At large scales, MD simulation can fail with a small probability (node failure,...).
The script has currently no precaution against that, it just loses all these structures.
It's low priority, but probably necessary.
You can set resubmit_failed = True as an argument to the AppManager. This will automatically resubmit all failed tasks. But I recommend this after resolving #2
At large scales, MD simulation can fail with a small probability (node failure,...). The script has currently no precaution against that, it just loses all these structures. It's low priority, but probably necessary.