[slurm-users] slurm-users Digest, Vol 60, Issue 20

Hemanta Sahu hemantaku.sahu at gmail.com
Wed Oct 19 11:49:43 UTC 2022


  I set the  "billing" parameter instead of "cpu" parameter for
"GrpTRESMins"  and it is working as expected.

>>
[testfac3 at login04 ~]$ sacctmgr  modify user name=testuser100
Account=testfac3_imf set GrpTRESMins=billing=0

[testuser100 at login04 testuser100]$ scontrol show assoc users=testuser100
accounts=testfac3_imf flags=assoc|grep GrpTRESMins

GrpTRESMins=cpu=0(92),mem=N(398787),energy=N(0),node=N(9),billing=0(185),fs/disk=N(0),vmem=N(0),pages=N(0),gres/gpu=N(0)
[testuser100 at login04 testuser100]$ vi testjob.sh
[testuser100 at login04 testuser100]$ sbatch testjob.sh
sbatch: error: AssocGrpBillingMinutes
sbatch: error: Batch job submission failed: Job violates accounting/QOS
policy (job submit limit, user's size and/or time limits)
[testuser100 at login04 testuser100]$
>>


Regards
Hemanta

On Wed, Oct 19, 2022 at 5:12 PM Hemanta Sahu <hemantaku.sahu at gmail.com>
wrote:

