The deadlocks occurred again this morning after four days of runtime. I'm still using the first
patch that was posted for GlassFish bug 3747, but will switch to the updated patch for the
next round of tests. Since it's a production system, I can only upgrade files in the evening,
so the testing process is going to take some time.
The attached file has the jstack thread dump of the application server.
There are 21 NodePollResultWrite message-driven bean threads that will not complete,
and 7 ScheduledRequestWrite message-driven bean threads that will not complete.
Just to be sure I installed the patch right, here's what I did:
1. Recreated the toplink-essentials.jar v2 b58g file with the patched files
2. Renamed the SJSAS_HOME/lib/toplink-essentials.jar file to toplink-essentials.jar.orig
3. Copied the patched toplink-essentials.jar file into SJSAS_HOME/lib
4. Renamed my ACC_HOME/lib/toplink-essentials.jar file to toplink-essentials.jar.orig
5. Copied the patched toplink-essentials.jar file into ACC_HOME/lib
6. Restarted the server and application client
I'm not entirely sure if the classloader for the server is going to properly skip past the
older toplink-essentials.jar.orig file due to its different file extension. Is it possible that
the classloader could still pick up the older file?
[Message sent by forum member 'skalynuk' (skalynuk)]
http://forums.java.net/jive/thread.jspa?messageID=241912