Say I have a cluster with 1 node and 36 CPU cores, but I want SLURM to manage and allocate only 20 of these cores. Is there a way to configure SLURM so it's 'world' only consists of this subset of the total CPUs?
Note this is different from setting limits to a particular job. I want a situation where in the example above if job1 requests 20 cpus, and job2 requests 5, SLURM waits for job1 to finish before allocating job2 (as opposed to just starting job2 right away because there are 36 cpus total).