitting jobs with --gres-flags=disable-binding but this has not made any difference. Jobs asking for GPUs are still only being run if a core defined in gres.conf for the GPU is free. Basically seems the option is ignored. -- Paul Raines (http://help.nmr.mgh.harvard.edu) On Sun, 24 Jan 2021 11:39am, Paul Raines wrote: Thanks Chris. I think you have identified the issue here or are very close. My gres.conf on the rtx-04 node for example is: AutoDetect=nvml Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia0 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia1 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia2 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia3 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia4 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia5 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia6 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia7 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia8 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia9 Cores=0-15 There are 32 cores (HT is off). But the daughter card that holds all 10 of the RTX8000s connects to only one socket as can be seen from 'nvidia-smi topo -m' Its odd though in that my tests on my identically configured rtx6000 partition did not show that behavior but maybe it is due to just the "random" cores that got assigned to jobs there all having a least one core on the "right" socket. Anyway, how do I turn off this "affinity enforcment" as it is more important that a job run with a GPU on its non-affinity socket than just wait and not run at all? Thanks -- Paul Raines (http://help.nmr.mgh.harvard.edu) On Sat, 23 Jan 2021 3:19pm, Chris Samuel wrote: On Saturday, 23 January 2021 9:54:11 AM PST Paul Raines wrote: Now rtx-08 which has only 4 GPUs seems to always get all 4 uses. But the others seem to always only get half used (except rtx-07 which somehow gets 6 used so another wierd thing). Again if I submit non-GPU jobs, they end up allocating all hte cores/cpus on the nodes just fine. What does your gres.conf look like for these nodes? One thing I've seen in the past is where the core specifications for the GPUs are out of step with the hardware and so Slurm thinks they're on the wrong socket. Then when all the cores in that socket are used up Slurm won't put more GPU jobs on the node without the jobs explicitly asking to not do locality. One thing I've noticed is that in prior to Slurm 20.02 the documentation for gres.conf used to say: # If
Re: [slurm-users] Job not running with Resource Reason even though resources appear to be available
I tried submitting jobs with --gres-flags=disable-binding but this has not made any difference. Jobs asking for GPUs are still only being run if a core defined in gres.conf for the GPU is free. Basically seems the option is ignored. -- Paul Raines (http://help.nmr.mgh.harvard.edu) On Sun, 24 Jan 2021 11:39am, Paul Raines wrote: Thanks Chris. I think you have identified the issue here or are very close. My gres.conf on the rtx-04 node for example is: AutoDetect=nvml Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia0 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia1 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia2 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia3 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia4 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia5 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia6 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia7 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia8 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia9 Cores=0-15 There are 32 cores (HT is off). But the daughter card that holds all 10 of the RTX8000s connects to only one socket as can be seen from 'nvidia-smi topo -m' Its odd though in that my tests on my identically configured rtx6000 partition did not show that behavior but maybe it is due to just the "random" cores that got assigned to jobs there all having a least one