Re: [Condor-users] Log file with ClusterId in the name creates two files

["R. Kent Wenger" <wenger@xxxxxxxxxxx>]

On Fri, 5 Mar 2010, Greg Langmead wrote:

> I am using a trick I learned on this mailing list in my condor submit files:
>
> error = ./job/job.error.$$([ClusterId])
>
> so that if users run into errors and resubmit the job, I can still look 
> at the old stderr data to do a post-mortem. It works great! Except...
>
> there are always TWO files written:
>
> job/job.error.$$([ClusterId])
> job/job.error.12345
>
> The second one has the actual stderr data, the first one is empty. One 
> of the condor daemons is "touching" the first one before the job has an 
> ID. Does anyone know a way to stop this from happening, or knows of a 
> better way to do what I'm trying to do? It happens whether I submit the 
> job with condor_submit, or whether it gets submitted as part of 
> condor_submit_dag.

I just tried this:

  error = job.error.$(cluster)

and I only get one file.

Kent Wenger
Condor Team