5 Replies Latest reply on Feb 17, 2009 11:34 AM by jbosstcs

    clustering achieved ?

      I have an apache web server in front of some jboss server instances ( running on different m/cs )... i have configured mod_jk to route requests to the app servers... in this respect i have some queries : -
      when the server instances on the different nodes are started , they automatically detect each other... also requests are routed from one to another as in if i shutdown an instance after a request has gone to it, the request is served by another instance...

      1. ) can i then claim to have achieved clustering amongst the servers ? or have i missed something... i.e. do i need to do changes in some configuration files or something like that ?

      2.) is there a concept of vertical clustering in jboss like in websphere or 2 instances of jboss on the same m/c are completely independent ?

      thanks

        • 1. Re: clustering achieved ?

          in continuation of the above architecture what i have now done is separate one m/c from a cluster by changing its PartitionName in cluster-service.xml.... the server logs show appropriate messages indicating the remaining members of the cluster... however failover is still occurring amongst the severs not in cluster... how is that possible ?

          • 2. Re: clustering achieved ?
            brian.stansberry

            Apache knows nothing about how you've organized the backend servers beyond what you configure in workers.properties. In workers.properties when you set up a load balancer worker you list the regular workers it should balance requests across. You need to limit that list to the workers that you want to receive requests.

            • 3. Re: clustering achieved ?

              exactly... apache will not care about how the app servers are arranged... you mean to say that just because the app servers are given in the workers.properties file and load balanced , failover will occur in case one of the constituents goes down ?? why will apache care about the failover part ?
              shouldnt that be the headache of the app servers ?

              • 4. Re: clustering achieved ?
                brian.stansberry

                The load balancer (Apache) is what handles failover. The request comes into apache and it decides how to handle it. If it sees the request's session is associated with a backend server that is no longer available, Apache picks a different server (i.e. fails over). If it passes the request through to a backend server and the request fails with a recoverable problem (e.g. ConnectException connecting to the backend server), it tries a different backend server (i.e. fails over).

                • 5. Re: clustering achieved ?

                  oh... now i get it... but consider the following situation... say i have 2 app servers load balanced by an apache server... after a request has gone to AS1 suppose it goes down... now my question is how does apache know whether AS1 has gone down and route the request to AS2 ( this will not be a case of ConnectException )... i mean this is exactly what i too have observed but i would like to know the reason of the behaviour...
                  does apache keep a track of the requests ?