When it reaches 12 hour it will stop.


Maybe some other process is stealing CPU resources from WCG.

WUs can vary considerably in length of time they take to process depending on whether they are a parent or child WU.

If you get a parent WU then it contains all targets for the WU and the HCMD app will try to process all of them until it comes up to the 60% in 6 hours rule or the hard 12 hour stop rule.

If a WU gets returned when its not complete then child WUs are generated on the remaining targets from the original parent WU and obviously since they contain fewer targets to be analyzed the child WUs often run for a shorter time.

To be precise, it's 12:00:00 hours, then to finish the position that is at that moment computing. If you switch on <checkpoint_debug> logging you can actually observe a position progress i.e. the last time a task checkpointed to give an idea from the recent checkpoints how long the next one could take.

Since your client ran 12 hours, we know the task had at least 60% done at 6 CPU hours, so it must have hit a few heavy ones on the course. One would expect at 60% / 6 hrs for a task to finis before the 12th hour. Some dont.

Mind you, dont know which client you run. If 6.10, then you see the wallclock time it took, not the CPU time. The CPU time you then can see in the task properties view.

Hope this helps to understand.

Happy Crunching.

edit: checkpoints in plural and more spelling

For the checkpoint reading details see this FAQ: http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=11332 . The instructions are right after the first screenshot and read on through the second paragraph else you miss it... adding a <logflag> in the cc_config.xml, which for Windows does not by default exist (will for the upcoming WCG skinned 6.10.58 I think)

It´s really easy. Just open the cc_config.xml with Notepad or Wordpad, then add the line, and save making sure the file does not change name. You really can't break BOINC when messing up this file... it will simply be ignored or just give a benign warning.