Not sure if this question fits better here or at Unix.SE but I figured it feels more appropriate here. I am running some computation-heavy research on a national cluster which uses SLURM for scheduling jobs.
I realized that a part of my batch script (which creates a bunch of child processes) is functioning much slower than I would expect it to and I am suspecting that there are too many threads being invoked by the analysis tool I use.
I was hoping to check if there are too many context switches, but it appears as I would need the PID to check the number of context switches. Checking the processes by ps or top didn't give any hits for the name of the particular tool I am running, so I suppose that the actual processes that are run by my script are "hidden" by SLURM at a different level. I suppose it makes sense, since the actual computer I am interacting with through ssh, and the node that carries out the calculations aren't the same.
But then is there a way by which I can check things like cpu utilization (e.g. % load) or number of context switches, from command-line?
