4 Replies Latest reply: Jul 13, 2014 11:39 AM by Marc Fielding RSS

    general question

    user13364377

      Hi All,

       

      In our environment we have a full node exadata box  and we have migrated some standalone databases and RAC databases to  the exadata box.

       

      My question is, eventhough we have clustered databases(4 node) inside one exadata box, how does this help  on a disaster recovery scenario since all the clusters are residing on the same box?

       

      If this one particular box stops working, then how does high availability feature works?

       

      OS:Oracle linux

      DB version:11.2.0.4

        • 1. Re: general question
          Marc Fielding

          Hello user13364377,

           

          I think the key point here is what the definition of a "box" is.  Exadata documentation deliberately avoids this term because of its ambiguity.  A half rack of Exadata X4-2 contains 4 database servers.  Each database server has its own CPUs, memory, etc, so if you configure 4 database servers in a RAC cluster, the cluster keeps running if a single server fails.  There is infrastructure that covers the entire rack like power distribution units and InfiniBand switches, but this infrastructure itself is configured redundantly so that the rack will continue functioning even if a single PDU or InfiniBand switch fails.

           

          HTH!

           

          Marc

          • 2. Re: general question
            user13364377

            Thanks Marc for the response.

             

            What i meant by box is the full set of exadata box that consists of all the database servers(considerng a half rac of exadata). When we purchase an exadata box, it is a single piece of box that consisits of database servers inside. My question or concern is , what will happen if something happens with the power supply unit of the exadata box or during a disaster recovery scenario .

             

            I am thinking with respect to a RAC set up. On a traditional 4 node RAC, we have 4 different nodes, if one node goes down, other nodes will work. But in a exadata setup, we have all the  4 nodes residing  inside on a single exadata box and in a disaster recovery scenario, how does this help on high availability comparing with a traditional RAC set up.

            • 3. Re: general question
              Srini Chavali-Oracle

              For DR, you should have an identical setup at a DR site, with the database being replicated to the DR rack - see the information at Exadata Database Machine MAA Best Practices

              • 4. Re: general question
                Marc Fielding

                Hello user13364377,


                Let's take your example of power supply unit (PSU) redundancy.  Quoting from the X4-2 datasheet (http://www.oracle.com/technetwork/database/exadata/exadata-dbmachine-x4-2-ds-2076448.pdf):


                Exadata Database Machine X4-2 Hardware

                Half Rack

                4 x Database Servers, each

                with:

                ...

                2 x Redundant Hot-Swappable Power Supplies

                 

                The key word here is each: every database server has its own set of 2 redundant PSUs, providing equivalent redundancy to 4 separate, commodity servers.

                 

                HTH!

                 

                Marc