Task virtual memory size lower than working set size

Author	Message
MichaelD Send message Joined: 31 May 16 Posts: 1 Credit: 610,291 RAC: 0	Message 105354 - Posted: 7 Mar 2022, 21:59:50 UTC Many new Rosetta tasks are running slowly and go well past the deadline causing a backlog of overdue tasks, even tuning the Rosetta tasks down to two at a time doesn't help. These tasks all show the virtual memory size allocated to this task lower than the working set size which seems wrong. Other Rosetta jobs with higher virtual memory size than working set size run fine and complete on time. System is 8 core, 64GB ram. Note the elapsed time versus CPU time below. Is virtual memory set by the task or the boinc client itself? is there any way to manually override the task virtual memory setting? or does anyone else know how to fix this? It appears Rosetta python project tasks exhibit this issue, not always, but mostly. Other Rosetta non-python jobs work fine. CPU time 00:03:46 CPU time since checkpoint 00:00:01 Elapsed time 17:39:24 Estimated time remaining 02:10:57 Fraction done 88.998% Virtual memory size 1.63 GB Working set size 2.79 GB Directory slots/0 Process ID: 105354 · Rating: 0 · rate: / Reply Quote

Sid Celery Send message Joined: 11 Feb 08 Posts: 2124 Credit: 41,208,781 RAC: 10,419	Message 105356 - Posted: 8 Mar 2022, 3:21:34 UTC - in response to Message 105354. Many new Rosetta tasks are running slowly and go well past the deadline causing a backlog of overdue tasks, even tuning the Rosetta tasks down to two at a time doesn't help. These tasks all show the virtual memory size allocated to this task lower than the working set size which seems wrong. Other Rosetta jobs with higher virtual memory size than working set size run fine and complete on time. System is 8 core, 64GB ram. Note the elapsed time versus CPU time below. Is virtual memory set by the task or the boinc client itself? is there any way to manually override the task virtual memory setting? or does anyone else know how to fix this? It appears Rosetta python project tasks exhibit this issue, not always, but mostly. Other Rosetta non-python jobs work fine. CPU time 00:03:46 CPU time since checkpoint 00:00:01 Elapsed time 17:39:24 You can clearly see here there is no CPU time being allocated to the task. Every time you see this, abort the task. Don't wait for it to get to 17 hours. Even 1 hour is too much. Whenever it happens, it <never> restarts. Don't worry why it happened. The fact that it's happening for you is enough. Maybe some other host can run it better, but it's not a concern of yours. Abort it and forget it. ID: 105356 · Rating: 0 · rate: / Reply Quote