5 Replies Latest reply on Apr 18, 2013 7:16 AM by mgencur

    infinispan and failover

    fnossair
    fnossair Apr 16, 2013 2:12 PM

      hi

      i have installed mod cluster with apache and JBoss 7 and apache.

      the problem is that the response time is so long.

       

      when i drop

       

      <cache-container name="web" aliases="standard-session-cache" default-cache="repl">

                          <transport lock-timeout="60000"/>

                          <replicated-cache name="repl" mode="ASYNC" batching="true">

                              <file-store/>

                          </replicated-cache>

                          <replicated-cache name="sso" mode="SYNC" batching="true"/>

                          <distributed-cache name="dist" mode="ASYNC" batching="true">

                              <file-store/>

                          </distributed-cache>

                      </cache-container>

       

      the response time become well but i have no failover.

       

      can any bady have any idea to have a goor time response and the failover.

       

      thanks.

      • 629 Views
      • Tags:
        • 1. Re: infinispan and failover
          mgencur
          mgencur Apr 17, 2013 3:11 AM (in response to fnossair)

          Hi Nossair,

          in my opinion, the biggest time consumer is the <file-store/> which is synchronous and so all the data that is written to the memory is also written to disk at the same time. Is it crucial for your application? Cannot you remove it or at least make it asynchronous (write-behind attribute)?

           

          Martin

            • Actions
          • 2. Re: infinispan and failover
            fnossair
            fnossair Apr 17, 2013 7:26 AM (in response to mgencur)

            thanks Martin.

             

            i remove it, but i still have a big response time.

             

            the thing that i see, is when one server turn, all seem fine and i have a goor time response. but when i turn on two server, the time reponse is so long. so i think that the share of data between servers that cause this problem.

            what do you think?

            there is a way to say what servers share? the objects shared between servers.

            Nossair

              • Actions
            • 3. Re: infinispan and failover
              mgencur
              mgencur Apr 17, 2013 1:01 PM (in response to fnossair)

              What does it mean "big response time" ? How many ops/s do you see? You can see an order of magnitude difference in response times when there's a replication over network as opposed to local mode where everything is stored in local memory.  You can't say what to replicate to the other server unless you use AtomicHashMap or another implementation of DeltaAware interface but that's probably not what you are looking for.

                • Actions
              • 4. Re: infinispan and failover
                fnossair
                fnossair Apr 17, 2013 3:20 PM (in response to mgencur)

                if i dont use cache web, i have a page that can be charged at 1s. but if i activate cache web the page is charged at least at 40s (that is a big response time for me. i'm not in a production env).

                 

                how can i manage the infinispan cache? how can i see what objects are in the cache?

                  • Actions
                • 5. Re: infinispan and failover
                  mgencur
                  mgencur Apr 18, 2013 7:16 AM (in response to fnossair)

                  This is really strange. The replication cannot normally take so long. It does not seem like a problem of replication but it's hard to say anything else without detailed logs and more information about the usage. ....You can see a few statistics of the cache over JMX - like number of entries in it. But you won't see individual entries and their values.

                    • Actions