Teradata ARC provides functions to roll back or roll forward databases or tables from a journal table. To identify the databases and/or data tables to recover, use a Teradata ARC statement. Multiple databases or data tables must all use the same journal table because a single recovery operation only uses change images from one journal.
Note: When a PPI table or a Column Partition (CP) table is restored, Teradata ARC automatically executes the Revalidate step during the BUILD phase. The Revalidate process increases the Util version of the table. This is considered to be a structural change to the table. Because of this change, a PPI table or a CP table cannot be recovered using either the RollBack or RollForward operation after it is restored.
A Teradata ARC statement also identifies whether the restored or current journal subtable is to be used in the recovery activity.
To recover data tables from more than one archived journal, perform the recovery operation in three steps:
1 Restore an archived journal.
2 Recover the data tables using the restored journal.
3 Repeat steps 1 and 2 until all the tables are recovered.
Teradata ARC completes the requested recovery action as long as the table structure is the same as the structure of the change images. If the structures are not the same, Teradata ARC stops the requested recovery for the data table and returns an error message. When recovering multiple tables, the stopped recovery of one table does not affect recovery of the other tables.
Teradata Database does not generate transient journal images during a rollback or rollforward operation. If the rollback or rollforward operation is not completed, the data tables being recovered are left in an unknown state.
If the rollback or rollforward is not completed because of a:
If the HUT locks on the data tables are released prior to restarting and completing the Teradata ARC rollback or rollforward operation, the state of the data is unpredictable.
Teradata Database does not generate permanent journal images during a rollback or rollforward operation. Consequently, a rollback operation might not be completed properly.
For example, assume:
1 A batch job updates a data table and creates both before- and after-images.
2 The batch job finishes.
3 Later (because of an error in the batch update) the table is rolled back using the before-images.
4 A disk is replaced and the most recent archive is restored. A rollforward is performed to incorporate any changes that have occurred since the archive was created.
If the same journal that was used in the rollback described above is used in the rollforward, it might be expected that the updates created by the batch job would not be reapplied because the batch job had been rolled back. Instead, the Teradata Database does not modify a journal table as the result of a recovery operation and the updates executed by the batch program are reapplied to the data tables during the rollforward.