I haven't found much, if any info in how you bypass the web to connect directly to an jboss cluster. I see there is a jbossha-client.jar, which I am assuming has to be used somehow. What I can't seem to find is any info how I get my swing client to use the multiple jndi values to round robin between jboss servers. I also don't see much in the way of discussion on different scenarios. For example, should a hardware load balancer be used and a swing client (possibly located at many locations with many different users all using it) only has the single IP to the load balancer? Can a load balancer properly maintain client/transaction session aware across multiple requests? Normally this is the domain of the web server and httpsession, however in our case we use direct calls to the jboss layer. So how is this properly load balanced?
I ask this because some of our clients refuse to open up any of the ports required by JNDI, and others are willing to open up only a single port but require clustering, and so forth. The easy way out is "Hey, it's this way or nothing.." however anyone that takes that stance stands a chance of not having a job in the near future. I'd love to address these issues, perhaps a new doc on client/ejb connection regarding clustering if one is not already available?
have you read the doco? It may help I think. You don-t need a LB or anything: the LB functionality is INSIDE the proxy you receive from your JNDI lookup.