3 Replies Latest reply on Mar 29, 2019 12:05 AM by Tmiddlet-Oracle

    Federation resync

    dadashy

      with Federated caches, my understating is that in the event of a cluster reconnection, we don't automatically get resync. and we have to manually perform a replicateAll

      is there any custom way that one can automatically fill the gap, assuming that there is some sort of unique sequence number available.

      and what does the replicateAll do? do we end up resending everything?

        • 1. Re: Federation resync
          Tmiddlet-Oracle

          Hi.

           

          When a destination cluster (participant) is reconnected, the source cluster will send any inserts/updates/deletes that were raised since the disconnect of the participant. (assuming that it was an active participant before)

           

          In 12.2.1.2 a new operation was added called startWithSync which will start a participant and do a replicateAll operation. This assumes that a stop operation was carried out on the participant.

           

          https://docs.oracle.com/middleware/12213/coherence/manage/oracle-coherence-mbeans-reference.htm#GUID-F7BB7F6A-93E7-4D36-…

           

          ReplicateAll does just that, replicates all cache data in the current cluster to the destination participant.

          • 2. Re: Federation resync
            dadashy

            just trying to reconfirm my understanding.

            Assume an active-passive situation. lets say the replication get interrupted:

            1- due to a network issue between source and destination cluster

            2- due an outage in destination cluster - has to be restarted for some business reasons etc

            3- due to a controlled stop. ie the "stop operation" you mentioned above was somehow triggered

             

            when destination cluster comes up. how can we ensure an "automatic" resync of the destination caches for each use case above

             

            thanks

            • 3. Re: Federation resync
              Tmiddlet-Oracle

              1- due to a network issue between source and destination cluster.

              Assuming that the destination cluster is still up and running, then when the source reconnects, any unsent changes will be sent.

              2- due an outage in destination cluster - has to be restarted for some business reasons etc.

              In this case if active persistence is used, then then after the destination cluster is restarted and recovers, scenario 1 will apply. If the destination is not using active persistence then a replicateAll must be sent.

              3- due to a controlled stop. ie the "stop operation" you mentioned above was somehow triggered

              by issuing startWith Sync

              You mention: when destination cluster comes up. how can we ensure an "automatic" resync of the destination caches for each use case above.

               

               

              Because of the nature of different scenarios, some of you describe above,  there are a number of permutations, so we don't have a automatic sync in case this is not required.

              There is the ability to set the initial-action for a participant of stop/start/pause after which the require action can be taken.