Hi JF,
We seem to have hacked our way around the second problem (as you may have seen below), but after 24 hours of perfect stability, the original problem for which I created this thread has arisen again. The CPU randomly spiked from a normal of about 33% to over 750% (8-core machine). I have a 4 thread dumps spanning almost an hour (the first 3 were taken automatically but monitoring script I had, the last I took when rebooting glassfish). After a reboot, glassfish went back to behaving normally. Do these thread dumps suggest the spinning state you were referring to?
I should note some configuration updates we had to make to solve the sluggishness problem described below. We are now running with 1 acceptor thread, 20 initial request processing threads, and 80 max request processing threads.
Thanks!
[Message sent by forum member 'rwillie6' (rwillie6)]
http://forums.java.net/jive/thread.jspa?messageID=339330