I am trying to prototype a simple 2 node cluster with JBoss 4.0 on Redhat Linux 9.0 and JDK 1.4.2.
I have 2 nodes (A & B), each running JBoss 4.0. I enabled the HAJNDI and HAJMS on both and deployed the same stuff on both. Both the Nodes have a queue called "testQueue" and a Simple Message MDB deployed watching that Queue.
From a JMS Queue client, I lookup "testQueue" using HAJNDI and start sending messages to it. Based on the HA JNDI lookup, the messages should end up in either of the Nodes. But, they are always going to Node B. The "java.naming.provider.url" is set to "jnp://A:1100,jnp://B:1100".
How does the Node B gets picked always? What algorithm does it follow?
When both the Nodes are up and running, then I start the Queue client to send messages to one of the JMS Queue (on either A or B). Let's say the messages are going to Node B. At this time, if I shutdown the Node B server, the client is getting its onException() method gets called. So, the client tries to lookup the testQueue again and start sending the messages. But, the messages never seems to reach Node A until Node A is restarted. That means, the log messages in the MDB on Node A never says it is processin the beans, until Node A is re-started.
Can any body see why Node A is not processing the messages right away?
Can somebody please give me a pointer with latest stuff on clustering - jms - queues - failover using JBoss 4.0? All I could find so far is references to 3.2.4