| name | slurm-workload-manager |
| description | Comprehensive guide and documentation for Slurm Workload Manager. Use when submitting jobs, managing queues, configuring Slurm, or troubleshooting Slurm-managed clusters. |
Slurm Workload Manager
This skill provides a complete reference for interacting with and managing a Slurm-based HPC cluster. It includes detailed command guides, job submission templates, and extensive configuration references.
Core Workflows
1. Job Submission and Management
Submit and monitor your computational tasks.
- Submit a batch script: Use
sbatch. See sbatch.md for full options.
- Run interactive tasks: Use
srun. See srun.md.
- Check job status: Use
squeue. See squeue.md.
- Cancel jobs: Use
scancel. See scancel.md.
2. Monitoring Cluster State
View nodes, partitions, and resource availability.
- Cluster overview: Use
sinfo. See sinfo.md.
- Node/Job details: Use
scontrol show [node|job] <id>. See scontrol.md.
- Accounting & history: Use
sacct. See sacct.md.
3. Cluster Administration
Manage Slurm configuration, users, and accounts.
- Manage accounts/users: Use
sacctmgr. See sacctmgr.md.
- Configuration files: The primary configuration is slurm.conf.md.
- Generic Resources (GRES): Configure GPUs and other resources in gres.conf.md.
Command Reference Library
Detailed documentation for all Slurm utilities:
Configuration Reference
Documentation for Slurm configuration files:
Specialized Guides
For any topic not listed above, use grep to search the references/ directory for specific keywords.
