That is correct. Connection Managers are local to each node and are not visible to other nodes. All a node knows is in which node a user session lives. So if nodeA knows that a user session lives in nodeB then nodeA will contact nodeB. It's up to nodeB to use the connection manager connected to nodeB for delivering packets to users connected through that connection manager.
Which problems are you seeing with this solution? BTW, note that a DNS server even though it does round robin to distribute the load it will fail to recognize when an IP address is not available.
Ah, ok so that is by design. I just wanted to be able to see what
connection mangers are connected to the cluster - I guess I'd need to
check each node individually for that info.
I know that a dns server will not know that an ip address isn't
available -- we reconnect when an connection manager isn't found. Or do
you mean that this is an issue when a node drops and a connection
manager is still connected?