Questions and Answers :
Sixtrack :
problem with a long running WU. Bug or data problem?
Message board moderation
Author | Message |
---|---|
Send message Joined: 14 Oct 08 Posts: 5 Credit: 215,057 RAC: 0 |
I'm about to cancel a WU which has some really odd properties. It's a WU which is supposed to finish about 6 days from now, but has about a 100ish hours of running time left. This will no doubt fail the deadline as the computer is only on for about 12 - 15 hours per day. Normally this system will complete a sixtrack WU in about 7½ to 8 hours, so I suspect something is wrong. I'm about to cancel this work, but I'd like to notify you before I do. Could this be caused by a power outage I experienced yesterday? Is this a freak WU? Is there a way to check the WU's data integrity to check for damage? Here are the details (In dutch, I hope that this isn't a problem) |
Send message Joined: 14 Oct 08 Posts: 5 Credit: 215,057 RAC: 0 |
I've cancelled the WU. The completion time wasn't dropping to an acceptable level. New LHC WU's are running with a normal projected completion time. |
Send message Joined: 29 Sep 04 Posts: 281 Credit: 11,859,285 RAC: 0 |
Hi Ocf81, As you say, most WUs run about 7 hrs but Eric is currently doing a study with a few longer WUs that may run 24Hrs, some 50Hrs and I currently have a couple that started with an estimate of 203Hrs. I expect both to complete about 100Hrs or so. There is some discussion on these longer WUs in Very long jobs in the News section and Long WU's in Number Crunching. The problem of the deadline for these longer WUs is being addressed and they now seem to be getting credits that reflect the longer runtime. You machine looks like it would be more than capable of tackling one of these longer tasks. |
©2024 CERN