[NLPL Task Force (A)] Max jobs per user

[Vinit Ravishankar]

Hi all,

I have an issue that might be easy to solve on your end — I want to run a (large) cluster of jobs on GPU, however, what I do /not/ want to do is hog the entire queue, because I can afford to wait for my results. There is currently no way for me to be considerate and monitor the queue to ensure I don’t take up more than a certain number of job slots, unless I write a complicated and annoying bash script to monitor and conditionally submit. Submitting an array job is not an option in my case.

The sacctmgr command should, in principle, allow me to set a hard limit to the number of running jobs associated with my account — so right now, for example, I could have run something like:

$ sacctmgr modify User vinitr where MaxJobs set MaxJobs=16

to ensure that no more than 16 of my GPU jobs are running at any given time. Unfortunately, I don’t have the right permissions to do this. It might be an idea to allow users to modify their own job limits — this lets us be more considerate, and would also likely help with the queuing annoyances we’ve had to deal with in the past. Do you think this might be possible?	

Best

– Vinit