JBossCache 1.2 Still accepting members with Multicast set to

[xavierpayne2]

I am using JBossCache as an mbean within jboss 4.0.1 and in my .xml descriptor snippet shown below:

<attribute name="ClusterConfig">
 <config>
 <!-- UDP: if you have a multihomed machine,
 set the bind_addr attribute to the appropriate NIC IP address -->
 <!-- UDP: On Windows machines, because of the media sense feature
 being broken with multicast (even after disabling media sense)
 set the loopback attribute to true -->
 <UDP mcast_addr="228.1.2.3" mcast_port="48866"
 ip_ttl="64" ip_mcast="false"
 mcast_send_buf_size="150000" mcast_recv_buf_size="80000"
 ucast_send_buf_size="150000" ucast_recv_buf_size="80000"
 loopback="false"/>

I have ip_mcast="false" but in my jboss server log. I am still seeing the other machines on my subnet join in...

INFO [TreeCache] viewAccepted(): new members: [apple:4440, pear:2225, carrot:32883]

Looking at the cache with the tree view it does not seem like state is being replicated to the other nodes (which is good!) But why are they still joining when other jboss instances come online on my same subnet and launch the treecache? I though (perhaps incorrectly) setting ip_mcast to false basically turned off clustering.

[xavierpayne2]

Just to add further details. The other instances that are showing up DO have their respective .xml files configured with ip_mcast="false" as well.

We are on a private development subnet and use the same jboss app server configuration from our own internal cvs server. So all jboss application server configurations on the subnet are the same.

[belaban]

ip_mcast=false means that instead of sending 1 multicast msg to N members, we send N datagrams (one per member).
If you want to turn off clustering, comment ClusterPartition in cluster-service.xml

[xavierpayne2]

Ahhh...

This may be a problem for me then.

We want to support just one deployment in the all configuration (which includes session, and entity beans which are configured to be clustered in their descriptors and thus bound to HAPartition) but for standard usage we want to basically limit the cluster size to 1(as most people will not wish to use the clustering capability).

We were hoping to find a simple descriptor attribute or the like that would essentially toggle the clustering capability on(Unlimited cluster size) or off(limit to a cluster of 1) but not have to change how the ejbs are deployed.

If I comment out default partition I have to redeploy my ejbs with new descriptors as their dependancy on DefaultPartition will fail :(

Maybe we will have to have two different deployments after all (one based on the "all" configuration and one based on "default")

[xavierpayne2]

Can I set mcast to true but then just set the bind address to 127.0.0.1?

Would that limit me to a cluster of exactly 1? :)

[belaban]

Yes, that works

Another possibility is to set ip_ttl=1