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

Re: [Condor-users] Transfer Files mechanism



On Aug 18, 2005, at 12:13 PM, Bruno Miguel Tavares Goncalves wrote:

Using a submit file like this one:

Executable = 1D.x
Universe = Vanilla
Output = results/1D/1D.out.$(CLUSTER).$(PROCESS)
Error = results/1D/1D.err.$(CLUSTER).$(PROCESS)
Log = results/1D/1D.log.$(CLUSTER)
Nice_User = True

Arguments = 1D.in.$(PROCESS) $(CLUSTER).$(PROCESS)
Transfer_Input_files = 1D.in.$(PROCESS)
Queue 10

and with a shared file system, will the input files be tranfered to the spool directory? And if so, when? At queue time? or at run time? The manual ( http://www.cs.wisc.edu/condor/manual/v6.7/2_5Submitting_Job.html ) isn't very clear on this aspect.
Should I be able to submit a cluster of jobs, modify the input files, resubmit another cluster, etc... and have each cluster have a different version of the input files? Is this the way to do it? Or should I just create several copies of the input files and have each cluster read a different set of files?

If you use the -r or -s command-line arguments to condor_submit, then the files will be copied into the spool directory. Otherwise, they will not and Condor will use the original copies when the job runs. With -r and -s, the output files are also stored in the spool directory. To retrieve them after the job completes, you need to run condor_transfer_data.

+----------------------------------+---------------------------------+

|            Jaime Frey            |  Public Split on Whether        |

|        jfrey@xxxxxxxxxxx         |  Bush Is a Divider              |

|  http://www.cs.wisc.edu/~jfrey/  |         -- CNN Scrolling Banner |

+----------------------------------+---------------------------------+