Trouble separating clusters

[ericl]

Hi,

I'm setting up several different clusters (development, qa, etc) on the same network. I'm separating the clusters by changing the mcast_addr setting in cluster-service.xml from mcast_addr="228.1.2.4" to mcast_addr="228.1.2.5", etc.


I'm seeing the following scenario:

start up node A run hello world
start up node B with same mcast_addr run hello world
see node A recognize the addition of node B (good)
see clients w/ node A in JNDI.properties load balance nicely between A and B (good)

shutdown node B
start up node B with different mcast_addr in cluster-services
see node A recognize the addition of node B (bad)
see clients w/ node A in JNDI.properties get served by A (good)
see clients w/ node B in JNDI.properties get served by A (bad)



In other words, I can't seem to make a client connect to the new cluster with the different mcast_addr setting. Any suggestions?

Thanks,
Eric

[ericl]

Has anyone set up two clusters on the same network successfully? Even a hint like "I just changed the mcast_addr and it worked for me" would at least give me a clue that I'm on the right track. Right now I don't know if I'm even changing the right setting, or if my address change makes any sense.

Thanks for your time, it's really appreciated.

Sincerely,
Eric

[jason1]

You also need to change the HA-JNDI settings for jnp discovery Group. That might be why your 2 different clusters are sharing the same JNDI.

Jason

[ericl]

"jason@cdot" wrote:
You also need to change the HA-JNDI settings for jnp discovery Group. That might be why your 2 different clusters are sharing the same JNDI.

Jason

Thanks for the suggestion Jason. I tried incrementing the following settings in cluster-service.xml under the "jboss:service=HAJNDI" mbean:

<!-- Multicast Address and Group used for auto-discovery -->
-attribute name="AutoDiscoveryAddress": 230.0.0.4
-attribute name="AutoDiscoveryGroup": 1102

I made AutoDiscoveryAddress 230.0.0.5 and AutoDiscoveryGroup 1103 in my second cluster. I left the other settings like PartitionName and BindAddress unchanged. Sadly, this didn't seem to alter the behavior I'm seeing. I still can't get a client to connect to the second cluster.

Is this what you meant? Have I made a mistake?

Thanks for your time,
Eric