Overview

Fsched is a deeply optimized derivative of the open-source Slurm scheduler (19.05 branch), designed for high-performance computing scenarios. This guide provides a quick reference for Fsched core commands to help users use cluster resources efficiently.

Intended Audience

• Basic users: quickly master common commands with this guide
• Advanced users: use advanced features together with the Slurm official documentation

Common Command Quick Reference

Below are some of the most commonly used commands:

• sbatch: Submit a job script to run. The script can also include one or more srun commands to start parallel tasks.
• srun: Run parallel jobs interactively in real time, usually for short tests, or combined with salloc and sbatch.
• salloc: Allocate resources for jobs that need real-time handling. A typical scenario is to allocate resources and start a shell, then use that shell to run srun commands to execute parallel tasks.
• sinfo: Show partition or node status with many filtering, sorting, and formatting options.
• squeue: Show jobs and job-step status in the queue, with many filtering, sorting, and formatting options.
• scancel: Cancel pending or running jobs or job steps, and can also send arbitrary signals to all processes in running jobs or job steps.
• sacct: Show accounting information for active or completed jobs or job steps (corresponding to billable compute time).
• scontrol: Display or set the status of Slurm jobs, partitions, nodes, etc.
• sacctmgr: Command-line tool for managing accounting data