Message boards :
Number crunching :
If your work ends on several short runs...
Message board moderation
Author | Message |
---|---|
Send message Joined: 13 Jul 05 Posts: 456 Credit: 75,142 RAC: 0 |
If this time around your work happens to end of several (say 6+) short runs, then this could happen to you next time: ... [posted by Ray in another thread] This could have happened because last time that participant got work from LHC on that box, however long ago that was, the last several WU all ran very short. The client expected that to happen again and asked for more to compensate. LHC often has short runs as if the beam hits the wall of the tube there is no point the simulation running any further, that would just be a waste of your cpu. By bad luck, sometimes several will come in a row. Even then there is no problem if there is one final full-length run - as this will reset things (when runs get longer, the client rests fast, it only reacts gradually when runs are getting shorter). By even more bad luck, sometimes there will be a row of short runs that comes right at the end of the work you got, leaving a timebomb for the next time that box gets work. So, my suggestion is: If you notice that the last 6 or more tasks all run short then it is worth resetting LHC (projects tab). This will prevent the 60 task flood hitting you next time around! If it does happen to you, then the responsible thing to do is exactly what Ray did, calculate how many you can really use, and abort the rest. Abort from the top downwards in the list, but not the one that is currently running. Update the project when the next following (non-aborted) task starts to run, and the aborted work will be automatically redistributed by the server. River~~ ![]() |
©2025 CERN