This could be an issue in the Java Web Start-specific logic, in the app client container, or in the ORB that the ACC relies on to communicate with the back-end.
We will do some checking on this end as well, but could you please try this:
1. Redeploy the app using the --retrieve <dir-name> option. (If you specified this already when you deployed the app there is no need to redeploy he app.)
2. Copy the ${install-dir}/domains/domain1/config/sun-acc.xml file to another name or location (say, for example, myacc.xml). Edit myacc.xml so that it contains <target-server> elements for each of your back-end instances.
3. Launch the client using the appclient script:
appclient -client <dir-name>/theDownloadedJAR.jar -xml <full-path-to-myacc.xml>
This experiment will help identify whether the Java Web Start-specific logic is involved or whether this is a broader issue that affects all clients.
It's not surprising that the client would always connect to the first node in the list. The list is used only for "bootstrapping" the connection from the client to the back-end. Once connected the ORB on the server should make sure the ORB on the client has the full, up-to-date list of all available instances in the cluster.
Of course, once you take down that node and retry the client you should see the client connect to the second node in the list. And that's what we need to track down.
- Tim
[Message sent by forum member 'tjquinn' (tjquinn)]
http://forums.java.net/jive/thread.jspa?messageID=210923