Hi JF,
I'm still having the intermittent spikes that were the original reason for the post and, because of increasing traffic, they've increased in frequency to every 24 hours or so. They are easy to see coming, because of increases in CPU and Interrupts / Context-Switches in my munin graphs, and can be remedied by restarting the app server. But, without a restart, the situation does not resolve itself. Eventually, the spikes go even higher and the server becomes non-responsive.
As an update, to resolve the sluggishness that we started experiencing after originally creating this thread, I cut the acceptors down to 1 and the thread-count down to 80 (I'm not sure why this helped, since we had been running 5 acceptors and 150 thread-count for 6 months without problem). After doing so, I'm now back to the original problem and am still hoping to find a solution, since waking up in the middle of the night to restart the server after it blew up is getting exhausting ...
I've attached two more thread dumps from when the CPU and Interrupts were escalating and I had to restart the server. Please let me know if these are indicative of the spin state you previously referenced or whether there is any other data I can provide.
[Message sent by forum member 'rwillie6' (rwillie6)]
http://forums.java.net/jive/thread.jspa?messageID=340113