Re: [HTCondor-users] Simple use case

[Brian Candler <b.candler@xxxxxxxxx>]

> Can Condor submit N jobs in parallel to a specific set of nodes; I want them to run once and only once on each node.

I haven't used it myself, but the parallel universe might do what you want; you can make a unique claim on N nodes and have a job run on those N at the same time. This is designed for jobs which need to use inter-process communication. None of the jobs start until all are ready.

Otherwise, you can submit N jobs each with a different requirements _expression_:

requirements = (Machine == "foo.example.com")
queue
requirements = (Machine == "bar.example.com")
queue
... etc

Those jobs are then constrained to run only on the machines you have requested. However they may not run at the same time, and there may be other jobs running on those machines at the same time (if the machine has multiple slots)

Normally you wouldn't express requirements this way; it's better to label the resources you're interested in and which machines have them, so that condor can choose the best machine to run them on. See
http://stackoverflow.com/questions/9864766/how-to-tell-condor-to-dispatch-jobs-only-to-machines-on-the-cluster-that-have

If you describe what you're trying to achieve in more detail, we may be able to find a better way of doing it.

Regards,

Brian.