Hi everyone,

This past Monday we upgraded from GW7 SP3 HP1 to GW8 HP1, and are having some stability issues with the GW webapp. On a fresh boot of the server, GW WA loads fine, and you can access it without any problems. However, after some random amount of time, the webapp unloads itself and users get a "503 Service Unavailable" error. After that, you can't just stop and restart Tomcat, you have to restart the server to get the webapp to come online again.

There are some odd things in the localhost_log file about some null pointer exceptions (attached,) but nothing that delineates "this is why Tomcat is yakking" because the webapp is online even with the NPE showing on load. From what I can tell from all the logs, the only difference between a "successful" load and a "failed" load is that the failed load has the messages about "removing web application at context path ..." after the NPE; there aren't any other errors in any logs I can find, and nothing appears to show up on the Logger screen. This is happening on two different servers, both NW 6.5 SP7 with all applicable post-SP patches. We are running Tomcat4 on both of these servers. When the webapp unloads, NoRM and Apache continue to function properly on the server. The WA agent continues to function properly with its already-established connections, but doesn't allow additional connections.

One thing that we are doing is we have the WA agents between the two servers linked together for failover (same encryption key and pointing the GWProviders at each other.) The problem was occuring before this was set up though, so I know it's not the instigator.

Any ideas?