An instance error occurs when the Oracle database instance (System Global Area (SGA) and the corresponding background processes) can no longer run.
An instance error can result from a hardware problem (such as a power failure) or a software error (for example, the crash of the operating system or of an Oracle background process).
An instance error generally results in an immediate abnormal termination of the entire instance. Even if the database system remains active, the data in the SGA is lost in any case, and the instance can no longer be shut down in the conventional way (withshutdown normal or shutdown immediate ).
Since only an abnormal termination is usually possible (shutdown abort ) the instance must be recovered. Only transactions completed normally ("COMMIT") can be processed; all others are rolled back. If you are working with a standard configuration, the system monitor (SMON) of the database system automatically carries out the recovery of the instance when you restart the instance (instance recovery). It uses the entries in the appropriate redo log files to do so.
You do not need to intervene during the recovery, provided no database files were changed. At least one copy of the control file, all data files, and at least one online redo log file in each online redo log group must exist.
Before you restart the database system, you should find out which background process was canceled and why. Check the ALERT and trace files.
If you want to perform a recovery after an instance error, shut down the instance (if it is still active) with the commandshutdown abort . Analyze and correct the problem. Restart the database with the command startup .
When you restart the database, the system monitor performs an automatic instance recovery, as described above.