2.1.5. Site-specific Slurm info

While the Running jobs in Slurm and Job scripts: Advanced topics pages provide basic and advanced information regarding Slurm, there are additional points to consider when when using Slurm on Tier-2 clusters hosted at UAntwerpen.

Converting from PBS to Slurm

Have a look at the quick PBS-to-Slurm conversion tables if you need help converting your PBS job scripts to Slurm. See also A list of important differences between Torque and Slurm.

Requesting memory

When requesting memory, keep in mind that no swap space is available in case your application runs out of memory. Swapping is disabled because the nodes don’t have drives suitable for the load caused by swapping and because swapping is extremely detrimental to performance.

Requesting nodes

On UAntwerp clusters we ask to only use the -N/--nodes option to request (and use) full nodes. This is because in the current Slurm configuration the empty cores cannot be used by other jobs. We furthermore ask to abstain from the --nodes=<minimum number>-<maximum number> option without consulting us first and discussing why it would make sense for your job.

Atools module

On UAntwerp clusters the most recent version of the atools package is available as the module atools/slurm.