core on the "right" socket. Anyway, how do I turn off this "affinity enforcment" as it is more important that a job run with a GPU on its non-affinity socket than just wait and not run at all? Thanks -- Paul Raines (http://help.nmr.mgh.harvard.edu) On Sat, 23 Jan 2021 3:19pm, Chris Samuel wrote: On Saturday, 23 January 2021 9:54:11 AM PST Paul Raines wrote: Now rtx-08 which has only 4 GPUs seems to always get all 4 uses. But the others seem to always only get half used (except rtx-07 which somehow gets 6 used so another wierd thing). Again if I submit non-GPU jobs, they end up allocating all hte cores/cpus on the nodes just fine. What does your gres.conf look like for these nodes? One thing I've seen in the past is where the core specifications for the GPUs are out of step with the hardware and so Slurm thinks they're on the wrong socket. Then when all the cores in that socket are used up Slurm won't put more GPU jobs on the node without the jobs explicitly asking to not do locality. One thing I've noticed is that in prior to Slurm 20.02 the documentation for gres.conf used to say: # If your cores contain multiple threads only the first thread # (processing unit) of each core needs to be listed. but that language is gone from 20.02 and later and the change isn't mentioned in the release notes for 20.02 so I'm not sure what happened there, the only clue is this commit: https://github.com/SchedMD/slurm/commit/ 7461b6ba95bb8ae70b36425f2c7e4961ac35799e#diff- cac030b65a8fc86123176971a94062fafb262cb2b11b3e90d6cc69e353e3bb89 which says "xcpuinfo_abs_to_mac() expects a core list, not a CPU list." Best of luck! Chris -- Chris Samuel : http://www.csamuel.org/ : Berkeley, CA, USA
Re: [slurm-users] Job not running with Resource Reason even though resources appear to be available
Thanks Chris. I think you have identified the issue here or are very close. My gres.conf on the rtx-04 node for example is: AutoDetect=nvml Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia0 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia1 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia2 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia3 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia4 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia5 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia6 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia7 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia8 Cores=0-15 Name=gpu Type=quadro_rtx_8000 File=/dev/nvidia9 Cores=0-15 There are 32 cores (HT is off). But the daughter card that holds all 10 of the RTX8000s connects to only one socket as can be seen from 'nvidia-smi topo -m' Its odd though in that my tests on my identically configured rtx6000 partition did not show that behavior but maybe it is due to just the "random" cores that got assigned to jobs there all having a least one core on the "right" socket. Anyway, how do I turn off this "affinity enforcment" as it is more important that a job run with a GPU on its non-affinity socket than just wait and not run at all? Thanks -- Paul Raines (http://help.nmr.mgh.harvard.edu) On Sat, 23 Jan 2021 3:19pm, Chris Samuel wrote: On Saturday, 23 January 2021 9:54:11 AM PST Paul Raines wrote: Now rtx-08 which has only 4 GPUs seems to always get all 4 uses. But the others seem to always only get half used (except rtx-07 which somehow gets 6 used so another wierd thing). Again if I submit non-GPU jobs, they end up allocating all hte cores/cpus on the nodes just fine. What does your gres.conf look like for these nodes? One thing I've seen in the past is where the core specifications for the GPUs are out of step with the hardware and so Slurm thinks they're on the wrong socket. Then when all the cores in that socket are used up Slurm won't put more GPU jobs on the node without the jobs explicitly asking to not do locality. One thing I've noticed is that in prior to Slurm 20.02 the documentation for gres.conf used to say: # If your cores contain multiple threads only the first thread # (processing unit) of each core needs to be listed. but that language is gone from 20.02 and later and the change isn't mentioned in the release notes for 20.02 so I'm not sure what happened there, the only clue is this commit: https://github.com/SchedMD/slurm/commit/ 7461b6ba95bb8ae70b36425f2c7e4961ac35799e#diff- cac030b65a8fc86123176971a94062fafb262cb2b11b3e90d6cc69e353e3bb89 which says "xcpuinfo_abs_to_mac() expects a core list, not a CPU list." Best of luck! Chris -- Chris Samuel : http://www.csamuel.org/ : Berkeley, CA, USA
