PooledInvokerHA

[joshid]

I have a cluster I am running and am in need of a way to limit the number of RMI threads used for remote calls. I would simply use the pooledinvoker but because the remote calls are over the cluster I assuming that I need to use the PooledInvokerHA. Does anyone have experience using a pooled invoker on a JBOSS cluster?

[joshid]

My apologies this is running on JBOSS 3.2.6 w/ Java 1.4.2_05

[joshid]

To respond to my posting, I've gotten the PoolInvokerHA working.

To do this I copied and pasted the PooledInvoker entry in jboss-service.xml, appended HA to the class name, and made the MBean name unique (e.g., jboss:service=invoker,type=pooledha). Once I had my PooledInvokerHA MBean defined I followed the instructions on http://www.jboss.org/wiki/Wiki.jsp?page=PooledInvokerConfig with the only difference being that I used the PooledInvokerHA MBean for the clustered <invoker-proxy-bindings> entries.

The only issue I ran into was that the PooledInvokerHA attribute, ClientConnectAddress, was initially set to ${jboss.bind.address}. For my environment the ${jboss.bind.address} resolves to 0.0.0.0. So after the JNDI lookup it appeared that the client was attempting to use 0.0.0.0 to incorrectly connect to itself (running the client and server locally was fine but running the client and server on separate boxes caused problems). To resolve I simply removed ClientConnectAddress entry entirely --which I read somewhere then calls InetAddress.getLocalAddress(). With the attribute removed the ciorrect IP of the server was set in the ClientConnectAddress.

After working past the above, the PooledInvokerHA is working well and the cluster is behaving much better under higher loads.