Questions and Answers : Windows : Major Problems with ATLAS Simulation 1.01
Message board moderation

To post messages, you must log in.


Send message
Joined: 23 Feb 09
Posts: 3
Credit: 2,998,777
RAC: 0
Message 36585 - Posted: 29 Aug 2018, 13:24:20 UTC

I have observed this issue on multiple machines. All running up-to-date windows 10 pro 64-bit and the latest Virtual box.
The remaining time estimator is wildly not accurate. (estimation around: 1 hour and 20mins, actual around 12 hours) Which causes the following problems:
1) Too many work units will be downloaded to the machine.
2) Jobs are not completed in time (mainly for the next reason)
3) After about 5mins I will consistently get "Postponed: VM job is unmanageable, restarting later" on my 8 core jobs (ATLAS Simulation 1.01 (vbox64_mt_mcore_atlas))
It doesn't restart later by itself. I have to close out boinc and reopen it before it tries again. In practice I get about 5mins of calculations every time I start up my computer.
So jobs are completed wildly late if I don't noticed and abort them after they have missed their deadline.
ID: 36585 · Report as offensive     Reply Quote

Send message
Joined: 13 Apr 18
Posts: 443
Credit: 8,438,885
RAC: 0
Message 36586 - Posted: 29 Aug 2018, 15:01:17 UTC - in response to Message 36585.  
Last modified: 29 Aug 2018, 15:13:05 UTC

16GB RAM is not enough for 8-core VBox tasks. Go to and set "Max # CPUs" to 2 which will get you 2-core tasks instead of 8-core tasks.
Also, set "Max # of jobs" to 2 which will prevent downloading too many tasks and limit the concurrent cores on ATLAS to 4. I think that's the most you can do with 16GB RAM.

If you want to run Theory, Sixtrack and LHCb tasks in addition to ATLAS then you might want to use an app_config.xml for finer control.
ID: 36586 · Report as offensive     Reply Quote

Questions and Answers : Windows : Major Problems with ATLAS Simulation 1.01

©2024 CERN