Slurm Job Scheduler#
The HPC cluster uses the Slurm Job Scheduler to assign users jobs to compute nodes. Jobs are allocated based on the requested resources, submission time, and the users previous usage. We use the FairShare algorithm which adjusts priority to balance usage across of our users.
Command Quick Reference
squeue lists your jobs in the queue
sinfo lists the state of all computers in the HPC cluster
sbatch submits batch jobs
sprio Displays the priorities of pending jobs in the queue
scancel can be used to cancel jobs
srun allocates a compute node for use
sacct display historical report data for jobs
seff displays job CPU and Memory efficiency
Partitions and Limits
How to select where your jobs run and the resources available to them.
Interactive Sessions
How to run interactive jobs with cluster resources
Batch jobs
How to submit batch jobs to the cluster and sample slurm batch job scripts
Job Monitoring and Management Commands
How to monitor and manage your active jobs
Job Resource Utilization
How to check resource utilization of completed jobs: how much memory did the completed job used? how many cores did the completed job utilized?