Problems with BOINC benchmarks?

Message boards : Number crunching : Problems with BOINC benchmarks?
Message board moderation

To post messages, you must log in.

AuthorMessage
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1953754 - Posted: 5 Sep 2018, 2:01:57 UTC

I am running with no changes to my BOINC or other programs that run (EVGA Precision, SIV, Win Task Manager, BoincTasks) with BOINC.

On the machine in question, when the benchmarks are run, BOINC crashes and restarts. I lose a couple of minutes of run time, but no other problems are apparent. If the benchmarks are at BOINC startup, this does not happen, just when BOINC decides (at random?) to run the benchmarks.

1) Any knowledge of why this would happen?

2) How can I stop BOINC from running the benchmarks to see if this is indeed the cause of my problem?

3) And how often does B run the benchmarks anyway?

Thanks for your help!
ID: 1953754 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1953755 - Posted: 5 Sep 2018, 2:17:32 UTC - in response to Message 1953754.  

In the options section of cc_config add ...
<options>
<skip_cpu_benchmarks>1</skip_cpu_benchmarks>
</options>
ID: 1953755 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1953811 - Posted: 5 Sep 2018, 8:37:33 UTC

There's a known problem in the recent v7.12.1 BOINC release which causes it to crash during benchmarks. It's connected with the newly-added feature:

* Windows 10: detect and report Windows Subsystem for Linux (WSL)
It should be fixed in the next BOINC release, but in the meantime you can:

Disable the automatic benchmark runs, as Brent suggests
Revert to the previous version of BOINC
Build your own copy of BOINC using VS2013 - the developer of the new feature missed it because there's no problem with updated build tools, but the release versions are built with VS2010 and the problem there wasn't spotted in testing.
ID: 1953811 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1953820 - Posted: 5 Sep 2018, 10:34:04 UTC - in response to Message 1953811.  

Richard - thanks for the info. Interesting.

However

1) I am running older versions of BOINC, not 7.12.1 (7.6.22 - has the problem, 7.6.33 - no problem)

2) Win 7, not 10

I have disabled the autobenchmarks, and will see.

Can you tell me what triggers the benchmarks? Is it a specific elapsed time interval, and, if so, what is it?

Thanks!
ID: 1953820 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1953823 - Posted: 5 Sep 2018, 11:32:02 UTC - in response to Message 1953820.  

I'll have to go and do some research to answer your questions about the older versions - v7.6 was several years ago now. But while you're waiting...

The WSL detection code runs on all versions of Windows. Obviously, on anything except Windows 10, it always gets the answer 'not installed', but the possibility to trigger the crash is still there.

The benchmarks used to run on a calendar - every 5 days, from memory. I can't remember when it was changed - I'll research that too - so I don't immediately know whether v7.6 behaves like that. The new versions only run benchmarks on startup, and only if the benchmark hasn't run within the same calendar delay. If you leave BOINC running continuously all month (as many here do), benchmarks will only run when you restart after the security updates.
ID: 1953823 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1953848 - Posted: 5 Sep 2018, 14:48:09 UTC
Last modified: 5 Sep 2018, 15:02:03 UTC

I've found an interesting old thread at BOINC: http://boinc.berkeley.edu/dev/forum_thread.php?id=8943. It starts off talking about task switching, but then we got on to automatic benchmarking.

Automatic repeat benchmarking should have been disabled from v6.11.8 (September 2010), but owing to a bug, it was still happening during that discussion in March 2014.

Now I'm going off to look at the current logs and code, to see if it still exists.

Edit - looks like it was cleaned up after that March 2014 discussion. See Client: remove unnecessary CPU benchmarking from April 1st (sic!) 2014.

Everything should be OK from v7.3.15 onwards.
ID: 1953848 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1953853 - Posted: 5 Sep 2018, 15:11:42 UTC - in response to Message 1953754.  

Back to the original question.

... when the benchmarks are run, BOINC crashes and restarts. If the benchmarks are at BOINC startup, this does not happen, just when BOINC decides (at random?) to run the benchmarks.
BOINC (your BOINC) should only run benchmarks at startup, and then, only if it's been over 30 days since the last benchmark.

I'm seeing lines like

21-Jun-2018 16:05:42 [---] Last benchmark was 49 days 21:41:40 ago
in the event log after startup, and benchmarks are duly run when that happens. So far, so good.

But in the BOINC discussion, Jord mentions that benchmarks are also run when you change the BOINC configuration manually - specifically, when you increase the number of CPUs BOINC is allowed to use. Could something like that be happening?
ID: 1953853 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1953909 - Posted: 5 Sep 2018, 22:07:50 UTC - in response to Message 1953853.  

But in the BOINC discussion, Jord mentions that benchmarks are also run when you change the BOINC configuration manually - specifically, when you increase the number of CPUs BOINC is allowed to use. Could something like that be happening?


The number of CPUs was increased, but a few months ago. Both machines went from dual 8-core Xeon E5-2670 (v1) to dual 10 core E5-2680 v2.
The machine that doesn't have the problem ran 24/7 for 140+ days, until I cleaned it recently. The machine with the problem has been crashing BOINC regularly (apparently with running benchmarks) every few days (about a week or so, which is why I asked about internal benchmark interval).
ID: 1953909 · Report as offensive
Profile bloodrain
Volunteer tester
Avatar

Send message
Joined: 8 Dec 08
Posts: 231
Credit: 28,112,547
RAC: 1
Antarctica
Message 1955475 - Posted: 15 Sep 2018, 6:51:15 UTC - in response to Message 1953754.  

could be the screen saver glitch. i had that happen before.
ID: 1955475 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1955485 - Posted: 15 Sep 2018, 11:28:40 UTC - in response to Message 1955475.  

I don't run the screen saver.
ID: 1955485 · Report as offensive
Profile bloodrain
Volunteer tester
Avatar

Send message
Joined: 8 Dec 08
Posts: 231
Credit: 28,112,547
RAC: 1
Antarctica
Message 1955575 - Posted: 15 Sep 2018, 21:17:00 UTC - in response to Message 1955485.  

i seen it where you choose to no have it. but some how it still runs. that was a bug i seen before..


but this is the only software giving this problem then?
ID: 1955575 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1955586 - Posted: 15 Sep 2018, 22:04:21 UTC

You could always switch to Linux.....





Runs away rapidly while be pelted with moldy Win7 OS CD-roms....


Tom
A proud member of the OFA (Old Farts Association).
ID: 1955586 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1955587 - Posted: 15 Sep 2018, 22:06:43 UTC - in response to Message 1953755.  

In the options section of cc_config add ...
<options>
<skip_cpu_benchmarks>1</skip_cpu_benchmarks>
</options>


How crucial are benchmarks to the proper automatic decision making of the scheduler?

Tom
A proud member of the OFA (Old Farts Association).
ID: 1955587 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1955603 - Posted: 15 Sep 2018, 22:53:47 UTC

Richard or Jord could definitively answer that I'm sure. I think it is only a little bit crucial for the schedulers. As when a new host is seen and the servers haven't developed an application APR rate yet for it. But I think if an APR is already set, that is what the servers use to decide how much work to send a host.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1955603 · Report as offensive

Message boards : Number crunching : Problems with BOINC benchmarks?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.