When Do You Need It?
Complete recovery brings back everything, including a bad change. Point-in-time recovery (PITR) stops before the mistake. Use it when someone drops a schema, truncates a table or deletes the wrong data and the change has already been committed.
Ways to Choose the Stopping Point
| Clause | Example |
|---|---|
UNTIL TIME | SET UNTIL TIME "TO_DATE('2026-11-10 14:29:00','YYYY-MM-DD HH24:MI:SS')" |
UNTIL SCN | SET UNTIL SCN 1234567 |
UNTIL SEQUENCE | SET UNTIL SEQUENCE 215 THREAD 1 |
| Restore point | SET UNTIL RESTORE POINT before_upgrade |
Whole Database PITR
Prefer a Smaller Tool
| Problem | Better tool |
|---|---|
| One table was changed | Flashback Table, or RECOVER TABLE |
| One application (PDB) was damaged | PDB point-in-time recovery |
| Wrong rows deleted recently | Flashback Query to read the old rows and re-insert |
PDB Point-in-Time Recovery
Other PDBs keep running and are not affected. The auxiliary destination is a temporary work area (not always needed when local undo is used).
Recover One Table
RMAN builds a temporary database, extracts the table as it was and loads it under the new name orders_before. You then compare and copy the rows you need.
Find the Right Moment
After the Recovery
- Check the data carefully before telling users.
- Take a new full backup immediately, because OPEN RESETLOGS started a new incarnation.
- Write down what happened and which transactions were lost.
Conclusion
Point-in-time recovery rewinds to just before a mistake. Always compare it with smaller tools like flashback first, because PITR costs you all later changes.

