Unreserving Problem

[Madpoo]

Originally Posted by endless mike

Madpoo's mention of it happening with a mostly idle server makes me think that maybe it has something to do with the CPU dropping down to its idle speed. 800 MHz instead of 3.4 GHz can be a huge difference when computing estimated completion dates. I've had it happen occasionally with my cluster when adding some triple checks for others. The estimated completion date reported to Primenet will end up over 200 days, and then after a day of running will be back to normal.

That's not a bad theory, however all of my servers are set to "high performance" mode so they should be at full clock speed all the time. I like having responsive systems that don't throttle up/down. That bugs me.

I actually just had it happen again today... restarted a system after installing updates and the rolling average went back to 1000. I could understand if it thought it was running slightly slower or whatever, but it seems telling that it went back to *exactly* 1000, like just resetting to default.

Oh well... on the hex-core CPUs doing all 6 cores in one worker, I set it to 4000 and we're back in business. It should actually be a bit more than 4000... it's still off by like 10% or something but it maxes out at 4K.