2 Replies Latest reply on Jan 19, 2009 11:00 AM by danieltravin

    Configuration issue

      Hi all,

      I have a object tree like this inside my jboss cache cluster
      /country/city
      So, there are many countries and many cities inside each country.

      I need to have only one JVM on the LAN that is aware of the whole object tree, so i use a default config for that.

      I also have a lot of other JVMs but they are logically linked to a "city" and "country"

      So, 1 single JVM can manipulate with only 2 nodes for example /Argentina/BuenosAires and /Argentina

      Basically i d like to say to cache
      "I want to replicate these nodes only Node1, Node2. Do not replicate other, i will not use them"

      What element in configuration is responsible for this?

        • 1. Re: Configuration issue
          manik

          Hmm - if there is no other shared data at all, you could use non-clustered caches in all of your JVMs that actually use the country/city data, configure them to use a TcpDelegatingCacheLoader, and then set up a single cache on a single JVM with all of your data, and stick a TcpCacheServer in front of it.

          • 2. Re: Configuration issue

            I think I understand you advise, but i doubt whether this appoach synchronize data between JVMs that use same node if none of them is a coordinator?

            Example of JVMs and nodes that they use

            Coordinaror JVM
            /UK
            /UK/Cardiff
            /UK/London
            /UK/Bristol
            /Spain
            /Spain/Madrid
            /Spain/Barselona

            JVM1:
            /Spain
            /Spain/Madrid

            JVM2:
            /UK/Bristol
            /UK

            JVM3:
            /UK/Cardiff
            /UK

            JVM4:
            /Spain
            /Spain/Barselona

            So, when JVM3 changes a leaf of /UK, the instance of JBC must propagate changes to JVM2 and to coordinator immediately.
            As far as I see from the source code TcpDelegatingCLassloader will send changes only to coordinator JVM node through CacheServer.

            And does this TCPDelegatingCacheLoader make use of UDP and multicasting?

            My problem is that I am trying to cluster a TreeCache between JVMs on the LAN. But I do not want to share the entire tree between all members of the cluster in order to reduce traffic.

            The configuration I expect to exist is like this
            For coordinator JVM

            <clustering mode="replication" clusterName="MySmartCluster" coordinatorNode="true">



            For JVM1
            <clustering mode="replication" clusterName="MySmartCluster" coordinatorNode="false">
            <fgnlist>
             <fqn>/Spain</fqn>
             <fqn>/Spain/Madrid</fqn>
            </fqnlist>


            Etc.

            I think my question is clear.