I got this from a cluster user...
The nodes have different sized scratch partitions. I need to get some jobs > on the nodes with more scratch space. > I know how to asked for a specific node, but it would be better to make a > request for scratch space, just like memory, > cpus, … I have looked on the web, but I have not seen any requests based > on scratch size. Have you come across anything > about a slurm command that deals with scratch space? Local disk is mentioned in Consumable Resources, but I have yet to find a way to specify local disk space in sbatch. This could be a moving target depending how well jobs are cleaning up after a run.
