Recovery Concepts and Strategies

The JADE database backup and recovery mechanisms provide an important safeguard for protecting critical data stored in a JADE database. Restoring and recovering a database from a backup enables the complete restoration of data over a wide range of potential system problems.

It is important to distinguish between restoring a database and recovering a database. Restoring a database means bringing back the files that comprise the database from a backup or archive repository. The database at this point is not in an operational state.

Recovering a database means bringing the restored database to the point of being fully operational, which may involve restoring journal files and performing a roll-forward recovery, for example.

As backing up and restoring a large database is a lengthy process, and backups of very large databases become impractical to perform on a regular basis for systems with high availability requirements, the Synchronized Database Service (SDS) provides an alternative recovery strategy. For details, see "Implementing a Database Recovery Strategy", in Chapter 1 of the JADE Synchronized Database Service (SDS) Administration Guide.

When a database is restored, a restoreinfo file is created in the database directory. The restoreinfo file contains similar information to the backupinfo file and it includes a record for each file contained in the backup. When files or partitions are excluded from a database restore operation, they are marked as offline in the restoreinfo file. When the database is first opened following a restore, the restoreinfo file in the database directory is processed. Any files or partitions marked as offline that are not resident are marked as offline in their respective control files, and all control file latest backup timestamps, and latest full backup timestamps, are updated to reflect the backup that was restored.

During database restart recovery:

• When a recovery redo or undo operation encounters an error such as an index checksum error, the file is marked as having a broken index.

• Recovery continues to redo and undo object updates to files with broken indexes but stops applying most index updates (with the exception of index block creates).

• At the end of recovery, an automatic reindex operation is performed on all files with broken indexes.

See also "Reindexing Database Files", later in this chapter.