Your bf_window may be too small.  From 'man slurm.conf':

  bf_window=#

         The number of minutes into the future to look when considering
         jobs to schedule.  Higher values result in more overhead and
         less responsiveness.  A value at least as long as the highest
         allowed time limit is generally advisable to prevent job
         starvation.  In order to limit the amount of data managed by
         the backfill scheduler, if the value of bf_window is increased,
         then it is generally advisable to also increase bf_resolution.
         This option applies only to SchedulerType=sched/backfill.
         Default: 1440 (1 day), Min: 1, Max: 43200 (30 days).

So since we have a 5 day option should bf_window=7200? What should bf_resolution be set to then?

But how does this affect/improve wait times?

 

>  On Tue, Jun 4, 2024 at 4:13 PM Ryan Novosielski <novosirj@rutgers.edu> wrote:
>
>  This is relatively true of my system as well, and I believe it’s that the backfill schedule is slower than the main scheduler.
>
>  --
>  #BlackLivesMatter
>  ____
>  || \\UTGERS,     |---------------------------*O*---------------------------
>  ||_// the State  |         Ryan Novosielski - novosirj@rutgers.edu
>  || \\ University | Sr. Technologist - 973/972.0922 (2x0922) ~*~ RBHS Campus
>  ||  \\    of NJ  | Office of Advanced Research Computing - MSB A555B, Newark
>       `'
>
>  On Jun 4, 2024, at 16:03, Robert Kudyba via slurm-users <slurm-users@lists.schedmd.com> wrote:
>
>  At the moment we have 2 nodes that are having long wait times. Generally this is when the nodes are fully allocated. What would be the other
>  reasons if there is still enough available memory and CPU available, that a job would take so long? Slurm version is  23.02.4 via Bright
>  Computing. Note the compute nodes have hyperthreading enabled but that should be irrelevant. Is there a way to determine what else could
>  be holding jobs up?
>
>  srun --pty  -t 0-01:00:00 --nodelist=node001 --gres=gpu:1 -A ourts -p short /bin/bash
>  srun: job 672204 queued and waiting for resources
>
>   scontrol show node node001
>  NodeName=m001 Arch=x86_64 CoresPerSocket=48
>     CPUAlloc=24 CPUEfctv=192 CPUTot=192 CPULoad=20.37
>     AvailableFeatures=location=local
>     ActiveFeatures=location=local
>     Gres=gpu:A6000:8
>     NodeAddr=node001 NodeHostName=node001 Version=23.02.4
>     OS=Linux 5.14.0-70.13.1.el9_0.x86_64 #1 SMP PREEMPT Thu Apr 14 12:42:38 EDT 2022
>     RealMemory=1031883 AllocMem=1028096 FreeMem=222528 Sockets=2 Boards=1
>     State=MIXED ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
>     Partitions=ours,short
>     BootTime=2024-04-29T16:18:30 SlurmdStartTime=2024-05-18T16:48:11
>     LastBusyTime=2024-06-03T10:49:49 ResumeAfterTime=None
>     CfgTRES=cpu=192,mem=1031883M,billing=192,gres/gpu=8
>     AllocTRES=cpu=24,mem=1004G,gres/gpu=2,gres/gpu:a6000=2
>     CapWatts=n/a
>     CurrentWatts=0 AveWatts=0
>     ExtSensorsJoules=n/s ExtSensorsWatts=0 ExtSensorsTemp=n/s
>
>  grep 672204 /var/log/slurmctld
>  [2024-06-04T15:50:35.627] sched: _slurm_rpc_allocate_resources JobId=672204 NodeList=(null) usec=852
>
>  --
>  slurm-users mailing list -- slurm-users@lists.schedmd.com
>  To unsubscribe send an email to slurm-users-leave@lists.schedmd.com
--
Dr. Loris Bennett (Herr/Mr)
FUB-IT (ex-ZEDAT), Freie Universität Berlin