DUNE global pool why Fermilab and some other sites ballooning in memory usage

DUNE / dist-comp

Action items for DUNE distributed computing, and common scripts that are used.

2 stars 0 forks source link

DUNE global pool why Fermilab and some other sites ballooning in memory usage #89

Closed StevenCTimm closed 4 months ago

StevenCTimm commented 4 months ago

This is now understood--there are some variables normally set by default in jobsub which justIN was clearing out just before it executed the user jobscript, thus tensor flow, etc were able to take all threads. Now fixed.