Slurm lab
Job scheduling on a real single-node cluster — the scheduler that runs most GPU clusters. Learn mode covers sbatch, resource requests and the queue; Break-fix mode fixes drained nodes, impossible resource requests and dead daemons.
Why this matters: On shared GPU infrastructure the scheduler is the interface — researchers do not ssh into nodes, they submit jobs. Knowing why a job is stuck in PENDING is core AI-infrastructure work and almost nobody practises it.
Slurm — getting started
New to job scheduling? Each walkthrough spins up a real Slurm cluster and teaches one core skill — submit a job, write an sbatch script, request resources, manage the queue — one command at a time. This is the scheduler that runs GPU clusters.
Loading scenarios…