| < Day Day Up > |
|
Database availability can be taken out of the hands of the DBA when an outage is related to a hardware problem on the database server. At Horatio's Woodscrews, they have had users complaining about losing their connection to the database intermittently. The DBA could find nothing wrong with their setup, but the system administrator ran some diagnostics that pointed to a flaky network interface card. The solution is to replace the network card as soon as possible. Doing so means taking the database down, shutting down the entire server, and replacing the card. With month-end reports running around the clock in order to complete on time, shutting down the database will mean that the reports will have to wait until after the database comes back up.
With Real Application Clusters, multiple nodes are accessing the database at the same time. As such, if a node fails due to hardware or operating system problems, that node can be taken offline and repaired while users are still accessing the database through the remaining nodes in the cluster. Once repaired, the node can be restarted and will rejoin the cluster automatically, making the instance again available to users. We discuss the setup and configuration of a RAC cluster in Chapter 4.
With Real Application Clusters and transparent application failover, should an instance on one node crash, it is possible for users connected to that instance to automatically fail over to one of the remaining nodes in the cluster, and have queries be restarted-and then to continue on uninterrupted. This is discussed in Chapter 11.
| < Day Day Up > |
|