I have a set of dependent Slurm jobs that are successfully submitting. The jobs are setup as:
a b \ / c | d | e I need to submit this set of jobs 1000s of times, each time parametrized slightly differently. If I was going to be submitting a large batch of jobs without dependencies, I would use a job array to be kind to other users and the scheduler. What is the best practice for submitting job arrays of dependent jobs?
Possible wrinkle: Each job (a/b - e) is parameterized slightly differently for SBATCH (nodes, tasks-per-node, etc.).