Repairing a corrupted database

If a database is corrupted and cannot start, this is indicated in the Databases widget.

In the Database Troubleshooting dashboard, you can attempt to repair a corrupted database.

Typically, corruption is an inconsistency between the description of the database, stored in metadata.json, and the files currently present in the database.

Select Dashboards > OLAP > Database Troubleshooting.
In the Databases widget, click the Repair icon of the database to repair.

Select a repair option. The options depend on the nature of the corruption, but these options are generally available:

Option	Description
Check database integrity again	Rechecks the consistency of the current database.
Go back to this valid save version	Reverts to a version of the database that was valid when it was saved. Typically, only one saved version is available.
Go back to this valid backup version	Reverts to a backup of the database that was valid when it was backed up.
Adjust the database description file to the saved files	Changes the description in the metadata to match the files currently present. This option makes the metadata and database files consistent and can enable the database to start. But that does not guarantee that the database performs as required.
Re-run database update	Re-runs the last database update.
Try to recover local database files	Re-runs database crash recovery. Conditions for success are that it is known which cluster node database was running, and that node is still part of the farm.

None of the repair options can be undone but, after a failed repair attempt, the other options remain available. The available options can depend on the outcome of a repair attempt.

By default, a repaired database starts automatically after the repair. Optionally, clear the option to start the database.
Click Repair, and then click OK to confirm.