I'm trying to set up a Parallel Environment in SGE, with allocation rule $pe_slots, but I'm having trouble with the scheduler.

Single-core jobs hog the slots; there are never enough slots open at one moment for multi-core jobs get scheduled.

Can SGE notice when multi-core jobs have been waiting a long time, and reserve slots for them, rather than letting single-slot jobs continuously butt in line?


It doesn't do this automatically but if you have a non-zero max_reservation in your sched_conf you can add -R y to your qsub options to enable reserving resources for a job when it is high priority. On our clusters we have a jsv that examines the job requests and adds -R y as needed.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.