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

Re: [HTCondor-users] Debugging NEGOTIATOR_PRE_JOB_RANK

On Tue, 2013-01-22 at 15:29 +0100, Chris Filo Gorgolewski wrote:
> Thanks, that should work but is too strict. I just realized that I
> wasn't precise in the way I described the problem. I'm not looking for
> a strict restriction but a priority. I want the jobs to be executed at
> the servers unless there are no available slots - then they should be
> directed to workstations. 

The negotiation algorithm currently assigns at most one job to a slot
per cycle, including partitionable slots.   That works against the kind
of behavior you are seeking.

There is an enhancement to the negotiator under review that should help
with this kind of use case:

Ticket #3435: RFE: negotiator-side resource consumption with consumption