Run a simulated checkpointable job on Spot capacity, create a checkpoint, interrupt the job, and resume it without losing the completed work.
Simulateur déterministe — tape une commande (help) pour commencer.
hpc@cluster:~$
Objectifs
- Create a Spot-enabled cluster.
- Submit a checkpointable job.
- Advance the job, save a checkpoint, and interrupt it.
- Resume and complete the job.
Indices
Voir la solution
hpc create-cluster --nodes 2 --instance-type c7i.4xlarge --spot true sbatch --job-name long-run --nodes 2 --time 12 --checkpoint true tick 5 checkpoint 1001 interrupt 1001 resume 1001 tick 8 sacct