"Stuck" unit
Anonymous
2 Jan 2005 20:39:59 UTC
Topic 10655
(moderation:
Thanks for the report.
Actually the new pt workunits crunch in three steps: Two runs similar to the previous ft runs (0-50%, 50-100%), then a "comparison" run that we thought to be short enough for not needing chekcpointing or a single percent of the done counter. Apparently this last step takes somewhat longer on your machine, I'll take a look at the WU and see if the reason is in the data.
BM
"Stuck" unit
)
> Under version 4.65, Unit pt11_I12_f59.998_b0.104_0 has been sitting on 100%
> completed for several hours now and not rolling over (still counting CPU time,
> nothing under "total time") This is on machine 1525. When I restart BOINC
> it starts at 50%, quickly jumps to 100%, then repeats the process.
> Might want to see if this is a reproduceable error. I'll keep a copy of the
> current file state, but I'm going to let it crunch on some other units to see
> if it's a corruption in my software.
I think we have now fixed this bug. The new science code was acting badly if it didn't find any source candidates in one of the data sets.
[Added Jan 8th] We're now testing a revised app version that hopefully will fix this problem. If our testing goes well it should be available in a day or so and will revive this stuck WU.
Bruce
>> Same problem here with
)
>> Same problem here with unit pt12_I12_f59.998_b0.104_9, crunching with einstein
> 4.71 on machine 1538, starts at just under 50%, gets to 50%, then upto 100%,
> and sits there.
Ooops - I thought we had this fixed... thanks for the report.
BM
BM