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

Re: [HTCondor-users] How to limit number of machines used in a long list of queues



On Tue, Sep 2, 2014 at 6:25 PM, <Greg.Hitchen@xxxxxxxx> wrote:
>
> MAX_JOBS_RUNNING = 50
>
> in the config of your submit node should do the trick.

The caveat here is that if you share the submit node with others, then
you'll constrain their jobs too (perhaps undesirably).

On Tue, Sep 2, 2014 at 9:41 PM, Ralph Finch <ralphmariafinch@xxxxxxxxx> wrote:
> Excellent, this does it. I also see on further reading of the HTCondor
> manual that DAGMan will do this and also provide a number of other desirable
> features (failure detection and retry for instance).
>
DAGMan will also do the trick. A third option is to use concurrency limits[1].

Depending on why you only want to use 50 of 80 slots, another approach
would be to allow the job to use all slots when the pool is idle, but
have preemption (or suspend, depending on job characteristics) policy
that allows other users to come in.

[1] http://research.cs.wisc.edu/htcondor/manual/v8.2/3_12Setting_Up.html#SECTION0041215000000000000000


Thanks,
BC

-- 
Ben Cotton
main: 888.292.5320

Cycle Computing
Leader in Utility HPC Software

http://www.cyclecomputing.com
twitter: @cyclecomputing