Re: [slurm-users] Job not running with Resource Reason even though resources appear to be available
On Saturday, 23 January 2021 9:54:11 AM PST Paul Raines wrote: > Now rtx-08 which has only 4 GPUs seems to always get all 4 uses. > But the others seem to always only get half used (except rtx-07 > which somehow gets 6 used so another wierd thing). > > Again if I submit non-GPU jobs, they end up allocating all hte > cores/cpus on the nodes just fine. What does your gres.conf look like for these nodes? One thing I've seen in the past is where the core specifications for the GPUs are out of step with the hardware and so Slurm thinks they're on the wrong socket. Then when all the cores in that socket are used up Slurm won't put more GPU jobs on the node without the jobs explicitly asking to not do locality. One thing I've noticed is that in prior to Slurm 20.02 the documentation for gres.conf used to say: # If your cores contain multiple threads only the first thread # (processing unit) of each core needs to be listed. but that language is gone from 20.02 and later and the change isn't mentioned in the release notes for 20.02 so I'm not sure what happened there, the only clue is this commit: https://github.com/SchedMD/slurm/commit/ 7461b6ba95bb8ae70b36425f2c7e4961ac35799e#diff- cac030b65a8fc86123176971a94062fafb262cb2b11b3e90d6cc69e353e3bb89 which says "xcpuinfo_abs_to_mac() expects a core list, not a CPU list." Best of luck! Chris -- Chris Samuel : http://www.csamuel.org/ : Berkeley, CA, USA
Re: [slurm-users] Job not running with Resource Reason even though resources appear to be available
Yes, I meant job 38692. Sorry. I am still having the problem. I suspect it has something to do with the GPU configuration as this does not happen on my non-GPU node partitions. Also, if I submit non-GPU jobs to the rtx8000 partition here, they use up all the cores on the nodes just fine. The upshot is on my 10 GPU nodes, I never see more than 6 GPUs in use and jobs just asking for 1 or 2 GPUs are just made to wait in the qeuue. Here is an example. The state of the nodes in rtx8000 queue before I queue jobs: rtx-04 CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=15,mem=120G,gres/gpu=5 rtx-05 CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=15,mem=328G,gres/gpu=5 rtx-06 CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=15,mem=224G,gres/gpu=5 rtx-07 CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=16,mem=232G,gres/gpu=6 rtx-08 CfgTRES=cpu=32,mem=1546000M,billing=81,gres/gpu=4 I then submit 10 jobs. Then the queue for rtx8000 is: NODELISTJOBID PARTITION ST TIME_LIMIT TRES_ALLOC TRES_PER rtx-04 40365 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-04 38676 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-04 38673 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-04 38670 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-04 38409 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-05 40214 rtx8000R 6-10:00:00 cpu=3,mem=128G,node= gpu:1 rtx-05 38677 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-05 38674 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-05 37450 rtx8000R 6-10:00:00 cpu=3,mem=128G,node= gpu:1 rtx-05 37278 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-06 40366 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-06 40364 rtx8000R 6-10:00:00 cpu=3,mem=128G,node= gpu:1 rtx-06 38648 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-06 38646 