I've seen
alot of people try to do this type of thing with WebSphere, but never with any true success. Requests get sent to that 'backup' server regardless of how the weights are set.
The weights can control which server the majority of requests will go to, but it will never, from what I have seen, cut a server off completely. The plugin is pretty liberal with regards to when it consideres a server 'down.' Just a slow repsonse from one cluster member will cause WebSphere to
spray requests to another server. This is good fo clients, but it stinks for us when we are trying to control the environment.
To truly implement a configuration like this, clients have implemented a switch in front of the cluster, that will spray requests to the second cluster member only when, according to pre-defined rules, the first server is down. But this is a pretty complex solution.
Cheers!
-Cameron McKenzie