[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] condor and torque software in the same cluster.



On Mon, Oct 31, 2011 at 5:55 AM, D.Yilmaz <d.yilmaz@xxxxxx> wrote:
> I wonder if anyone is using condor and torque software in the same cluster.
> I would like to use the both of them in the same cluster and I would like to
> know all your experiences about it.
>
We've been using Condor to soak up idle PBSPro cycles for years. Our
recipe is at:

  https://condor-wiki.cs.wisc.edu/index.cgi/wiki?p=HowToScavengeCycles

The big issue that I currently see is that short PBS jobs
unnecessarily kill Condor jobs. For example, a PBS job that starts on
a 24-core node can kick off up to 24 Condor jobs. If the PBS job dies
immediately, then the Condor work is lost. Depending on local policy,
you may consider suspending jobs for a few minutes before vacating.


-- 
Ben Cotton
Systems Research Engineer
IT Research Systems
Purdue University