Partitions

Nodes in the cluster are grouped into Slurm partitions (sometimes referred to as a queue for legacy reasons) based on their installed hardware and primary function.

The default partition on Apocrita is compute, which contains the standard compute nodes. Nodes configured for parallel workloads are grouped into the parallel partition, while GPU-enabled nodes are assigned to the gpu partition. Additional partitions may exist for nodes that cannot be classified under any of the partitions mentioned above.

Any cluster user may submit jobs to the compute, computeshort, gpushort, highmem and parallel partitions, as they are open access. Other partitions are access-controlled, with access evaluated on a per-user basis according to individual circumstances.

Use sinfo to view all partitions available to your account (note this does not mean you have access to run jobs in those partitions). To see only specific partition(s), for example either only compute, or compute and parallel:

sinfo --partition=compute
sinfo --partition=compute,parallel