When running cases on Niagara, it is reccomended to follow the following guidelines:
As mentioned on the Getting Started page, run cases on your scratch directory, then save final data files to your project directory. This avoids issues of the group running out of space on project (remember, we only have 20TB shared between everyone).
Try to avoid spamming runs at the queue. Both the environmental impact and monetary cost of running simulations is significant. Think twice before submitting a lot of very large jobs!
Be conservative with the data files you save on project. For example, a very large unsteady optimization will have many solution and grid files that can take up a lot of space, and which may not all be necessary to store on project. If you do need to store larger files or a significant amount of data, you can use your archive (or HPSS) directory. Talk to Alex about details for transfering files between scratch/project and archive.
Look out for files named "core.[numbers]". These often appear after failed runs as a data-dump from the processors. Although in theory these could be useful in determining why a job failed, they rarely are, and take up up to hundreds of useless GB. If you find that a directory is taking up an unexplainably high amount of space (you can check with the command "du -hs path/to/directory"), it is likely that you have some these core files there. You should remove these whenever possible as they are unecessary. (tip: search for such files with "find /path/to/directory -type f -name 'core.*'")