I have two separate machines: a Windows 10 gaming rig which is also my primary productivity machine, and an Ubuntu workstation. I’ve got the workstation set up so I can ssh into it from Windows over LAN. I’m thinking of adding SLURM to it so that I can submit compute workloads from Windows without keeping an ssh session open (e.g. for when I put my machine to sleep). In essence, it would be like a mini-HPC for me. I don’t have any experience doing sysadmin type stuff, so if anyone has suggestions or advice that would be appreciated.
I set it up back in the day when it was basically just slurmd and munge. It looks like it has gotten a bit more complicated now. If you’re in academia and/or HPC slurm isn’t a terrible idea. If you’re in industry or data engineering you might want to look at https://k3s.io/ instead of slurm. It’s just one or two steps to install k3s and Kubernetes seems to carry more weight on a resume outside HPC.
Depending on your workload slurm on only one system may be a bit overkill. I did an internship in a hpc company and it is a bit tricky to set up. If you want to just run a process and kill the ssh session then something like tmux on your ubuntu machine may be enough. Tmux is basically a terminal window in a terminal window and can be detached and attached without killing the processes inside tmux.
Hmm, this is a good point. I might stay in HPC after grad school, but this looks interesting to learn too. Although I know that lot of scientific computing groups in industry use on prem HPC with something like SLURM or Platform LSF.
Thanks for the suggestion. I’ll look into tmux. SLURM as the nice property of batching things though. I guess on a single system, batching can just be done with a simple script.