2 Replies Latest reply on Feb 1, 2013 8:42 AM by user2877716

    Reinstatiation  out-of-sync database

      Hi all

      I am replicating a 3TB database from Oracle Single Instance AIX (source) to RAC (two nodes and this is target) Linux.

      The replication is setup and working, the initial instantiation took long time though due to different platform, used transportable tablespaces and the entire process took around 20 hours.

      Now that the replication is setup I would like to avoid run initial instantiation again. I would like to know if this is possible, I will run RMAN backup in both RAC and Single Instance Database plus archived logs and the trails. If for some reason the target is out of sync (for example a 5000 million rows table, in this case is faster restore instead of initial load) can I restore the trails, RMAN backup, recover to a specific scn and run the replicat with afterscn or atscn option? Do I have to reposition the replicat to start from a specific rba? Do I need to "manipulate" the checkpoint table located in the source database?

      Thanks in advance
        • 1. Re: Reinstatiation  out-of-sync database
          Sure, don't let things get out of sync and you'll be fine.

          You can piecemeal your replication. If you put everything into one process group, not such a good idea. If you can group your groups, so to speak, you won't have a single point of failure with respect to the data being replicated.

          Aside from what GoldenGate shows (long running, clobs, pk/fk type of relationships), you can separate schemas or applications.

          Make it easy to access older archived redo logs. If you need to reach back in the past, having the relevant ARLs available makes life easier.

          Probably trial and error, but figure out what your cut-off time is for trying to fix things via GoldenGate via just starting over.
          • 2. Re: Reinstatiation  out-of-sync database
            Hi Steve

            We will try the best to sync the data of course.

            We are currently replicating one schema only. There are a couple of tables which occupies 5000 million and 9000 million of rows and getting around 50 million inserts per table per day, our worry is that if one of these tables are out of sync to resync a table which such activity and amount of data the procedure can be tricky.

            The procedure I mentioned with RMAN is straight forward in terms of Oracle but I have doubts how to deal with the OGG checkpoint table and the OGG trails in the target site when the target database is restored and recovered, how to get these synced with the database just recovered.

            Today is the 3TB database but in a couple of months time if OGG works well we will start another project to replicate a 45TB database.