ajp thread spikes in newly ported application
We recently (last December) migrated a large j2ee application from weblogic to jboss. Additionally, there is an apache webserver in front of the jboss as, with requests being directed on from apache to jboss via ajp. Now, mostly every day, we experience a situation where the number of ajp threads in use spikes up and the users report a major degradation in performance. These spikes typically last about 30 minutes, and then resolve themselves withough intervention. We have been unable to associate this spiking behavior with anything happening in the application, and I'm curious if anyone else has a similar configuration and has seen a similar problem. I'm attaching a couple of images of graphs of the spikey behavior, along with a thread dump taken while the problem is happening and a summary of the thread states (runnable vs. Object.wait). Any thoughts about this are most welcome.
