Slurm in BCM and the performance lab

In the NVIDIA Base Command Manager environment, Slurm is the job scheduler. BCM integrates Slurm to provision nodes, configure GPU resources, and provide monitoring.

The performance lab tests:

1. sinfo to check node/partition status
2. scontrol show node <name> to inspect a problem node
3. scontrol update NodeName=<n> State=RESUME to fix a drained node
4. sbatch or srun to submit a test job
5. squeue and sacct to verify the job ran

Practice these five steps until they are instinct.