Partitions / Queues

From HPCwiki
Revision as of 09:06, 16 June 2026 by Haars0011 (talk | contribs) (Phase 1 § 4 P1.4.2: split Partitions / Queues out of Using Slurm § Queues (via create-page on MediaWiki MCP Server))
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

The SLURM scheduler on Anunna divides compute nodes into separate partitions (sometimes called queues). When you submit a job, you choose which partition it runs in by setting --partition on your sbatch script.

Available partitions

  • main — the default partition for CPU jobs. Use this unless you have a specific reason to pick another.
  • gpu — for jobs that need NVIDIA GPUs.
  • gpu_amd — for jobs that need AMD GPUs.

To see the partitions currently available on the cluster, run:

sinfo

This lists every partition together with its state and the nodes that belong to it.

Selecting a partition

The default partition is main. If you don't specify --partition, your job runs there. To select a different one, add a directive to your sbatch script:

#SBATCH --partition=gpu

For GPU partitions you also need to request the GPU itself as a generic resource — see Choosing a node (constraints) for the details.

See also