Rechenkraft.net e.V.

- progress indicator and estimated run time are only guesses. But in most cases the values fits to the reality. Nevertheless it can happen that a wu needs more than 100%. In most cases the wu will finish before 100%.

- no checkpoints

yoyo

http://www.rechenkraft.net/yoyo/workuni ... id=1454919
( evo_A1227051903-1027-0_17.2MB_31.9 )
- is now at 123% and still rising. This is the first time I have had a WU so high.
Have you a recommendation how high should be allowed before cancelling ?

It finished normally - must have been close to 130%.
Panic over.

JohnMD hat geschrieben:http://www.rechenkraft.net/yoyo/workuni ... id=1454919
( evo_A1227051903-1027-0_17.2MB_31.9 )
- is now at 123% and still rising. This is the first time I have had a WU so high.
Have you a recommendation how high should be allowed before cancelling ?

It finished normally - must have been close to 130%.
Panic over.

Can be going and going - even over 200 %. The WU time is just a bad approximation. Best thing for evo@home is just to let your computer run 24/7. Because up to now checkpointing is not integrated - not in the pure client.

After manually suspending a WU and later on resuming it again, BOINC says it is active but it doesn't progress nor use CPU time. Between suspending and resuming WU BOINC was active non-stop. Tried this with 2 WUs.

After a reboot I got three new WUs but they dont even crunch . The wrapper starts but then it goes quiet. Aborted them and suspended project.

I have 2 w/u showing 0% completed. One has run 11h:32m evo_A1233838803-1143-0_34.3MB_47.9 and the other 4h:45m evo_A1233838803-1160-0_42.9MB_47.9. Both seemed to be running ok in taskmanager (I have suspended them for now). Are they stuck? Should I abort or let them continue? I have had a few long running ones before but I have never seen them stay at 0%.

Hello,
both wu names are ending with 47.9, which means they are estimated to run 47.9 hours on a 500MHz Pentium. So they are long runners. Depending on your computer you can slightly estimate the runtime.
If progress indicator does not increase does not mean that the wu does not run. It may also be the case, that the progress, which is written by the wrapped application, is somehow not fetched by the wrapper.
So I would let run them to the end.
yoyo

Yoyo, I will trust your judgement and let them run. Similar ones on my machines have finished in 11-13h range. They normally show over 100% though!! Fingers crossed.
Thanks.

EDIT: first one finished after 12h:17m. stderr out was filled with ''can't open eProgress.txt file''. Looks like you were spot on yoyo. EDIT

about 20h (140%)

for task: evo_A1259803203-44-0_34.3MB_63.84_0
application: evolution@home 1.12
process: S005r6gSimulatorWin32.exe
on: windows7 64bit RCm quadcore

some values:
processor_time progress
15:14 - 111,775%
16:22 - 120,053%
17:38 - 130,841%
18:13 - 135,724%

Tasks that run past 100% for long periods of time are very likely to be returned after the deadline as the BOINC CPU scheduler believes that they are within seconds of being done. So the CPU scheduler may wait to re-schedule them until just a few minutes before the deadline. Many hours later they are still crunching, and late.

At 123%, at least 2 hours of crunching past 100% and still going.

I just caught one that had been stuck at ~97.4% / ~11 hours CPU time, not consuming CPU time anymore.

When I stopped BOINC, the Evo process (not the wrapper, the program itself) didn't exit, I had to kill it from the task manager.

It is this one : http://www.rechenkraft.net/yoyo/workuni ... id=5001081

Well, in about 11 hours (as it restarted at 0%) we will probably know wether the problem persists or has been a "one-bug wonder".


edit : Done within 8.5 hours, no trouble after the retry.

Is there any chance we may see checkpointing for evo in near future? I am just asking since this is my favourite yoyo-project and I may have to abandon my XP Cruncher soon. I had a look at the instructions for Linux, but I do have some doubts I will be able to get it working on wine with my very limited Linux-skills.

To the best of my knowledge, there are no current plans for E@H checkpointing.

Michael.