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-06 37267 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-07 40760 rtx8000R 50:00 cpu=4,mem=32G,node=1 gpu:2 rtx-07 38675 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-07 38672 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-07 38671 rtx8000R 7-00:00:00 cpu=3,mem=24G,node=1 gpu:1 rtx-07 37451 rtx8000R 6-10:00:00 cpu=3,mem=128G,node= gpu:1 rtx-08 40785 rtx8000R 50:00 cpu=4,mem=32G,node=1 gpu:2 rtx-08 40786 rtx8000R 50:00 cpu=4,mem=32G,node=1 gpu:2 (Priorit40794 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 (Priorit40793 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 (Priorit40792 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 (Priorit40791 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 (Priorit40790 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 (Priorit40789 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 (Priorit40788 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 (Resourc40787 rtx8000PD 50:00 cpu=4,mem=32G,node=1 gpu:2 [root@mlsc-head ~]# scontrol show job=40787 JobId=40787 JobName=sjob_5 UserId=raines(5829) GroupId=raines(5829) MCS_label=N/A Priority=19836243 Nice=0 Account=sysadm QOS=normal JobState=PENDING Reason=Resources Dependency=(null) Requeue=1 Restarts=0 BatchFlag=1 Reboot=0 ExitCode=0:0 RunTime=00:00:00 TimeLimit=00:50:00 TimeMin=N/A SubmitTime=2021-01-23T12:37:51 EligibleTime=2021-01-23T12:37:51 AccrueTime=2021-01-23T12:37:51 StartTime=2021-01-23T13:08:52 EndTime=2021-01-23T13:58:52 Deadline=N/A SuspendTime=None SecsPreSuspend=0 LastSchedEval=2021-01-23T12:38:36 Partition=rtx8000 AllocNode:Sid=mlsc-head:1268664 ReqNodeList=(null) ExcNodeList=(null) NodeList=(null) SchedNodeList=rtx-07 NumNodes=1-2 NumCPUs=4 NumTasks=1 CPUs/Task=4 ReqB:S:C:T=0:0:*:* TRES=cpu=4,mem=32G,node=1,billing=11,gres/gpu=2 Socks/Node=* NtasksPerN:B:S:C=0:0:*:1 CoreSpec=* MinCPUsNode=4 MinMemoryNode=32G MinTmpDiskNode=0 Features=(null) DelayBoot=00:00:00 OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null) Command=/autofs/cluster/batch/raines/sjob_5 WorkDir=/autofs/cluster/batch/raines StdErr=/autofs/cluster/batch/raines/sjob_5.err40787 StdIn=/dev/null StdOut=/autofs/cluster/batch/raines/sjob_5.out40787 Power= TresPerJob=gpu:2 MailUser=(null) MailType=NONE [root@mlsc-head ~]# scontrol show node=rtx-04 NodeName=rtx-04 Arch=x86_64 CoresPerSocket=16 CPUAlloc=15 CPUTot=32 CPULoad=18.21 AvailableFeatures=intel,cascade,rtx8000 ActiveFeatures=intel,cascade,rtx8000 Gres=gpu:quadro_rtx_8000:10(S:0) NodeAddr=rtx-04 NodeHostName=rtx-04 Version=20.02.3 OS=Linux 4.18.0-193.28.1.el8_2.x86_64 #1 SMP Thu Oct 22 00:20:22 UTC 2020 RealMemory=1546000
Re: [slurm-users] Job not running with Resource Reason even though resources appear to be available
I think job 38687 *is* being run on the rtx-06 node. I think you mean why job 38692 is not being run on the rtx-06 node (the top prio pending job). I can't see the problem... This (and other info) does seem to indicate that there is enough resource for the extra job: CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=16,mem=143G,gres/gpu=5 If I were debugging this, I'd submit some test jobs that just request resource and sleep, and look for if a node ever allocates more than 16 cores/cpus or 5 gpus. Maybe the answer is in the comprehensive info you posted and someone will see the gem. Not me, sorry. Gareth -Original Message- From: slurm-users On Behalf Of Paul Raines Sent: Friday, 22 January 2021 7:12 AM To: slurm-users@lists.schedmd.com Subject: [slurm-users] Job not running with Resource Reason even though resources appear to be available I am in the beginning of setting up my first SLURM cluster and I am trying to understand why jobs are pending when resources are available These are the pending jobs: # squeue -P --sort=-p,i --states=PD -O "JobID:.12 ,Partition:9 ,StateCompact:2 ,Priority:.12 ,ReasonList" JOBID PARTITION ST PRIORITY NODELIST(REASON) 38692 rtx8000 PD 0.0046530945 (Resources) 38693 