2 Replies Latest reply on Sep 30, 2009 11:07 AM by 641965

    Unicast communication in a cluster

      Hi all,
      I have a weird problem I'd like your help with.
      I have a domain with two virtual machines, running 3 pairs of clusters. I've setup the cluster
      address for each one, but left the unicast communication channel empty resulting in the default
      channel being used.

      Recently we copied the two virtual machines, changed ip's and hostnames (the two new virtual
      machines are located in a different physical host, than the original ones) and set up one more
      environment for our needs.

      However I discovered today a message in the logs of the 2nd managed server in the first cluster (in the old virtual machines):
      <Sep 23, 2009 6:59:16 PM EEST> <Error> <RJVM> <BEA-000506> <Closing: weblogic.rjvm.t3.MuxableSocketT3$T3MsgAbbrevJVMConnection@6925a2 because of Server expected to route a message received over an uninitialized connection: 'JVMMessage from: '-5925955035663835806S:newvm2.company.com:[7003,7003,-1,-1,-1,-1,-1]:newvm1.company.com:7003,newvm2.company.com:7003:my_domain:ManagedServer2' to: '2423355813275438826S:newvm2.company.com:[7003,7003,-1,-1,-1,-1,-1]:my_domain:ManagedServer2' cmd: 'CMD_IDENTIFY_REQUEST', QOS: '101', responseId: '-1', invokableId: '-1', flags: 'JVMIDs Sent, TX Context Not Sent, 0x1', abbrev offset: '214''>

      Are the 4 virtual machines messing up their unicast communication? How can I avoid this? Cut them off from the firewall, or set a unicast communication channel for each cluster?


      PS: As a naming convention, assume the old vm hostnames are: vm1.company.com, vm2.company.com

      Edited by: dvm on Sep 23, 2009 9:12 AM
        • 1. Re: Unicast communication in a cluster
          By default weblogic server uses multicast for cluster communication and not unicast, are you sure you have configured your clusters to communicate over unicast ? Its not clear from your problem statement.

          If you are using multicast, and havent set any unique values for it for each cluster, then you will end up having some network errors caused by multicast address conflict.

          when setting up several clusters in the same subnet, you should ensure that you have defined explicit listen address to each managed server in the cluster, the cluster address should be defined with no ambiguity (ie taking care of multi IP resolution). When using DNS names ensure that it resolves in a single IP only.

          Unicast communication is one to one and hence the chances of messing up is rare, as long as you have configured things properly.
          • 2. Re: Unicast communication in a cluster
            Hi, thanks for the reply.

            Yes, I'm sure it's Unicast. I forgot to mention that I'm using 10.3 so Unicast is the default there. The docs advise to use multicast only for legacy support.

            I'm afraid of misconfiguration because as I've said the two virtual machines resulted from a copy of the existing virtual machines. So it probably has
            many common values etc, it's not a new installation. I'll take a look at the multi-ip resolution as I've used domain names for the cluster address though.