[gridengine users] Is there a way to let big jobs through?

Chester Langin cl.research at siu.edu
Mon Jun 5 21:04:49 UTC 2017


Large jobs seem to be stuck in "qw" indefinitely on our HPC cluster while a steady input of small jobs take up the slots.  Is there an easy way to make the small jobs wait in "qw" longer so that the big jobs can proceed to "r" in a timely manner?

Our nodes have 20 cores each.  One user is submitting 8-slot jobs by the thousands with a script.  Each of these jobs lasts approximately 15 minutes.  Another user is submitting a small number of 80-slot jobs.  The 80-slot jobs never make it past "qw" unless I email the first user and ask him to temporarily stop his script which submits the small jobs.  Is there a better way to handle this?


Chet Langin, SIU

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gridengine.org/pipermail/users/attachments/20170605/2746b7b3/attachment.html>

More information about the users mailing list