rtx8000 PD 0.0046530945 (Priority) 38694 rtx8000 PD 0.0046530906 (Priority) 38695 rtx8000 PD 0.0046530866 (Priority) 38696 rtx8000 PD 0.0046530866 (Priority) 38697 rtx8000 PD 0.208867 (Priority) The job at the top is as follows: Submission command line: sbatch -p rtx8000 -G 1 -c 4 -t 12:00:00 --mem=47G \ -o /cluster/batch/iman/%j.out --wrap='cmd .' # scontrol show job=38692 JobId=38692 JobName=wrap UserId=iman(8084) GroupId=iman(8084) MCS_label=N/A Priority=19989863 Nice=0 Account=imanlab QOS=normal JobState=PENDING Reason=Resources Dependency=(null) Requeue=1 Restarts=0 BatchFlag=1 Reboot=0 ExitCode=0:0 RunTime=00:00:00 TimeLimit=12:00:00 TimeMin=N/A SubmitTime=2021-01-21T13:05:02 EligibleTime=2021-01-21T13:05:02 AccrueTime=2021-01-21T13:05:02 StartTime=2021-01-22T01:05:02 EndTime=2021-01-22T13:05:02 Deadline=N/A SuspendTime=None SecsPreSuspend=0 LastSchedEval=2021-01-21T14:04:32 Partition=rtx8000 AllocNode:Sid=mlsc-head:974529 ReqNodeList=(null) ExcNodeList=(null) NodeList=(null) SchedNodeList=rtx-06 NumNodes=1-1 NumCPUs=4 NumTasks=1 CPUs/Task=4 ReqB:S:C:T=0:0:*:* TRES=cpu=4,mem=47G,node=1,billing=8,gres/gpu=1 Socks/Node=* NtasksPerN:B:S:C=0:0:*:1 CoreSpec=* MinCPUsNode=4 MinMemoryNode=47G MinTmpDiskNode=0 Features=(null) DelayBoot=00:00:00 OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null) Command=(null) WorkDir=/autofs/homes/008/iman StdErr=/cluster/batch/iman/38692.out StdIn=/dev/null StdOut=/cluster/batch/iman/38692.out Power= TresPerJob=gpu:1 MailUser=(null) MailType=NONE This node shows it has enough free resources (cpu,mem,gpus) for the job in the partition # scontrol show node=rtx-06 NodeName=rtx-06 Arch=x86_64 CoresPerSocket=16 CPUAlloc=16 CPUTot=32 CPULoad=5.77 AvailableFeatures=intel,cascade,rtx8000 ActiveFeatures=intel,cascade,rtx8000 Gres=gpu:quadro_rtx_8000:10(S:0) NodeAddr=rtx-06 NodeHostName=rtx-06 Version=20.02.3 OS=Linux 4.18.0-193.28.1.el8_2.x86_64 #1 SMP Thu Oct 22 00:20:22 UTC 2020 RealMemory=1546000 AllocMem=146432 FreeMem=1420366 Sockets=2 Boards=1 MemSpecLimit=2048 State=MIXED ThreadsPerCore=1 TmpDisk=600 Weight=1 Owner=N/A MCS_label=N/A Partitions=rtx8000 BootTime=2020-12-30T10:35:34 SlurmdStartTime=2020-12-30T10:37:21 CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=16,mem=143G,gres/gpu=5 CapWatts=n/a CurrentWatts=0 AveWatts=0 ExtSensorsJoules=n/s ExtSensorsWatts=0 ExtSensorsTemp=n/s # squeue --partition=rtx8000 --states=R -O "NodeList:10 ,JobID:.8 ,Partition:10,tres-alloc,tres-per-job" -w rtx-06 NODELIST JOBID PARTITION TRES_ALLOC TRES_PER_JOB rtx-0638687 rtx8000cpu=4,mem=47G,node=1 gpu:1 rtx-0637267 rtx8000cpu=3,mem=24G,node=1 gpu:1 rtx-0637495 rtx8000cpu=3,mem=24G,node=1 gpu:1 rtx-0638648 rtx8000cpu=3,mem=24G,node=1 gpu:1 rtx-0638646 rtx8000cpu=3,mem=24G,node=1 gpu:1 In case this is needed # scontrol show part=rtx8000 PartitionName=rtx8000 AllowGroups=ALL AllowAccounts=ALL AllowQos=ALL AllocNodes=ALL Default=NO QoS=N/A DefaultTime=04:00:00 DisableRootJobs=NO ExclusiveUser=NO GraceTime=0 Hidden=NO MaxNodes=UNLIMITED MaxTime=7-00:00:00 MinNodes=0 LLN=NO MaxCPUsPerNode=UNLIMITED Nodes=rtx-[04-08] PriorityJobFactor=1 PriorityTier=4 RootOnly=NO ReqResv=NO OverSubscribe=NO OverTimeLimit=NONE PreemptMode=OFF State=UP TotalCPUs=160 TotalNodes=5 SelectTypeParameters=NONE JobDefaults=(null) DefMemPerNode=UNLIMITED
[slurm-users] Job not running with Resource Reason even though resources appear to be available
I am in the beginning of setting up my first SLURM cluster and I am trying to understand why jobs are pending when resources are available These are the pending jobs: # squeue -P --sort=-p,i --states=PD -O "JobID:.12 ,Partition:9 ,StateCompact:2 ,Priority:.12 ,ReasonList" JOBID PARTITION ST PRIORITY NODELIST(REASON) 38692 rtx8000 PD 0.0046530945 (Resources) 38693 rtx8000 PD 0.0046530945 (Priority) 38694 rtx8000 PD 0.0046530906 (Priority) 38695 rtx8000 PD 0.0046530866 (Priority) 38696 rtx8000 PD 0.0046530866 (Priority) 38697 rtx8000 PD 0.208867 (Priority) The job at the top is as follows: Submission command line: