1) Message boards : :
Number crunching :
Long Running Work Units
Posted 1026 days ago by Richard Kimber
9hrs is to long, please abort it or try a host restart. Maybe it helps.
I will forward this to the admin.
This is still a big problem. I may need to detach from the project, it's requiring too much monitoring
|
2) Message boards : :
Number crunching :
Long Running Work Units
Posted 1075 days ago by Richard Kimber
The other issue is that once a work unit has been going for a very long time, boinc runs it as a high priority job because all the chess jobs have very short deadlines. This means that it is never swapped out according to the preferences for allocating resources between projects.
I currently have chess960_52758_117_3 running as high priority. It has been running for 9:17:30 and boincmanager shows 18:39:17 to completion, but it also shows 0.000% progress. The deadline is Wed 09 June 2010 04:15:46 BST.
How long should I leave it? Normally I would abort it.
|
3) Message boards : :
Number crunching :
Long Running Work Units
Posted 1076 days ago by Richard Kimber
The position seems to be right for me but you found many moves and the calculating time of an unit depends ever on the number of moves.
The problem is that on many jobs boincmanager shows an elapsed time much larger than the 3 minutes or so given as the original completion time (maybe 12 hours, if I haven't kept an eye on it) and at the same time it shows *zero* progress and an increasing 'To completion' time. If it showed *some* progress, I wouldn't abort it, but if after many hours no progress is shown, that suggests there's something wrong, surely. All the other projects I work on show progress being made and at some point a decreasing 'Time to completion'.
|
4) Message boards : :
Number crunching :
Long Running Work Units
Posted 1081 days ago by Richard Kimber
The problem should be solved now. The error was an incorrect position in some units.
It was solved for quite a while, but the problem has returned again - for me at least since mid-May.
|
5) Message boards : :
Number crunching :
Boinc preferences
Posted 1333 days ago by Richard Kimber
I guess I know the answer to this, but I'll ask it anyway.
Is there any way of getting chess960 to observe the Boinc Manager preferences?
I have configured Boinc Manager such that Boinc projects only use three out of the 4 CPUs. Chess960 does not observe this, and when its work units are running all four processors are used, and my machine runs hotter than I would like.
|
6) Message boards : :
Number crunching :
Long Running Work Units
Posted 1358 days ago by Richard Kimber
I just use boincmgr to abort the one that's not doing anything and everything goes back to normal - until the next one that sticks. It does mean one has to remember to keep an eye on the progress that's being made.
|
7) Message boards : :
Number crunching :
work unit ends but no finish file
Posted 1373 days ago by Richard Kimber
It seems to be segfaulting
glaurung[7996]: segfault at 7fffccd462c5 ip 0000000000416571 sp 00007fffcc929018 error 4 in glaurung[400000+21000]
|
8) Message boards : :
Number crunching :
work unit ends but no finish file
Posted 1373 days ago by Richard Kimber
I am getting errors saying that a task has exited with zero status but no 'finished' file, and that I should reset the project, which I have done.
|
9) Message boards : :
Number crunching :
Work Needed
Posted 1822 days ago by Richard Kimber
Yes. I don't seem to have had any work since 28 April - although I have been offline for the last week because of ISP incompetence.
|
10) Message boards : :
Number crunching :
Chess960 ignores preference
Posted 2186 days ago by Richard Kimber
I only have one processor and have no problems. The latest Bonic is 5.8.16 try upgrading.
Forgive me, but since you only have one processor I don't see the relevance of your not having any problems. The 64 bit Boinc I have is the most up-to-date for my distribution and since work units for other projects didn't have this issue, I think it's unlikely that an upgrade would solve the problem.
Also for the last several days all WUs were hanging at the 98% mark (no cpu usage.) I rebooted and now they complete normally.
I mostly don't have any problems running work units. From time to time one unit seems to hog the CPU without appearing to do work (according to BOINC Manager) and in this case I just abort the job and everything returns to normal. Oddly, the symptom of this is that gkrellm shows the job just running on one processor :-( Aaarrrgh.
I guess I'll just detach when the weather gets hot.
|
