2 Replies Latest reply on Oct 2, 2012 5:57 PM by jacob1111

    Strange clustered clients behavior

    jacob1111
    jacob1111 Sep 17, 2012 5:49 PM

      Hi,

       

      I see unexpected behavior of my clustered clients. Instead of adding

      object to the shared cache, my 2nd client overrides the 1st object in

      the shared cache. The 1st client always works properly always adding

      objects to the cache.

       

      If I start the 2nd client first, the behavior reverses accordingly.

       

      Details:

       

      My two clustered clients/embedded servers  run on two different nodes

      using one common cache.

       

      My 1st client (the one that I started the first) stores 3 objects:

      Input new command > list

      Ticket{name='c', show='d'}

      Ticket{name='a', show='b'}

      Ticket{name='e', show='f'}

       

      My 2nd client correctly shows all 3 objects:

      Input new command > list

      Ticket{name='c', show='d'}

      Ticket{name='a', show='b'}

      Ticket{name='e', show='f'}

       

      My 2nd client tries to add new object, but instead overrides the first object:

      Input new command > book

      Enter name: g

      Enter show: h

      Booked ticket Ticket{name='g', show='h'}

      Input new command > list

      Ticket{name='c', show='d'}

      Ticket{name='g', show='h'}

      Ticket{name='e', show='f'}

       

      My 1st client correctly shows 3 objects

      adds another one and correctly shows 4 objects:

      Input new command > list

      Ticket{name='c', show='d'}

      Ticket{name='a', show='b'}

      Ticket{name='e', show='f'}

      Input new command > book

      Enter name: i

      Enter show: j

      Booked ticket Ticket{name='i', show='j'}

      Input new command > list

      Ticket{name='c', show='d'}

      Ticket{name='g', show='h'}

      Ticket{name='i', show='j'}

      Ticket{name='e', show='f'}

      Input new command > 

       

      Question: is it expected behavior?

      I don't think so. Both clients should work the same way.

       

      Here are my nodes parameters:

      cluster.xml file

        

      <infinispan>

          <global>

              <transport

                     nodeName="Infinispan-Node1"

                     clusterName="infinispan-cluster" >

              </transport>

          

          </global>

          <namedCache name="clusteredCache">

         

              <clustering mode="replication">

          

                   <stateTransfer

                      chunkSize="0"

                      fetchInMemoryState="true"

                      timeout="240000">

                   </stateTransfer>

         

                   <sync replTimeout="20000"/>

          

                </clustering>

         

          </namedCache>

       

      </infinispan>

       

      Both nodes run CentOS 5.5 using JDK 1.6_31 and infinispan-5.1.6.FINAL, JGroups 3.0.11.

      The 1st client runs as host, while the 2nd one is a virtual machine on MacBook running 10.7.

      Lion OS. However, they behave the same way when I reverse their roles.

       

      Thank you for your help.

       

      Jacob Nikom