"Captain, she's only hitting on 4 cores...."

[PoppaGeek]

dual Operton 8356, 4 cores each. Running Debian. Has been crunching for over a year close to 2. No problems.

Was running a mix of WCG and E@H, 50/50. Noticed one day it was running 4 E@H and 2 WCG with more of both in queue. Rebooted and all 8 used.

Yesterday I noticed again 1 WCG and 4 E@H with plenty of both in queue. Rebooted. Now running 4 E@H. All else Waiting to run. Since then I have:

Shut off and restarted. uninstalled Boinc and reinstalled. Checked BIOs even though I had made no changes. All looked ok.
Checked settings at E@H and WCG. Detached from WCG, aborted all tasks, removed any reference to WCG in Boinc files. Changed settings like % of CPU and then back. Checked all settings on web site and Boinc multiple times. Another dual 8356 is running same E@H and WCG profile and no problems. Checked for heat.

With Top I notice that both CPUs are being used, 2 cores each and which cores being used does change, more than 15 minutes and less than 30. Installed Ubuntu 12.0.4.2 and same. Logs show no errors. dmesg shows all 8 "cpus" detected and started, Boinc Event Log shows 8 cores.

It has a good Corsair PSU, will check that next, maybe CPUs not getting enough juice?? I am at a loss. Do not know what to do next. Have googled everything I can think of and nothing.

Ideas?

[Evantaur]

what does top say... having zombie processes?

[PoppaGeek]

None. No zombies

[D_A]

Check and reset your BOINC preferences for WCG. It sounds like they might have been corrupted somehow.

[Evantaur]

you could try to borrow cpus from your other rig and see if it's still acting up with those

[OldChap]

Heat? When in failed condition what does system monitor show under resources? (not used debian before so made a couple of assumptions here)

[Evantaur]

Heat? When in failed condition what does system monitor show under resources? (not used debian before so made a couple of assumptions here)

dmesg would have spam like this in it

[ 1199.201405] CPU4: Core temperature above threshold, cpu clock throttled (total events = 1)
[ 1199.201406] CPU0: Core temperature above threshold, cpu clock throttled (total events = 1)

[D_A]

Yes, I should have thought to get him to post dmesg output when it's happening. Might be some clues there.

OC, you might want to copy and paste the output into a text file and attach it, as it can be quite large. Alternatively use the command

dmesg > dmesg.txt

and it will send the output to the file for you, then just post/attach/link to the file.

[PoppaGeek]

While it certainly did look like CPU throttling due to heat there was no other evidence of it. Logs and dmesg showed nothing. In the end I again uninstalled Boinc but this time deleted the /var/lib/boinc-client directory. The data directory. Reinstalled and all good now. Looks as if there was a corrupted file somewhere in the data directory.

All good now:


Thanks for the ideas of things check!