SLURM

scrontab is Slurm’s “cron”—it lets you schedule recurring batch jobs (hourly, daily, weekly, etc.) to run on the cluster without you being logged in. It works a lot like Linux crontab, but submits jobs to Slurm. 

How you use it

Open your schedule: scrontab -e (list with scrontab -l, clear with scrontab -r). 

Put Slurm options at the top using #SCRON lines (partition, account, time, outputs, etc.). 

Add cron-style lines (5 time fields + the command/script to run). Times are typically interpreted in UTC on many systems. 

Use full paths and a script that loads your modules/conda env, just like a normal sbatch job. 

Minimal example

Runs a script at the start of every hour, on partition high, account sub1, for up to 1 minute, appending output:

#SCRON -p high

#SCRON -A sub1

#SCRON -t 00:01:00

#SCRON -o /home/you/logs/hourly-%j.out

#SCRON --open-mode=append

@hourly /home/you/bin/run_my_task.sh

Tips

Put any environment setup inside your script (e.g., module load ...). The scron environment is minimal. 

Use sacct -Dj <JOBID> to view accounting for scrontab-run jobs (the -D flag shows duplicates sometimes created by recurring runs).