> Thanks Ole for pointing to necessary documents.
>
> So Do you mean to say I need to configure the "billing" parameter instead
> of "GrpTRESMins" .
>
> >>
> [testuser100 at login04 3072]$ scontrol show assoc users=testuser100
> accounts=testfac3_imf flags=assoc|grep GrpTRESMins
>
> GrpTRESMins=cpu=0(88),mem=N(380538),energy=N(0),node=N(9),billing=N(176),fs/disk=N(0),vmem=N(0),pages=N(0),gres/gpu=N(0)
> >>
>
> Regards
> Hemanta
>
> On Wed, Oct 19, 2022 at 2:01 PM <slurm-users-request at lists.schedmd.com>
> wrote:
>
>> Send slurm-users mailing list submissions to
>>         slurm-users at lists.schedmd.com
>>
>> To subscribe or unsubscribe via the World Wide Web, visit
>>         https://lists.schedmd.com/cgi-bin/mailman/listinfo/slurm-users
>> or, via email, send a message with subject or body 'help' to
>>         slurm-users-request at lists.schedmd.com
>>
>> You can reach the person managing the list at
>>         slurm-users-owner at lists.schedmd.com
>>
>> When replying, please edit your Subject line so it is more specific
>> than "Re: Contents of slurm-users digest..."
>>
>>
>> Today's Topics:
>>
>>    1. Re: slurm-users Digest, Vol 60, Issue 19 (Hemanta Sahu)
>>    2. Re: slurm-users Digest, Vol 60, Issue 19 (Ole Holm Nielsen)
>>    3. GPU utilization of running jobs (Vecerka Daniel)
>>
>>
>> ----------------------------------------------------------------------
>>
>> Message: 1
>> Date: Tue, 18 Oct 2022 19:39:39 +0530
>> From: Hemanta Sahu <hemantaku.sahu at gmail.com>
>> To: slurm-users at lists.schedmd.com
>> Subject: Re: [slurm-users] slurm-users Digest, Vol 60, Issue 19
>> Message-ID:
>>         <CAH5Hmwf4jRXYFYbwwBe=+
>> aRtrzWo22ycuy_akGch06W_OPo5qg at mail.gmail.com>
>> Content-Type: text/plain; charset="utf-8"
>>
>> Hi Ole,
>>
>> I confirm that the Slurm database has been configured and "
>> AccountingStorageEnforce" parameters have been set.
>>
>> >>
>> [admin2 at login01 ~]$  scontrol show config | grep AccountingStorageEnforce
>> AccountingStorageEnforce = associations,limits,qos,safe
>> >>
>>
>> My Question : If I have multiple users under a slurm Account and  I
>> want to limit user xxx to  max 1000 CPU core-minutes and user yyy  to
>> max 2000 CPU core-minutes for all past,present and future jobs , what
>> would be the best way to achieve  this ?
>>
>>
>> Thanks
>>
>> Hemanta
>>
>>
>> On Tue, Oct 18, 2022 at 5:31 PM <slurm-users-request at lists.schedmd.com>
>> wrote:
>>
>> > Send slurm-users mailing list submissions to
>> >         slurm-users at lists.schedmd.com
>> >
>> > To subscribe or unsubscribe via the World Wide Web, visit
>> >         https://lists.schedmd.com/cgi-bin/mailman/listinfo/slurm-users
>> > or, via email, send a message with subject or body 'help' to
>> >         slurm-users-request at lists.schedmd.com
>> >
>> > You can reach the person managing the list at
>> >         slurm-users-owner at lists.schedmd.com
>> >
>> > When replying, please edit your Subject line so it is more specific
>> > than "Re: Contents of slurm-users digest..."
>> >
>> >
>> > Today's Topics:
>> >
>> >    1. How to implement resource restriction for diffrent slurm
>> >       users under same slurm account (Hemanta Sahu)
>> >    2. Re: How to implement resource restriction for diffrent slurm
>> >       users under same slurm account (Ole Holm Nielsen)
>> >
>> >
>> > ----------------------------------------------------------------------
>> >
>> > Message: 1
>> > Date: Mon, 17 Oct 2022 20:21:59 +0530
>> > From: Hemanta Sahu <hemantaku.sahu at gmail.com>
>> > To: slurm-users at lists.schedmd.com
>> > Subject: [slurm-users] How to implement resource restriction for
>> >         diffrent slurm users under same slurm account
>> > Message-ID:
>> >         <CAH5HmweLsQ7uUkH=D=
>> > 6XqbzrxLXsGDvPftSBt6T5snqbqQckXQ at mail.gmail.com>
>> > Content-Type: text/plain; charset="utf-8"
>> >
>> > Dear All,
>> >
>> >      I want to implement resource restriction for different slurm users
>> > under
>> > the same slurm account by setting  "GrpTRESMins" Flag. For testing
>> purposes
>> > I set the "GrpTRESMins=cpu=0"  and submitted the Job.
>> >
>> >   I expect the job submission should fail but it is not happening. The
>> jobs
>> > are still going to Q and running.
>> >
>> > Kindly help me if I am missing anything obvious. Command outputs given
>> > below for reference.
>> >
>> >
>> > >>
>> > [testfac3 at login04 export_bin]$ sacctmgr  modify user name=testuser100
>> > Account=testfac3_imf set GrpTRESMins=cpu=0
>> >  Modified user associations...
>> >   C = param-shakti A = testfac3_imf         U = testuser100
>> > Would you like to commit changes? (You have 30 seconds to decide)
>> > (N/y): y
>> >
>> > [testuser100 at login04 ~]$ sacctmgr show assoc where Account=testfac3_imf
>> > user=testuser100 format=Account%15,User%15,GrpTRESMins,QOS%30
>> >         Account            User   GrpTRESMins
>> > QOS
>> > --------------- --------------- -------------
>> > ------------------------------
>> >    testfac3_imf     testuser100         cpu=0
>> >  testfac3_imf
>> >
>> > [testuser100 at login04 testuser100]$ sacctmgr show qos sacctmgr show qos
>> > testfac3_imf
>> >
>> >
>> format=Name%20,MaxWall,Flags%20,GrpTRESMins%20,MaxSubmitJobsPerUser,MaxSubmitJobsPeraccount,GrpTRESRunMin,Priority
>> >                 Name     MaxWall                Flags
>> GrpTRESMins
>> > MaxSubmitPU MaxSubmitPA GrpTRESRunMin   Priority
>> > -------------------- ----------- --------------------
