Submitting Jobs#
Note: this training assumes you’ve completed the “Getting Started with ROAR Collab” training
At the most rudimentary level, our interest in ROAR Collab comes from a desire to take advantage of the cluster to do work that we would otherwise be unable to do, either because of (memory) space or (execution) time constraints. Doing work on ROAR Collab requires the submission of a job, which is simply just the “formal” ask to perform a computational task. Penn State’s ICDS has an entire section of ROAR Collab dedicated to submitting jobs, so please review this link.
There are two primary methods of running jobs with the SLURM Workload Manager (formerly the “Simple Linux Utility for Resource Management”):
Interactively. Please be mindful to only execute scripts or submit a job from an interactive node. The default (submit) node that you are brought to after connecting to Penn State’s cluster is not suitable for scripting and scheduling batch jobs.
We’ve put together a more in-depth training presentation on jobs at this link. The Roar Collab team has also developed a training on data management in the Cluster. Find it here.