On Mon, Feb 08, 2021 at 12:36:06PM +0100, Ansgar Esztermann-Kirchner wrote:
> Of course, one could use different partitions for different nodes, and > then submit individual jobs with CPU requests tailored to one such > partition, but I'd prefer a more flexible approach where a given job > could run on any large enough node. After scouring the docs once more, I've noticed DefaultCpusPerGpu, which seems to be exactly what I was looking for: jobs request a number of GPUs, but no CPUs; and Slurm will assign an appropriate number of CPUs. The only disadvantage is the fact that this is a partition parameter, so to retain full flexibility, jobs will have to mention all partitions (since there is no wildcard); but this shouldn;t be a problem for us since we have an automated submission tool that can take care of this. I have run some simple tests to ensure the parameter works as expected, but more thorough testing needs to be done. A. -- Ansgar Esztermann Sysadmin Dep. Theoretical and Computational Biophysics http://www.mpibpc.mpg.de/grubmueller/esztermann
smime.p7s
Description: S/MIME cryptographic signature