>> --------------------
>> > ----------- ----------- ------------- ----------
>> >         testfac3_imf  3-00:00:00  DenyOnLimit,NoDecay
>> cpu=210000000
>> >         100         500                    10000
>> > [testuser100 at login04 testuser100]$
>> >
>> > [testuser100 at login04 testuser100]$ scontrol show job 949622|grep
>> JobState
>> >    JobState=COMPLETED Reason=None Dependency=(null)
>> > [testuser100 at login04 testuser100]$
>> >
>> >
>> > [testuser100 at login04 testuser100]$ cat testjob.sh
>> > #!/bin/bash
>> > #SBATCH -J testjob                # name of the job
>> > #SBATCH -p standard               # name of the partition: available
>> > options "standard,standard-low,gpu,gpu-low,hm"
>> > #SBATCH -n 2                      # no of processes
>> > #SBATCH -q testfac3_imf
>> > #SBATCH -A testfac3_imf
>> > #SBATCH -t 01:00:00               # walltime in HH:MM:SS, Max value
>> > 72:00:00
>> > #list of modules you want to use, for example
>> > module load compiler/intel-mpi/mpi-2020-v4 compiler/intel/2020.4.304
>> >
>> > #name of the executable
>> > exe="uname -n"
>> >
>> > #run the application
>> > mpirun -n $SLURM_NTASKS $exe
>> >
>> > [testuser100 at login04 testuser100]$ sbatch testjob.sh
>> > Submitted batch job 949622
>> >
>> > [testuser100 at login04 testuser100]$ squeue
>> >              JOBID PARTITION     NAME     USER ST       TIME  NODES
>> > NODELIST(REASON)
>> >             949622  standard testjob- testuser  R       0:04      2
>> > cn[304-305]
>> > >>
>> >
>> > Thanks in advance
>> >
>> > Best Regards
>> > Hemanta
>> > -------------- next part --------------
>> > An HTML attachment was scrubbed...
>> > URL: <
>> >
>> http://lists.schedmd.com/pipermail/slurm-users/attachments/20221017/7b137ddb/attachment-0001.htm
>> > >
>> >
>> > ------------------------------
>> >
>> > Message: 2
>> > Date: Tue, 18 Oct 2022 07:41:29 +0200
>> > From: Ole Holm Nielsen <Ole.H.Nielsen at fysik.dtu.dk>
>> > To: <slurm-users at lists.schedmd.com>
>> > Subject: Re: [slurm-users] How to implement resource restriction for
>> >         diffrent slurm users under same slurm account
>> > Message-ID: <e780481e-9314-9a3e-790b-40c08e3f66bd at fysik.dtu.dk>
>> > Content-Type: text/plain; charset="UTF-8"; format=flowed
>> >
>> > On 10/17/22 16:51, Hemanta Sahu wrote:
>> > >  ? ? ?I want to implement resource restriction for different slurm
>> users
>> > > under
>> > > the same slurm account by setting ?"GrpTRESMins" Flag. For testing
>> > > purposes I set the "GrpTRESMins=cpu=0" ?and submitted the Job.
>> > >
>> > >  ? I expect the job submission should fail but it is not happening.
>> The
>> > > jobs are still going to Q and running.
>> > >
>> > > Kindly help me if I am missing anything obvious. Command outputs given
>> > > below for reference.
>> >
>> > Job submission should not fail due to resource limits.
>> >
>> > Read the slurm.conf manual page to make sure you have set this parameter
>> > correctly, for example:
>> >
>> > $ scontrol show config | grep AccountingStorageEnforce
>> > AccountingStorageEnforce = associations,limits,qos,safe
>> >
>> > You should also read this documentation:
>> > https://slurm.schedmd.com/resource_limits.html
>> >
>> > I assume that you have configured a Slurm database?
>> >
>> > /Ole
>> >
>> >
>> >
>> > End of slurm-users Digest, Vol 60, Issue 19
>> > *******************************************
>> >
>> -------------- next part --------------
>> An HTML attachment was scrubbed...
>> URL: <
>> http://lists.schedmd.com/pipermail/slurm-users/attachments/20221018/662f720c/attachment-0001.htm
>> >
>>
>> ------------------------------
>>
>> Message: 2
>> Date: Wed, 19 Oct 2022 08:22:22 +0200
>> From: Ole Holm Nielsen <Ole.H.Nielsen at fysik.dtu.dk>
>> To: <slurm-users at lists.schedmd.com>
>> Subject: Re: [slurm-users] slurm-users Digest, Vol 60, Issue 19
>> Message-ID: <76e0cc7c-8ec0-140f-beef-d82498ad63ef at fysik.dtu.dk>
>> Content-Type: text/plain; charset="UTF-8"; format=flowed
>>
>> Hi Hemanta,
>>
>> On 10/18/22 16:09, Hemanta Sahu wrote:
>> > Hi Ole,
>> >
>> > I confirm that the Slurm database has been configured and "
>> > AccountingStorageEnforce"?parameters have been set.
>> >
>> >  >>
>> > [admin2 at login01 ~]$ ?scontrol show config | grep
>> AccountingStorageEnforce
>> > AccountingStorageEnforce = associations,limits,qos,safe
>> >  >>
>> >
>> > My Question : If I have multiple users under a slurm Account and I want
>> to
>> > limit user xxx to max 1000 CPU  core-minutes and user yyy to max 2000
>> CPUcore-minutes for all past,present
>> > and future jobs , what would be the best way to achieve this ?
>>
>> Slurm works with a "Associations", see some information in
>> https://wiki.fysik.dtu.dk/Niflheim_system/Slurm_accounting/#associations
>>
>> Under each account in the hierarchy you must define the different users
>> that you want.  They must have different associations (under the same
>> account), and you can give them different limits.  See this nice
>> tutorial:
>> https://slurm.schedmd.com/slurm_ug_2012/SUG_Oct2012_DBD.pdf
>>
>> You can only put limits on future jobs, not on past and running jobs.  If
>> you want to make limits for all of history, you need to setup "banking",
>> see https://slurm.schedmd.com/SLUG19/Banking.pdf
>>
>> /Ole
>>
>>
>>
>>
>>
>> ------------------------------
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20221019/990a31d9/attachment-0001.htm>


More information about the slurm-users mailing list