BOINC - New behavior noted

[Bryan]

Last night (and this morning) I saw something on a 24 thread machine I've never seen before. I had one running WCG FAH2 but I suspended that and brought up Seti V8 from Seti Beta. Initially all 24 threads began crunching but within the hour 8 threads went to "waiting to run". I couldn't figure out what was going on but after about 15 minutes those 8 threads began crunching again so I figured it was just some anomaly.

6 hours later I looked at that machine again and 16 threads were running and 8 were waiting to run. All 24 WU were started at the same time and I had 8 that had run for 8 hours, 8 @ 5 hours, and 8 @ 4 hours so during the night the waiting to run thing had happened multiple times. I rebooted the machine which cleared the WCG out of memory and the same thing happened with the Seti V8. It isn't a memory problem, heat, or disc.

I changed the "use at most" setting from 0, which is supposed to mean no restriction, to 100% and all threads came to life. I've watched it over the past 3 hours and none of the threads have gone to waiting to run.

I then noticed a 2nd 24 thread machine that was running WCG was doing the same thing only slightly different. It began turning in WU. After the 1st 8 were submitted they began crunching new WU. After 20 minutes those 8 WU went to waiting to run. I changed the use at most setting to 100% and all threads became active. Here again in the past 3 hours no threads have dropped off line.

I've never seen this before so I don't have a clue what is actually going on but so far changing the value from 0 to 100% seems to have rectified the problem.

[FourOh]

Most of my machines are running WCG - but I have it set to low priority versus NRG/CAS (trying to catch some intermittent work!). Even at resource zero, WCG is taking over the work scheduling. When NRG updates (using Boinc Updater) it isn't calling for CPU work even when the cache of WCG is only a few work units more than the number of threads.

Definitely something to keep an eye on.