I’m working with a two server system that utilizes JBoss (tcp) clustering to enable server failover. After a JBoss server comes up it can not immediately process failovers from the other web server in the cluster. This is because it takes time for the http session information from the other server to be processed. Is there any way to programmatically determine if the http session replication is complete?
In other words, suppose that there are two JBoss servers serverA and serverB. JBoss (tcp) clustering is utilized so that if serverA is restarted all active http sessions on serverA will failover to serverB. Once serverA finishes starting it can not immediately accept http session failovers from serverB. This is because there is a delay while the http session information from serverB is transferred to serverA. Is there any way to programmatically determine if the http session replication is complete and serverA can accept failovers from serverB?
The JBoss version being used is 4.2.3.GA.
Hi Tim, I am not really sure if thats true. If I remember 4.2 correctly, when you start the second server it joins the cluster, deploys the apps and tranfers the state. That would mean that when the server logs "Server started in ... s" the failover should be already possible. So are you saying this is not the case?