[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [Condor-users] questions about remote queue submission
- Date: Fri, 18 Feb 2005 10:39:49 -0500
- From: "Ian Chesal" <ICHESAL@xxxxxxxxxx>
- Subject: RE: [Condor-users] questions about remote queue submission
> No, that's not correct. It has nothing to do with central
> managers at all.
> -n is really meant for submitting to another schedd on the
> same machine, or at least with the same view of the
> filesystem. condor_submit (or condor_submit -n) tells the
> condor_schedd to create a new entry in the job queue, and
> what files need to be copied into the spool directory - but
> it doesn't actually send the files to the schedd.
Ahh. So because all my windows machines are using the same shared drive
space for files and such I'm okay doing this. Great.
> -r also tells the schedd create a new entry in the job queue
> for the new job, and what files are needed - but then sends
> the files needed to the schedd, so you don't need a shared
> filesystem between the two of them (it's actually a lot
> smarter than that - it creates a job on the remote schedd
> with that job in the 'HOLD' state, transfers the files over,
> and then releases the job, so we can recover from crashes)
That explains why the jobs were held on the machine even though the file
transfer via UNC crashed the remote schedd.
These explanations should really be in the help for condor_submit!