2 Replies Latest reply on Mar 7, 2013 8:19 AM by Randall Hauch

    Experience with Clustering

    Clement Pang Newbie

      I am just beginning to explore clustering Modeshape and I think I am running into MODE-1769 which I thought was fixed in the trunk. I downloaded JBossAS 7.1.1.Final from the website, installed ModeShape 3.1.3.Final, configured a distributed cache as follows:

       

      <cache-container name="modeshape" default-cache="dshell">
          <transport lock-timeout="60000"/>
          <distributed-cache name="dshell" mode="ASYNC" batching="true">
              <transaction mode="NON_XA" locking="PESSIMISTIC"/>
              <file-store relative-to="jboss.server.data.dir" path="modeshape/store/dshell" passivation="false" purge="false"/>
          </distributed-cache>
      </cache-container>
      

       

      And modeshape configured as:

       

      <subsystem xmlns="urn:jboss:domain:modeshape:1.0">
          <repository name="dshell" cache-name="dshell" cache-container="modeshape"/>
      </subsystem>
      

       

      All of these are in the domain.xml for JBossAS on the controller running with full-ha profile using full-ha-sockets.

       

      Upon start-up, we see the exception from the bug: Object of type class org.infinispan.schematic.internal.SchematicEntryLiteral expected to be marshallable

       

      Has anyone tried setting up a simple 2 node cluster using the just the downloads from the website? How did you make it work? I might try to debug into the code to see if the workarounds committed in the code did in fact execute properly (i.e. the interceptor was indeed removed).

        • 1. Re: Experience with Clustering
          Clement Pang Newbie

          After spending some time tinkering with the settings, the way to get it to work is to enable write-behind on the file-store, else the dreaded SchematicEntryLiteral expected to be marshallable would occur:

           

          <cache-container name="modeshape" default-cache="dshell">
                <transport lock-timeout="60000"/>
                <distributed-cache name="dshell" mode="ASYNC" batching="false">
                       <transaction mode="NON_XA" locking="PESSIMISTIC"/>
                       <file-store relative-to="jboss.server.data.dir" path="modeshape/store/dshell" passivation="false" purge="false">
                              <write-behind/>
                       </file-store>
               </distributed-cache>
          </cache-container>
          
          • 2. Re: Experience with Clustering
            Randall Hauch Master

            Clement Pang wrote:

             

            Upon start-up, we see the exception from the bug: Object of type class org.infinispan.schematic.internal.SchematicEntryLiteral expected to be marshallable

            Is it exactly the same stack trace as in MODE-1769? If not, I'd like to know what it is because that likely means Infinispan might have another bug (other than ISPN-2748).

             

            For AS 7.1.1.Final users using synchronous distribution: this issue is should not pose a problem.

             

            For AS 7.1.1.Final users using asynchronous distribution: I think that Clement's solution of using write-behind is the only way to get this to work. (I've updated MODE-1769 with a comment reflecting this.)

             

            For AS7.2 users: this should be fixed, and you should not have to enable write-behind when using asynchronous distribution.

             

            For those not using AS7 at all: the fix in ModeShape 3.1.1.Final appears to work, but we also recommend you override ModeShape's BOM and upgrade to the latest Infinispan 5.2.x version (which has fixed the problem). See our documentation for details.

            1 of 1 people found this helpful