Slow Crash Recovery

---

Even after the DBA restarted the database, it seemed to be hung at the startup command. There were no errors, but when he checked the alert log, he noted that the last line indicated the database was performing instance recovery (also known as crash recovery). This recovery must be done to recover any changes in the redo logs that have not been committed to the datafiles. Depending on the size of the redo logs, and the frequency of checkpoints, this can take a considerable amount of time. When there is a lot of pressure to make a database available as quickly as possible, waiting on a poorly tuned crash recovery session can be excruciating.

RAC Handles Instance Recovery from Surviving Nodes

With Real Application Clusters, should an instance that is part of the cluster crash, the instance recovery is handled immediately by one of the surviving instances in the cluster. We discuss this in Chapter 5. In addition, Oracle10g has, through Enterprise Manager, a new feature called the Redo Log Advisor, which will allow you to automatically tune redo logs so that checkpoints occur at a rate that allows for faster instance recovery in the event of a crash. This is discussed in Chapter 3.

---