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

Re: [Condor-users] Disabling the restarting of jobs

On Feb 28, 2008, at 7:37 PM, Robert E. Parrott wrote:

We have a certain set of users whose parallel code is fairly strict
in the way it names and manages files. They are looking for an option
to disable the restarting of jobs after the jobs have been damaged
(for example by an OOM condition on a node), because a restarted job
will otherwise overwrite the partial, and still usable, data files.

Is there a submit file expression, of a config expression, with a
boolean to be added to the config file, that would put all jobs that
would otherwise restart into a "hold" state?

As a secondary question, would there be a way to update the
"HoldReason" classad expression with something relevant?

Adding this to your submit file will put the job on hold if it doesn't exit with status 0:
on_exit_hold = ExitCode =!= 0

You can also prevent the jobs from rerunning if accidentally released like so:
requirements = NumJobStarts  =?= 0 || NumJobStarts =?= Undefined
periodic_hold = NumJobStarts > 0 && JobStatus == 1

Thanks and regards,
Jaime Frey
UW-Madison Condor Team