1) Message boards : Number crunching : Some task run longer than usual (Message 3777)
Posted 5 Sep 2019 by Profile Beyond
I agree with EG.
The realities of the DC world that have been demonstrated over and over again:
Less credit = fewer crunchers.
More credit = more crunchers.

Admin, the choice is yours.
2) Message boards : Number crunching : Project has no tasks available (Message 3403)
Posted 24 Feb 2019 by Profile Beyond
Good to hear. I haven't had a problem with this, but glad it's resolved for you. I'm running a very small cache on all of my machines and pretty much exclusively Universe on the CPUs at the moment. I've noticed that problems on project work fetch can occur if I try to keep a large cache or run many projects at once. In that case some projects will download too much for the BOINC settings and others will starve.
3) Message boards : Number crunching : cheating? (Message 3321)
Posted 6 Feb 2019 by Profile Beyond
Ok, some explanation.

This user host was doing wrong and his tasks should be not valid but...
As you remember, I had made server upgrade some time ago. Because of problems with validator I had to use old one, changed a bit by me long time ago. Unfortunately, with new BOINC server software the validator didn't work properly (which wasn't obvious as it was no problems during the time). I'm in contact with the user and it looks like his host was compute with errors but no it is fixed.

One last thing on this subject. While this mess was probably unintentional, this single rather anemic machine was awarded around 17,000,000 credits in the 4 days that it was spewing bad results. Manually adjusting that to a realistic level might be reasonable. Here's the info from BoincStats:

4) Message boards : Number crunching : cheating? (Message 3301)
Posted 5 Feb 2019 by Profile Beyond
I'm working on it.
Why don't you all calm down ??
no reason to panic...

Good general advice for life. The above sounds a bit like an excerpt from a Douglas Adams book. :-)

BTW, I also prefer the fixed credit system because creditnew is crazy and also smells bad...
5) Message boards : Number crunching : Zero credit for valid Work Unit (Message 3300)
Posted 5 Feb 2019 by Profile Beyond
There are hundreds if not thousands of these valid but zero credit WUs. Sounds like the admins are working on it.

Patience is a virtue. ;-)
6) Message boards : Number crunching : Linux new app version (Message 3253)
Posted 3 Feb 2019 by Profile Beyond
Looks like windows is already on v0.12

The Windows version jumped from v0.06 to v0.11 and then quickly to v0.12. All v0.11 WUs failed immediately but v0.12 seems to be running normally. While it's hard to say for sure because of WU variation, so far it seems that v0.12 is a little slower than v0.06.
7) Message boards : Number crunching : cheating? (Message 3251)
Posted 2 Feb 2019 by Profile Beyond
Looks like you ferreted something out. Makes one wonder about the validation routines...
8) Message boards : Number crunching : extreme long wu's (Message 2597)
Posted 6 Feb 2018 by Profile Beyond
As a result of the Meltdown/Spectre U@H shutdown, the density of long-runners has increased (because these are the tasks getting bounced from host to host). I have received four tasks in the last couple of days that have been bounced from host to host. All four have been completed and validated by one host, but with a CPU time of zero seconds.

Exactly what I've observed. Sometimes rebooting starts them again. Often it does not. BTW, I'd call them stuck WUs, not long-runners.
9) Message boards : Number crunching : extreme long wu's (Message 2496)
Posted 15 Nov 2017 by Profile Beyond
I don't think that they're long WUs, they're stuck WUs. So far the ones I get don't checkpoint at all. When I look in the slot folder there is no update in any file even after many, many hours. My sample is very small but so far it's happened on my Intel CPUs but not on my AMD.

Copyright © 2024 Copernicus Astronomical Centre of the Polish Academy of Sciences
Project server and website managed by Krzysztof 'krzyszp' Piszczek