Message boards : Number crunching : Crappy result configuration
Author | Message |
---|---|
Ananas Send message Joined: 1 Jan 06 Posts: 232 Credit: 752,471 RAC: 0 |
max # of error/total/success tasks 1, 2, 1 One wingman couldn't start, the other one aborted his result - gives me : Validate error Of course, counting aborted results as error results is most likely a bug implemented by a well known BOINC despot - and "couldn't start" is not really a workunit error either - so the best would be to set those limits much higher - but wait ... why did it even deliver the result a second and even a third time if the first one already had an error? One more ancient BOINC bug (oh well, some will call it "feature") Do you even care about your volunteer crunchers? Then fix this bug and those in the old threads that get a refresh from new bug reports nearly everyday :-( |
Polian Send message Joined: 21 Sep 05 Posts: 152 Credit: 10,141,266 RAC: 0 |
Of course they care. They've said on multiple occasions how important this work is. But, realistically, this is such a fringe case that happens so rarely that if I was the same shoes it wouldn't be where I focused my energy first. They have a lot of other very important things to do in that lab, only one of them is maintaining BOINC. |
Ananas Send message Joined: 1 Jan 06 Posts: 232 Credit: 752,471 RAC: 0 |
As this is just a configuration issue, not a single line of code to be written, it wouldn't take much time to fix it. Oh, and I doubt that it is so very rare, I rather think that many people don't notice it as the client doesn't report the error. The problems with the extra long running results giving extra low credits is much harder to fix but as it makes many people unhappy, they should work on it. |
Link Send message Joined: 4 May 07 Posts: 356 Credit: 382,349 RAC: 0 |
max # of error/total/success tasks 1, 2, 1 There's nothing wrong with the configuration, the numbers like that make sense. The real issue is, that the server creates one additional result after receiving 2 errors, which is IMHO not a configuration issue, but a bug in the server code, since new results should not be created after the WU was marked as bad. Also unsent results should be turned into "didn't need" once the WU is marked bad. At least that's how recent BOINC versions work, so it might be an issue with the ancient version they use. So eventually not so easy to fix. The "Too many total results" status itself is not an issue btw, I had that one recently with 1 error, 1 timed out and me, the timed out returned his result even before me, but both our results was validated and we both got credit (without that the credit awarding thing they use here had to help). . |
Message boards :
Number crunching :
Crappy result configuration
©2025 University of Washington
https://www.bakerlab.org