Hello, I am looking for some information on items that we do currently on our scheduler that have a Slurm parallel, and if not what we can do to help facilitate our transition.

1) We would like to limit the amount of CPUs on nodes.  Within slurm.conf, are we able to specify CPUs = Any# < Actual CPU count(eg, actual is 8, but we would like to specify 4), or are we forced to specify the CPU count.  If we must keep the actual amount of CPUs in this parameter, what is the accepted way to limit the amount of 'slots' on a Node?

2) Is it possible within Slurm to hold all jobs for a single user at a time, or does this require a loop to go through each of a user's jobs to requeue one by one, constantly to ensure they do not run any work?

3) Is it possible within Slurm to requeue a single user's jobs, or does this require a loop to go through each of a user's jobs to requeue one by one?

4) What is the proper way within Slurm to 'Pause' scheduling.  Expected behavior is to still accept new jobs but not allow any work to start.

5) Does Slurm have a concept of 'purgetime'?  As in, a way to see only completed jobs, and you can only see the completed jobs within a past window, 
  ex: running command X shows you all completed jobs from the past 12 hours, but anything beyond the 12 hour window requires further digging within sacct.