Boinc automatically assigning CPU projects when CPU WUs are disabled! BUG!

Message boards : Questions and problems : Boinc automatically assigning CPU projects when CPU WUs are disabled! BUG!
Message board moderation

To post messages, you must log in.

AuthorMessage
ProDigit

Send message
Joined: 8 Nov 19
Posts: 718
United States
Message 98718 - Posted: 21 May 2020, 17:22:04 UTC

I've noticed multiple times now, when a project has GPU projects, and I only enable GPU WUS, (disable the CPU WUs),
and my system runs out of CPU jobs, it will enable CPU jobs for that project, resulting in me needing to abort tens of WUs that got downloaded and shouldn't have!

I regularly disable CPU jobs on Milkyway, Asteroids, Moo!Wrapper, Einstein.
I think it's a bug that started with the newest version (which gives the annoying 'No CPU for project' notice.
That notice sometimes disappears.
Usually this happens when i'm trying to calibrate GPU projects, and suspend all CPU projects I'm running.
I prefer to run CPU only projects on the CPU, and those projects that have GPU WUs, run GPU WUs for them.

I constantly have to go back to the GPU enabled websites of each project, to disable CPU projects.
It's a very annoying bug!
ID: 98718 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15478
Netherlands
Message 98721 - Posted: 21 May 2020, 17:39:24 UTC - in response to Message 98718.  
Last modified: 21 May 2020, 17:40:28 UTC

The client does not change project preferences on the website, it cannot do that, it doesn't have the capability to do that - no write options for those preferences, only for the global computing preferences. Project preferences changing by themselves is something you report to the project. Maybe someone has access to your account.

Also make sure that on the project preferences website you uncheck lines saying things like "If no work for selected applications is available, accept work from other applications?", because if that's set to Yes, that will give you work even for resources you have unchecked.
ID: 98721 · Report as offensive
ProDigit

Send message
Joined: 8 Nov 19
Posts: 718
United States
Message 98738 - Posted: 22 May 2020, 20:34:58 UTC

Thanks, I'll double check that option!
ID: 98738 · Report as offensive
ProDigit

Send message
Joined: 8 Nov 19
Posts: 718
United States
Message 98740 - Posted: 23 May 2020, 2:48:16 UTC - in response to Message 98738.  

So it appears that this option is not available on many projects that I've seen re-enable the CPU.

I am beginning to think that when clicking 'synchronize with BoincStatsBAM' this issue appears.
And since it's many projects, (asteroids, Moo!, Milkyway, Einstein, ...) I don't think the issue has anything to do with the projects.
Some of the projects (like Asteroids today) show CPU disabled on my account page on their website, but still it downloaded CPU jobs.
ID: 98740 · Report as offensive
robsmith
Volunteer tester
Help desk expert

Send message
Joined: 25 May 09
Posts: 1283
United Kingdom
Message 98741 - Posted: 23 May 2020, 7:26:05 UTC

A couple of things to think about that may help.
Cache size - small caches are reported as being very helpful, a "store at least" size of one or less, and "store extra" no more than 10% of that figure.
Resource shares - BOINC struggles has trouble when they are all the same or very similar. BOINC works on a deficit system when trying to work out which project to call on next, and this is deficit over time, not instantaneous and this can be upset by micro managing by suspending a project. Think which project you want to favour and set that one to 1000, then drop the other projects down from that down to the one you are least interested in being set at say 10.
Don't expect all projects to run al the time, let BOINC manage things and it will eventually get into a stable pattern.
ID: 98741 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5078
United Kingdom
Message 98742 - Posted: 23 May 2020, 8:12:04 UTC - in response to Message 98741.  

The old advice (10 years or more ago) was something like:

Look at your active projects, and find the one that sets the shortest deadlines - mine is GPUGrid, at 5 days.

Divide that number by the number of active projects you have sharing the same resource - CPUs only, in those far-off days. So, if you had five projects sharing, and the shortest was 5 days, your absolute maximum cache size should be 1 day, and work down from there.

The reasoning behind this advice was to prevent resource share becoming unbalanced by episodes of EDF ('earliest deadline first') - the technique by which BOINC attempts to reconcile the unreconcilable demands placed on it by users ("cache 20 days of work and return it all within 5 days, when I only turn the computer on for a couple of hours a day"). EDF is how BOINC prioritises work, and runs tasks out of sequence, to get the science done when it's wanted.

But people took no notice then, and I don't expect them to take any notice now.
ID: 98742 · Report as offensive
Profile Dave
Help desk expert

Send message
Joined: 28 Jun 10
Posts: 2518
United Kingdom
Message 98743 - Posted: 23 May 2020, 10:46:13 UTC - in response to Message 98742.  

But people took no notice then, and I don't expect them to take any notice now.


Yep. I remember numerous posts in the early days of CPDN about problems with scheduling and resource share etc. Though with tasks that could take over six months on what were then fast machines, problems with one project that was such an outlier were perhaps inevitable for those running several projects. By only running anything else if CPDN is out of work for me micro managing isn't too much of a hassle and mostly consists of suspending main site stuff when tasks appear from the testing branch.
ID: 98743 · Report as offensive
ProDigit

Send message
Joined: 8 Nov 19
Posts: 718
United States
Message 98749 - Posted: 23 May 2020, 15:05:07 UTC
Last modified: 23 May 2020, 15:07:03 UTC

project priority for me is between 7.500 and 0.
Most values are:
0
1
100
1000
2000
4000
5000
and 7500

And are based on project priority based on my own preference on how WUs scale to other projects I find less important; and how much a project has been stable.
Asteroids for instance dropped down to 1000 after it had issues with WU uploads, and I still have hundreds of WUs pending to be verified. Such projects drop in priority rating compared to the others.
Only 2 projects (1 CPU, 1 GPU project) is rated at 7500.
Einstein rose from 1000 to 7500, after it now runs good and fully (@90%) utilizes my RTX2080Ti GPUs.
Important projects remain at 5000,
And projects that I want my PPD score to more or less stay the same, I set to 100, so a few WUs per day are processed, and my rank won't plummet..

What I wanted to say is that Asteroids isn't even close to being a number one priority project.
ID: 98749 · Report as offensive

Message boards : Questions and problems : Boinc automatically assigning CPU projects when CPU WUs are disabled! BUG!

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.