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

Re: [Condor-users] Priority issue



Nicolas,

> I started a first job (72), which is made of about 150 queued jobs.
> Then I later started a second one (74), which I need first.  So,
> once started, I modified the 74's priority with :
> condor_prio -p 500 74
> I also modified the 72's priority to -15.
> 
> Now my problem is that only one of the 74 job runs and other CPUs
> are used by 72. Even when a 72 job finishes, if a 74 is running, it
> doesn't launch any new 74.

As far as I know, this is the expected behavior.  The job priority is
only of concern when a new match is made, not when an existing match
is being reused.

When an existing match is being reused, the next best match from the
same cluster is used before looking for jobs in another cluster.

If you want to turn machines over gracefully, do something like:

$ condor_hold -constraint 'ClusterId == 72 && JobStatus == 1'

To put all of the idle cluster 72 jobs on hold.  You can later release
the held jobs once the cluster 74 jobs have grabbed the machines.

You could also just do:

$ condor_vacate_job 72

If you don't mind those jobs being restarted from scratch.

-- 
Daniel K. Forrest	Laboratory for Molecular and
forrest@xxxxxxxxxxxxx	Computational Genomics
(608) 262 - 9479	University of Wisconsin, Madison