sbatch -p rtx8000 -G 1 -c 4 -t 12:00:00 --mem=47G \ -o /cluster/batch/iman/%j.out --wrap='cmd .' # scontrol show job=38692 JobId=38692 JobName=wrap UserId=iman(8084) GroupId=iman(8084) MCS_label=N/A Priority=19989863 Nice=0 Account=imanlab QOS=normal JobState=PENDING Reason=Resources Dependency=(null) Requeue=1 Restarts=0 BatchFlag=1 Reboot=0 ExitCode=0:0 RunTime=00:00:00 TimeLimit=12:00:00 TimeMin=N/A SubmitTime=2021-01-21T13:05:02 EligibleTime=2021-01-21T13:05:02 AccrueTime=2021-01-21T13:05:02 StartTime=2021-01-22T01:05:02 EndTime=2021-01-22T13:05:02 Deadline=N/A SuspendTime=None SecsPreSuspend=0 LastSchedEval=2021-01-21T14:04:32 Partition=rtx8000 AllocNode:Sid=mlsc-head:974529 ReqNodeList=(null) ExcNodeList=(null) NodeList=(null) SchedNodeList=rtx-06 NumNodes=1-1 NumCPUs=4 NumTasks=1 CPUs/Task=4 ReqB:S:C:T=0:0:*:* TRES=cpu=4,mem=47G,node=1,billing=8,gres/gpu=1 Socks/Node=* NtasksPerN:B:S:C=0:0:*:1 CoreSpec=* MinCPUsNode=4 MinMemoryNode=47G MinTmpDiskNode=0 Features=(null) DelayBoot=00:00:00 OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null) Command=(null) WorkDir=/autofs/homes/008/iman StdErr=/cluster/batch/iman/38692.out StdIn=/dev/null StdOut=/cluster/batch/iman/38692.out Power= TresPerJob=gpu:1 MailUser=(null) MailType=NONE This node shows it has enough free resources (cpu,mem,gpus) for the job in the partition # scontrol show node=rtx-06 NodeName=rtx-06 Arch=x86_64 CoresPerSocket=16 CPUAlloc=16 CPUTot=32 CPULoad=5.77 AvailableFeatures=intel,cascade,rtx8000 ActiveFeatures=intel,cascade,rtx8000 Gres=gpu:quadro_rtx_8000:10(S:0) NodeAddr=rtx-06 NodeHostName=rtx-06 Version=20.02.3 OS=Linux 4.18.0-193.28.1.el8_2.x86_64 #1 SMP Thu Oct 22 00:20:22 UTC 2020 RealMemory=1546000 AllocMem=146432 FreeMem=1420366 Sockets=2 Boards=1 MemSpecLimit=2048 State=MIXED ThreadsPerCore=1 TmpDisk=600 Weight=1 Owner=N/A MCS_label=N/A Partitions=rtx8000 BootTime=2020-12-30T10:35:34 SlurmdStartTime=2020-12-30T10:37:21 CfgTRES=cpu=32,mem=1546000M,billing=99,gres/gpu=10 AllocTRES=cpu=16,mem=143G,gres/gpu=5 CapWatts=n/a CurrentWatts=0 AveWatts=0 ExtSensorsJoules=n/s ExtSensorsWatts=0 ExtSensorsTemp=n/s # squeue --partition=rtx8000 --states=R -O "NodeList:10 ,JobID:.8 ,Partition:10,tres-alloc,tres-per-job" -w rtx-06 NODELIST JOBID PARTITION TRES_ALLOC TRES_PER_JOB rtx-0638687 rtx8000cpu=4,mem=47G,node=1 gpu:1 rtx-0637267 rtx8000cpu=3,mem=24G,node=1 gpu:1 rtx-0637495 rtx8000cpu=3,mem=24G,node=1 gpu:1 rtx-0638648 rtx8000cpu=3,mem=24G,node=1 gpu:1 rtx-0638646 rtx8000cpu=3,mem=24G,node=1 gpu:1 In case this is needed # scontrol show part=rtx8000 PartitionName=rtx8000 AllowGroups=ALL AllowAccounts=ALL AllowQos=ALL AllocNodes=ALL Default=NO QoS=N/A DefaultTime=04:00:00 DisableRootJobs=NO ExclusiveUser=NO GraceTime=0 Hidden=NO MaxNodes=UNLIMITED MaxTime=7-00:00:00 MinNodes=0 LLN=NO MaxCPUsPerNode=UNLIMITED Nodes=rtx-[04-08] PriorityJobFactor=1 PriorityTier=4 RootOnly=NO ReqResv=NO OverSubscribe=NO OverTimeLimit=NONE PreemptMode=OFF State=UP TotalCPUs=160 TotalNodes=5 SelectTypeParameters=NONE JobDefaults=(null) DefMemPerNode=UNLIMITED MaxMemPerNode=UNLIMITED TRESBillingWeights=CPU=1.24,Mem=0.02G,Gres/gpu=3.0 Scheduling parameters from slurm.conf are: EnforcePartLimits=ALL LaunchParameters=mem_sort,slurmstepd_memlock_all,test_exec MaxJobCount=30 MaxArraySize=1 DefMemPerCPU=10240 DefCpuPerGPU=1 DefMemPerGPU=10240 GpuFreqDef=medium CompleteWait=0 EpilogMsgTime=300 InactiveLimit=60 KillWait=30 UnkillableStepTimeout=180 ResvOverRun=UNLIMITED MinJobAge=600 Waittime=5 SchedulerType=sched/backfill SelectType=select/cons_tres SelectTypeParameters=CR_Core_Memory,CR_CORE_DEFAULT_DIST_BLOCK,CR_ONE_TASK_PER_CORE PreemptType=preempt/partition_prio PreemptMode=REQUEUE SchedulerParameters=\ default_queue_depth=1500,\ partition_job_depth=10,\ bf_continue,\ bf_interval=30,\ bf_resolution=600,\ bf_window=11520,\ bf_max_job_part=0,\ bf_max_job_user=10,\ bf_max_job_test=10,\ bf_max_job_start=1000,\ bf_ignore_newly_avail_nodes,\ enable_user_top,\ pack_serial_at_end,\ nohold_on_prolog_fail,\