In our current cluster, we have used NFS to store the job files. When we use qsub to submit a job under a shared directory, don’t want to stagein. When the job is done, don’t want to stageout. Because that is wasting the I/O.
How to disable stagein/stageout in job life cycle?