I have BOINC Manager set to store at least 0.1 days work and up to an additional 0.1 days of work. And I have an app_config set to 12 max concurrent tasks. Whatever logic sends out tasks was particularly generous this afternoon and has given me 708 tasks, 704 of which are CPU tasks. I have an old 12 core x99 XEON (some random engineering sample), if I averaged 10 hours per task (which I won't) I have 7000 hours of work downloaded. There is no way I can complete what has been sent before its deadline. I will finish as much as I can, but figured I would report this. See Computer 12769859
Copyright © 2024 Einstein@Home. All rights reserved.
bfromcolo wrote: I have
)
Boinc is funky sometimes, just abort what you can't finish and it will go out to someone else instead, they will show up under the "Error" column but don't worry about them. Hopefully over time Boinc and your pc will come to some kind of agreement about how much work you need and kinda sorta keep that amount in your cache.
Hi BFROMCOLO, First, if
)
Hi BFROMCOLO,
First, if you haven't done so, choose NO for your "Run test applications?" on your Einstein's Account webpage, choose Preferences > Project, and then de-select the projects that you ABSOLUTELY can live without. I would suggest just one GPU task, not the 5 that you have. On Einstein alone you have 5 different tasks to down load and your CPU and GPU can't handle all of them. Plus, you have several other tasks listed that your computer will also struggle with while doing Einstein..
After doing the above, save your settings in the lower left corner of the page, and then go to your BOINC Manager and select Einstein and then update it.
Proud member of the Old Farts Association
bfromcolo wrote:I have BOINC
)
Your problem may be due to the use of max_concurrent in app_config.xml. I've never used it but I seem to recall (quite some time ago) that people doing so were experiencing the boinc client repeatedly asking for more tasks every minute under certain circumstances. You should be able to find the details by searching for max_concurrent in the community forums. I don't remember if this ever got fixed. If it did, a search should find it.
Do you really need to use app_config.xml? If you're just trying to limit your 12C/24T CPU to 12 tasks, can you just change your preferences to allow boinc to use 50% of the processors? That's a simpler option.
There is no logic at the project that causes the scheduler to send extra tasks above what was requested. You need to identify why your client keeps asking for more and more work.
I had a look at the last scheduler contact link for your host as it shows on the website. The date showing at the time I looked was 18 Feb 2023 02:48:09 UTC. Your client was requesting 34,560 secs of CPU work. But here are some interesting snips from the full contact log.
2023-02-18 02:48:09 [debug] not updating db prefs because they are not valid
2023-02-18 02:48:09 [version] Skipping CPU version - user prefs say no CPUs
2023-02-18 02:48:11 [debug] ... MSG(high) Invalid global preferences supplied ...
The middle message (repeated multiple times) seems to indicate that you have disabled all CPU searches. The other two seem to suggest that there is some sort of error in the way your global preferences (including any overrides) are stored on your machine. Those are in the file global_prefs.xml and global_prefs_override.xml in the BOINC directory. Your client will send those prefs to the scheduler when it makes a work request. The scheduler is complaining about invalid settings. Have you perhaps made manual edits to either of those files? If you have, you'd better check for typos.
Cheers,
Gary.
The problem with
)
The problem with max_concurrent use causing repeated requests for work exceeding the set daily limits was fixed finally and the code was merged into the latest 7.20.5 clients.
So either remove the max_concurrent statements or upgrade your BOINC to the latest.
Thanks, Keith.
)
Thanks, Keith.
Cheers,
Gary.