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

Re: [HTCondor-users] distributing a limited number of jobs evenly

----- Original Message -----
> I have some memory intensive jobs which for example I would only like to have
> 4 of them running on each node, and want others to stay 'unclaimed' until
> the preceding jobs finish running. I would be much grateful if you could
> provide any inputs.

Possibly the simplest approach would be to advertise partitionable slots and then use request_memory in the job submit file.   As long as the values given by request_memory have reasonable fidelity, jobs will stop matching each slot when its available memory is exhausted.

There are other approaches with varying additional complexity if more logic is required.