Re: [Condor-users] Dagman and PeriodicRemove

[Peter Doherty <doherty@xxxxxxxxxxxxxxxxxxx>]

On Jun 3, 2010, at 18:18 , R. Kent Wenger wrote:

> On Thu, 3 Jun 2010, Peter Doherty wrote:
>
> > I have a Condor DAG, and I have a PeriodicRemove statement in my  
> > classad that times out jobs after 30 minutes.
> > I have Retry = 1 in my *.dag file to retry jobs that fail, either  
> > with a non-zero exit code, or that fail the POST script.
> >
> > But if the job is removed by the PeriodicRemove statement, I don't  
> > want to run the POST script, and I don't want the job retried.  I  
> > just want it removed, and nothing more done with it.
> >
> > Can I make condor do this?
>
> There's no way to avoid running the POST script.  However, if the  
> POST script can recognize that the job was removed by the  
> PeriodicRemove expression, you can avoid retrying the node.  You can  
> do this by having
> the POST script return a "special" exit code in that case, and using  
> the
> UNLESS-EXIT option in your retry statements:
>
>  RETRY JobName NumberOfRetries UNLESS-EXIT value


Thanks Kent.  I was able to make your suggestion work.  I just had the  
POST script grep through the job's log file to find out if it was  
removed due to PeriodicRemove.  I wasn't sure if there was a more  
elegant way of doing this.

Peter