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

Re: [HTCondor-users] Running short-lived jobs on Condor

It sure looks like most of the slots *never* see action. The same slots get used over and over while half the slots sit there permanently unclaimed. On a totally idle cluster with far more jobs than slots, I would expect to see all the slots at least changing state from time to time.

From: HTCondor-users [htcondor-users-bounces@xxxxxxxxxxx] on behalf of Dimitri Maziuk [dmaziuk@xxxxxxxxxxxxx]
Sent: Thursday, June 18, 2015 2:54 PM
To: htcondor-users@xxxxxxxxxxx
Subject: Re: [HTCondor-users] Running short-lived jobs on Condor

On 06/18/2015 01:23 PM, Rowe, Thomas wrote:

> What can I do to make it so that short running jobs don't result in
> a mostly idle cluster?

IME part of the issue is what you see: jobs are running faster than
condor_status updates its display.

In my case the short ones are post-processing the results of the
long-running parent so I have the dag running them breadth-first and
that mostly evens it out.

Dimitri Maziuk
BioMagResBank, UW-Madison -- http://www.bmrb.wisc.edu