On Tue, 2 Aug 2016, Joshua Baker-LePain wrote:
On Mon, 1 Aug 2016 at 1:16pm, Thomas M. Payerle wrote
So jobs of "paying" customers can cut ahead of jobs of "non-paying" and
"scavenger" in the queue, and will bump "scavenger" jobs. They will have
to wait for standard priority jobs that are already running, but these have
stricter walltime limits (although can still run for several days). Jobs
on the standard priority partition will cut ahead of scavenger jobs in the
queue, and bump running scavenger jobs if needed. And those that can live
with being preempted, etc. can submit scavenger jobs to use up any spare
CPU cycles w/out having their allocations charged.
Thanks so much for the details. What kind of preemption are you using?
We do QOS based preemption (standard and high-priority partitions have a
handful of QOSes available, scavenger partition must use scavenger QOS which
is preemptible by the others). We do PreemptMode=REQUEUE with no grace time.
Scavenger QOS has UsageFactor=0 so there is no charge for such jobs, though
we still require users to have access to a valid allocation account in order
to access the system to submit scavenger jobs (i.e. either "paying" customers,
or "non-paying" customers granted an allocation account by applying to a faculty
committee).
--
Joshua Baker-LePain
QB3 Shared Cluster Sysadmin
UCSF
Tom Payerle
IT-ETI-EUS paye...@umd.edu
4254 Stadium Dr (301) 405-6135
University of Maryland
College Park, MD 20742-4111
