Progress 0.000%
26 Aug 2009 22:51:20 UTC
Topic 194511
(moderation:
The progress bar in boinc tasks has been 0.000% now for days for einstein and primegrid. 1 einstein task has been running now 18:39:04 with a 0.000% progress. 4 of 10 tasks running are einstein yet cpu usage is 32%.
I tried a reset project and reinstall of boinc to no success. 6.6.36
Rosetta milkyway ABC and seti-cuda all working fine, about a week now einstein and primegrid are not.
Any ideas...
Progress 0.000%
)
Could be related to your host-os W7-X64 or a bug in your boinc-version 6.6.37.
RE: The progress bar in
)
Consider looking in the Boinc "slots" subfolder(s) at the contents of file "stderr.txt". I believe that this file gets sent to E@h server at process completion. It has progress info and may tell you just what the process is "hung up" on.
There was a posting in the past few days regarding a "hang up" accessing the lock file and the process just checked every minute or so and "slept" between tries. "Message boards : Problems and Bug Reports : Not Downloading"...Here! Possibly similar issue for you.
Stan
RE: Could be related to
)
I got it twice on my machines on Win 7 64 using exactly that client and had to detach twoo machines
Why detaching, you can
)
Why detaching, you can download a newer boinc-client 6.10.7. Seems to be running without problems.
OK -- might be the best way
)
OK -- might be the best way to make it go away (0% hangups). Just for us dummies, kindly recite the procedure necessary to download the latest (newest) version of BOINC ... thanks
If I've lived this long - I gotta be that old!
RE: OK -- might be the best
)
Go to the BOINC home page (boinc.berkeley.edu) and follow the instructions. I am using the 6.6.40 Linux version and it works.
Tullio
RE: OK -- might be the best
)
Just a word of warning about the very latest versions - in this case 6.10.x. This whole series is for the development of an ATI GPU aware BOINC. If you are not crunching on an ATI GPU (Milkyway and Collatz Conjecture projects) you are crazy to buy into the potential grief of all the extra bugs likely to be introduced during the development process.
Many would probably argue that the 6.6.x series was equally unstable so what does it matter. Well at least that series has had a bit of time to mature and at last they have made a new release to replace 6.6.36 which (by all accounts) did have scheduling problems.
If you don't need the latest NVidia capabilities, you could even give the 6.6.x series a miss as well and think about 6.4.7/6.5.0. I've seen Paul D. Buck mention numerous times and in various places, the problems of 6.6.x and that he found 6.5.0 to be the best for normal and stable operation and free from the more insidious bugs. Paul does spend a lot of time testing BOINC versions on various machines so that's got to be worth something. On the other hand I spend virtually no time at the bleeding edge. I've never run a 6.6.x version and I probably never will and I'm only passing on the experiences of others. Most of my Win machines run 6.4.7 or 6.5.0 and Linux machines run 6.2.15.
When you go to the BOINC website and click on the download link you will be offered the current recommended version for your OS. If you want a somewhat wider (but far from complete) choice, choose the 'all versions' link at the bottom of the page. If you want to choose from every single version that ever existed even for a fleeting moment in time, they don't make it easy for you to shoot yourself in the foot - pun intended with a nick like yours :-). You have to manually enter the URL boinc.berkeley.edu/dl/ into the address bar of your browser and then scan through and carefully select the appropriate version to fit your needs. It's a huge list with all sorts of oddball and experimental versions and it would be quite easy for the uninitiated to get it wrong.
Cheers,
Gary.
I gate-crashed the v6.6.xx
)
I gate-crashed the v6.6.xx testing process at about v6.6.12, with a very hot-headed opening post when David Anderson mentioned in passing that the v6.6 range didn't support Anonymous Platfrom for CUDA apps. After that somewhat rocky start, I've tested pretty much every version since v6.6.14 (which did support a_p :-) ).
I don't find the scheduler to be anywhere near as bad as Paul has reported. Sure, there were bugs - I've found and reported my fair share - and there are probably still bugs. Along the way, there were some complete disasters - don't even think about .34 or .35, even though they'll still be in the dl directory - but most got gradually better.
I think that, like the rest of us, BOINC v6.6 doesn't react too well to stress. If you are attached to a modest number of projects, give them reasonable resource shares, and don't try to cache too much work, you won't notice any problems. But if you push the envelope too far in any or all directions, the cracks may start to show and you may get as frustrated as Paul.
RE: I gate-crashed the
)
The differences between Richard and myself are that my computers tend to be "wider" than his and I also tend to do more projects per machine (30+) than he. And so, yes, I tend to expose more cracks ...
The problem is that my mind is quite literal and the UCB glossy advertising is that you can run multiple projects and unattended and on and on and on ... so I take them at their naked word ... and sadly you cannot.
As I have said in other threads I am having decent luck with 6.10.3, 6.10.7, and 6.10.11 with an operational span more like Richard's (WCG has my sole CPU attention, GPUS are another matter).