Well finally, we are seeing results. Sadly one of our jboss servers takes 5 minutes to start, so it has been a very slow day. However, I am happy to say we are seeing our java swing client properly connect/balance between two jboss servers.

We have seen something we can't readily explain though. We specify two ip addresses in the client jndi.properties file. For some reason, the client always seems to try the highest IP address first, every time. When we shut down that server, it takes a little bit of time for the two servers to both spit out "dead member: 1", then the client finally connects to the other server. When we plug the network cable back into the first box, after a short while (10 seconds or so), we see both servers resolve each other again, dead members: 0 shows. From that point on, the client still connects to the 2nd server. Is this normal? I thought the round robin approach was in use?

Oh, we are using JBoss 3.2.2RC1, and I do see the HASingleton-service.xml in both servers /farm dir. So I am going to guess that the Master node thing is working in this case, and that is why we are not seeing round robin? Sacha/Bill, any words on what we are seeing?

The biggest problem, although maybe I misunderstood how farming works, is in redeploying an app. For some reason, when we redeploy it on one server in the /farm dir, it undeploys and redeploys on that server. On the other server I see that it undeploys it and deletes it, but then it never gets redeployed. Sacha, is this a bug still? I know this was the case in a previous thread, so I am thinking it is still a bug. Or is this a configurable option to indicate you may not want to re-farm it? Although that wouldn't make much sense since it breaks the notion of a syncronized cluster.

One of our servers is linux redhat 9, and the other is a Windows XP Pro workstation, if that helps clear anything up.