| [jetty-users] Diagnosing cause of crash |
Hi! My company - http://www.ninjapost.com - provides forum software where threads update in real time thanks to Jetty and Google Web Toolkit. However, I've noticed Jetty dies when there are as few as 50-75 concurrent users. Our server has 1024 MB RAM, serves 10,000+ pages per day, and has not been re-started for at least one year. Jetty version is 6.1.3. Jetty runs on port 8080 is triggered by Javascript and PHP on the front end of the site. The primary error message is: "exception while dispatching incoming RPC call ... null pointer exception". I'm not sure how to interpret this but uploaded some screen caps and a log file: - http://www.ninjapost.com/crash/jetty_normal.png -- Jetty working normally - http://www.ninjapost.com/crash/jetty_crash.png --- Jetty after a crash - http://www.ninjapost.com/crash/firebug1.png - http://www.ninjapost.com/crash/firebug2.png - http://www.ninjapost.com/crash/firebug3.png - http://www.ninjapost.com/crash/excerpt_from_2012_02_28.request.log Most traffic comes from the site http://forum.simplyshredded.com. You can take a look at the long polling Jetty calls in Firebug when viewing a thread. I previously had real time updates to the main page ("all threads" page) to detect new threads but I turned this off to reduce the number of connections and prevent Jetty from crashing all the time. My sense is that Jetty is most vulnerable when it needs to push a new post with lots of content to multiple subscribers but have not been able to test this theory. Any ideas to fix this problem? My initial fix is to write a script that re-starts Jetty any time it crashes but I don't understand why it crashes under a relatively light load. Mike -- Mike Wilt Ninja Post - Killer forum software http://www.ninjapost.com mike@xxxxxxxxxxxxx (404) 539-8032
