0 Replies Latest reply on Jun 16, 2003 9:48 PM by Kevin

    It's alive!

    Kevin Newbie

      Alrighty, regarding my other thread: http://www.jboss.org/modules/bb/index.html?module=bb&op=viewtopic&t= it appears I finally got them talking with one another. Today, I can go home and sleep.

      What I am hoping to do tomorrow is get our client application to talk to either and/or both of the clustered servers, and get our server app deployed on both. We have a Swing client app, using the jboss-client-all.jar and such to talk to our jboss server directly. We are not going through web services or jsp/servlets.

      So, while I have read the doco a bit, I am still unclear about how a client Swing app is able to load balance to either of the two servers. From what I understand, RMI itself, or is it JBoss that sends a "smart" rmi stub to our client indicating the server names/ports to look jndi up through? I recall it saying something about each server first looks in its local, then the other nodes local, then in a global shared, but don't quite understand what that is about?

      Do I have to do anything extra on our client, other than provide the two IP addresses via the jndi.properties file to the two clustered nodes? Does the jboss-all-client and such libraries automatically handle load balancing (round-robing or random if I read correctly) when I assign two jndi server ips via the jndi.properties? Or do I need to make some other settings?

      I am looking forward to seeing this work! I am hoping I can deploy our server .ear file to each node, see it get deployed, then run the client and watch it connect to both of them (round robin).

      What are some ways ya'll test this configuration? I am looking at doing two nodes in the one partition for now, and making sure that via round robin my client is able to hit both. Then I'll bring one node down, and make sure that all requests go to the other node. Then bring it back up and make sure that request are again submitted to both via round-robin, and then the same routine for the other node. Does that sound alright?