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

Re: [Condor-users] To force all job from same cluster need to run in same machine.

Johnson koil Raj <johnson.raj@xxxxxxxxx> wrote:
>     Is there any way to specify all jobs form same cluster should run on 
> same machine.
>     x.1 x.2 in machine   yyy only.

As Matt noted, one solution is to run one job through, note what
machine it got, then have the other jobs specify that machine.
The first job could be a throwaway /bin/true whose only purpose
is to land on a machine.  This could be implemented with a a
simple DAG; the first node is a throwaway to match a machine,
the.  The real job is a child node.  The first node's POST script
modifies the submit file to specify the machine chosen.

If you want to keep landing on the same machine because the first
job lays some files down on disk that other jobs need or strongly
want, you might use the STARTD_CRON_* settings to detect the
files and publish "JOB_SUCH_AND_SUCH_RAN_HERE=TRUE", then the
other jobs could requirements=JOB_SUCH_AND_SUCH_RAN_HERE.

Alan De Smet                              Condor Project Research
adesmet@xxxxxxxxxxx                http://www.cs.wisc.edu/condor/