Hi, I’m currently responsible for a mini-cluster of three hexacore machines (condensed matter theory group at a university). All three machines are identical and share a common filesystem (via NFS on a 1Gb ethernet connection).
All of our jobs are either single or multi-threaded, but we don’t do anything more exotic than that (no MPI, etc). Currently, to start a job, we ssh around until we find a machine that is mostly free and then start the job there.
I’ve been looking for “job scheduling” software, but all of the packages seem to have a non-free license and be far more complicated that we need. I have looked through the software manager in YaST, but haven’t found anything that sounds like it will do what I need it to (but maybe I just don’t know what to search for).
Specifically, I need a package that will make it when a user executes a job, that the job gets executed on the computer with the smallest load average. The user must also be able to see (and kill) jobs they have running on all machines.
Thanks in advance!