Crash recovery is the recovery of a single-instance database that occurs failure, or failure after all instances in the RAC database have occurred. After the RAC database crash, the first restart of the RAC in instance is responsible for crash recovery.
Instance recovery refers to the failure of a RAC environment in which the remaining surviving instance of the Smon process has been instance in the case of recovery. View gv$instance_recovery can be monitored when instance is restored
Instance/crash recovery Neither requires an administrator to participate in the manual.
The goal of the Instance/crash recovery is to restore the data blocks that were modified in the cache when the Instance was aborted and to close the redo thread that is still in the open state. Instance/crash recovery only uses online log files and current online data files.
The Instance/crash recovery consists of two stages: roll forward and roll back. Roll forward applies the committed, uncommitted transaction records in the online connection log file to the online data file; Rollback is the change to the original state of the uncommitted transactions.
Instance/crash Recovery has the following common features:
-Redo changes to the current online data file
-Only online connection log files are required, no archived logs required
-The required time and the number of instance that occurred failure, the amount of redo generated (since the last checkpoint), the user's configuration (such as the size of the redo log, checkpoint frequency, whether to open parallel, etc.)
The process of Instance recovery:
1. When performing an instance recovery, Oracle recognizes the last checkpoint that occurred.
2. From the last checkpoint, apply those and write to the online log file, but not yet write the data file. Use of the RBS and user data in the roll forward phase
3. The roll-forward ends, the database is opened, and the user can connect to the database. However, uncommitted transactions lock some data
4. Performing a rollback operation
5. End of recovery process
Instance/crash Recovery of the database