Problem with Loadbalancing of Sateless Session Beans across

[tvsrkp]

I have setup a cluster with two nodes. I have a state less ejb which works as facade. I have clients installed on many machines which uses HA JNDI to lookup this state less ejb and call some methods on it. I am using RoundRobin policy and it works fine if one client sends multiple requests. each request would go to different server instance. But if two different clients sends two requests, both the requests would go to the same server(coz both the clients lookup for ejb separately).
How can i achieve RoundRobin across the clients.
I thought using FirstAvailableIdenticalAllProxies logic would do the trick. But i think AllProxies means all proxies for a given view id (FamilyClusterInfo have current view id in it.).
Somebody please give me direction. Thanks in advance.

[ben.wang]

I have never tried it myself. But how about RandomRobin? Will that meet your need?

[jboss2005_01]

Interesting point you mention. I guess this is normal since EJBs rely on the dynamic proxy to decide to which node the request from the client application should be redirected. It actually contains a list of all nodes and this list is being iterated when you use round robin.

Activate load-balancing between different clients should be nice but I don't think you can actually do something about this behaviour. In other words, I don't think it's possible to control this... unless you start using hardware loadbalancers instead of software loadbalancers, but I'm not 100% sure about it :-S

Not a real response to your question, but perhaps it might be usefull for you...


Kurt