21 January 2004
I'm not particularly happy. I just had to reboot my server for the second time within 24 hours. It just dies with no response. It's running the 2.6.1 kernel, and it was up 11 days before the first crash. I see no logs of anything wrong.
Nearly a day before the first crash, I plugged in and used my USB CD burner for the first time since I booted the new kernels. If it crashes again, I'll try it without the burner.
Update (1:56pm): Andrey volunteered that before each crash, he had issued a stop request to his tomcat instance on my box -- maybe it's an extreme threading issue or something (as java tends to do). I'm off to search linux kernel mail list archives to find some clues. Of course, trying to stop his server again has had no adverse affects. Thank you Andrey for the hint!
Update (10:27pm): After some thorough testing with Andrey, Master of Server Burnination, I got the kernel to OOPS and drop a reference to the "BSD Process Accounting" kernel code. I recompiled without that option and we've been good. I had also tried without "Preemptive Kernel", but it still crashed. I suspect the new process/thread organization could be what's causing the problem in accounting. I'll submit my bug report to lkml.