WMCore icon indicating copy to clipboard operation
WMCore copied to clipboard

Running steps after early step interruption for memory usage

Open vlimant opened this issue 1 year ago • 0 comments

Apologies if this is already addressed somewhere else, and/or if not relevant. As part of looking into a memory issue in RunIII2024Summer24DRPremix, I stumbled on the fact that in a stepchain setup, if one of the early step is interrupted by the memory watch-guard, the rest of the steps are ran regardless, and to my knowledge, since the job is marked as failed, any of the output will be discarded anyways.

Would it be better to stop the job at the step that was interrupted and not waste the cpu running things that will be thrown away ?

vlimant avatar Dec 17 '24 11:12 vlimant