Constant crashes, connection refused on Tomcat restart

[Interkey JeffG]

Running Zimbra Network M4, every couple of hours most of the Tomcat system goes down. IMAP dies (still listening on the socket but times out on connect), and users cannot log into the web interface (can log in, then sits at "Loading ..." screen indefinitely. None of the Zimbra or Tomcat logs contain anything that could be perceived as indicating any sort of problem. When trying to restart Tomcat, a "java.net.ConnectException: connection refused" error is encountered; the process needs to be killed manually, and then restarted. It's driving our testers nuts.

Anyone have any ideas what's up?

[marcmac]

kill -HUP the tomcat pid, and check /opt/zimbra/tomcat/logs/catalina.out for a stacktrace - send us the stacktrace (attach it to a forum post)

[Interkey JeffG]

Originally Posted by marcmac

kill -HUP the tomcat pid, and check /opt/zimbra/tomcat/logs/catalina.out for a stacktrace - send us the stacktrace (attach it to a forum post)

When I kill -HUP the process it just hangs with no output until I kill -9 it.

[KevinH]

Try "kill -QUIT <pid>"

[Interkey JeffG]

kill -QUIT results in the same thing; the process stays hung, doesn't output anything to the logs, and doesn't quit. It has to be killed with kill -9 in order to restart it.

[KevinH]

Interesting... After you restart it can you try a "kill -QUIT <pid>" to see if Java will dump it's thread stacks then. It's pretty rare to get recent JDK's this hung but I have seen it before on older JDK's.

[Interkey JeffG]

Originally Posted by KevinH

Interesting... After you restart it can you try a "kill -QUIT <pid>" to see if Java will dump it's thread stacks then. It's pretty rare to get recent JDK's this hung but I have seen it before on older JDK's.

Yes, it dumps all of the threads correctly after a fresh restart. When it's hung, it's hung hard.

[Interkey JeffG]

Bump because this thing still can't seem to go six hours without crashing.

[dijichi2]

what OS are you running, and have you tried upgrading to GA-156? i had all these problems with the betas but the GA is solid as a rock.

[Interkey JeffG]

Originally Posted by dijichi2

what OS are you running, and have you tried upgrading to GA-156? i had all these problems with the betas but the GA is solid as a rock.

Oh, whoops, I knew I forgot a bunch of critical information in my post.

Zimbra 3.0 Network GA is running on CentOS 4.2 with a 2.6.9 kernel. The system is a 2.4 GHz Pentium 4 with 512 MB of system memory. At the time of the Zimbra crash, there is no abnormal CPU or memory usage, nor anything logged to the system log (though it's hard to tell thanks to Zimbra logging service status for all its services once per minute to /var/log/messages for some reason that I, despite far too much contemplation, still cannot discern).