[Mauiusers] Backfill and node reservation
Denis
denismpa at gmail.com
Mon Nov 15 10:37:05 MST 2010
2010/11/15 Arnau Bria <arnaubria at pic.es>
> Hi Denis and all,
>
>
> I tried to reproduce it and here are results:
>
> # showconfig |grep -i backfill
> BACKFILLPOLICY[0] NONE
> BACKFILLDEPTH[0] 0
> BACKFILLPROCFACTOR[0] 0
> BACKFILLMAXSCHEDULES[0] 10000
> BACKFILLMETRIC[0] PROCS
>
>
> $ pbsnodes td115.pic.es
> td115.pic.es
> state = offline
> np = 8
> properties = slc5_x64
> ntype = cluster
> jobs = 0/13890645.pbs02.pic.es, 1/13892037.pbs02.pic.es, 2/
> 13894222.pbs02.pic.es, 3/13894254.pbs02.pic.es, 4/13892138.pbs02.pic.es,
> 5/13891930.pbs02.pic.es, 6/13892881.pbs02.pic.es
>
>
>
> $ qsub -q short -l nodes=td115.pic.es:ppn=8 -N backfill_test sleep.sh
> 13894790.pbs02.pic.es
>
> $pbsnodes -c td115.pic.es
>
> > what does a diagnose -p report?
> > Is it possible that the jobs which are running before your highest
> > priority job are not being backfilled but having a higher priority
> > instead due to the weights of the other metrics?
> > I see that the CREDWEIGHT is set to 1 while QOS for example is set to
> > 100.
>
>
> Job PRIORITY* Cred( User:Group:Class) FS(
> User:Group: QOS)
> Weights -------- 1( 1: 1: 1) 1( 2:
> 10: 100)
>
> 13894790 100000 100.0(10000: 0.0: 0.0) 0.0( 0.0:
> 0.0: 0.0)
> 13894957 -134 0.0( 0.0: 0.0: 0.0) 100.0( 0.0:
> -0.3:-133.)
> 13894958 -134 0.0( 0.0: 0.0: 0.0) 100.0( 0.0:
> -0.3:-133.)
> [...]
>
> **** my job is first.
> **** farm is at 99,9%, only that slot is free.
>
> # pbsnodes td115.pic.es
> td115.pic.es
> state = job-exclusive
> np = 8
> properties = slc5_x64
> ntype = cluster
> jobs = 0/13890645.pbs02.pic.es, 1/13892037.pbs02.pic.es, 2/
> 13894222.pbs02.pic.es, 3/13894254.pbs02.pic.es, 4/13892138.pbs02.pic.es,
> 5/13891930.pbs02.pic.es, 6/13892881.pbs02.pic.es, 7/13894957.pbs02.pic.es
>
>
> >
> > Also there are some groups with priority really high ( 100000)
> My group.
>
>
> Does it help in any way?
>
> Well...weird. Right now I dont have any other suggestion.
What if you increase the loglevel and try to check what maui is doing behind
the scene?
>
> Cheers,
> Arnau
> _______________________________________________
> mauiusers mailing list
> mauiusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/mauiusers
>
--
Denis Anjos,
www.versatushpc.com.br
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/mauiusers/attachments/20101115/4b696536/attachment.html
More information about the mauiusers
mailing list