Hi,
I am trying to understand backfill chunking in Maui + Torque
environment. From what I understand, the configuration will allow Maui
to accumulate resources freed by jobs, for a configured time, before
allocating them for other backfill jobs.
In my Maui.cfg, I set up the following:
BFCHUNKDURATION 00:05:00
BFCHUNKSIZE 5
The default walltime for torque queues is set to a large value (several
thousand hours). The rest of the Maui configuration is default
configuration.
Now, on a 14 node cluster, I did the following:
echo "sleep 3600" | qsub -l nodes=7
echo "sleep 3600" | qsub -l nodes=14 (This got queued, and got a
reservation)
echo "sleep 3600" | qsub -l nodes=3 -l walltime=0:30:00
echo "sleep 3600" | qsub -l nodes=4 -l walltime=0:30:00 (Both these jobs
ran)
echo "sleep 3600" | qsub -l nodes=7 -l walltime=0:30:00 (This got queued)
echo "sleep 3600" | qsub -l nodes=3 -l walltime=0:30:00 (This also got
queued)
Now, when I deleted the running 3 node job using qdel, the last 3 node
job immediately ran.
Shouldn't Maui have waited for BFCHUNKDURATION time before running this
? Can you please help me understand whether backfill chunking should
have worked here ?
Thanks
hemanth
_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers
