I have enabled Jetty to log response times. I notice that in our production or load test environment, we have a temporary spike in Jetty response times. Jetty logs would trace response times of 200-400ms for a webservice that would normally take 2 ms. I did not notice particular burst of traffic at that time.
Further i traced through servlet debug logs that we have put in our application. And i never found a log that would reflect the servlet processing took more than 3 ms. For every single invocation which Jetty would report response time as 200-400ms, servlet debug logs would only show 2-3ms of processing.
I am guessing that once servlet execution starts it takes only 2-3 seconds. And latency is due latency between receiving of a request to scheduling of request in a sevlet. So i tried to increase threadpool, even though it did not make sense to increase thread pool. As my traffic comes @ 50-100 RPS and if it takes 2-3 ms to process a request 10 default thread count should be enough. But i did increase the thread count to 20 and lowThreadCount to 25. But still saw spike in response times in jetty logs.
Can anyone throw some light or show some path for further investigation ? I am using jetty-6.1.26
Thanks !
