Dec 7th, 2011, 10:34 AM
Here is how we managed to somehow replicate this issue. You need a WebSphere version 7 fixpack 17 and a cluster with 2 nodes all of this is on AIX with IBM JVM.
Your greatest chances of success is to deploy the same application twice with different context, start them both then restart WAS and make sure it starts both of them at startup. It may take a few attempts but it will happen.
We managed to connect a remote debuger to the WAS instance and we had a look at MessageHandlerChain.handlers and saw differences in the way the XML looks and how the handlers looked.
the List<MessageHandler> however had
service-activator 4 completely missing sa3 but having the expected number of handlers.
What was more interesting is that the ServiceActivatingHandler, MethodInvokingMessageProcessor, and MessagingMethodInvokerHelper were different but the targetObject was the same between sa1 & sa3 causing our flow to go nuts.
Hope this helps.
Dec 12th, 2011, 08:58 PM
So, any idea how to fix it?
Dec 20th, 2011, 03:42 AM
Created JIRA bug entry for it, also attached code to replicate
Dec 20th, 2011, 06:08 AM
I replied in JIRA, but you or your friend has sent me the same code over the weekend via email and I've replied to that as well. As of right now the attached test case is not reproducing that's been described here. Please provide more details.