Backup Validation

On this page Carat arrow pointing down

CockroachDB provides backup validation tools to check that backups you have in storage are restorable. Although a successful restore completely validates a backup, the validation tools offer a faster alternative and return an error message if a backup is not valid.

You can validate a backup of a cluster, database, or table backup with one of the following SHOW BACKUP or RESTORE statements. The three options result in increasing levels of backup validation:

  1. SHOW BACKUP ... WITH check_files: Check that all files belonging to a backup are in the expected location in storage. Refer to Validate backup files for an example.
  2. RESTORE ... WITH schema_only: Restore the schema from the backup to verify that it is valid without restoring any rows. Refer to Validate a backup is restorable for an example.
  3. RESTORE ... WITH schema_only, verify_backup_table_data: Run a schema_only restore and have the restore read all rows from external storage, verify checksums, and discard the rows before writing them to disk. To use verify_backup_table_data, you must include schema_only in the statement. Refer to Validate backup table data is restorable for an example.

The options that give the most validation coverage will increase the runtime of the check. That is, verify_backup_table_data will take a longer time to validate a backup compared to check_files or schema_only alone. Despite that, each of these validation options provide a quicker way to validate a backup over running a "regular" restore.


Cockroach Labs recommends implementing the following validation plan to test your backups:

  1. Very frequent schema_only restores: Checks your credentials, schema issues, and version compatability.
  2. Frequent verify_backup_table_data restores: Tests that all data files are present and uncorrupted.
  3. Somewhat frequent "full" restores: Validates a backup completely.

A schema_only restore provides high confidence in the recoverability of your backup and a verify_backup_table_data restore provides a slightly higher confidence as it checks for corruptions. Only a complete RESTORE can provide full confidence in validating a backup. The actual frequency you should run backup validation tests depends on the criticality of your data, the rate of change in your database, and your organization's disaster recovery and business continuity requirements.

Supported products

The feature described on this page is available in CockroachDB Dedicated, CockroachDB Serverless, and CockroachDB Self-Hosted clusters when you are running customer-owned backups. For a full list of features, see Backup and restore product support.

Validate backup files

Using SHOW BACKUP with the check_files option, you can check that all SST and metadata files that belong to a backup are present in the storage location.

  1. Take a backup that we'll use for each of the examples on this page:

    BACKUP DATABASE movr INTO "s3://bucket?AWS_ACCESS_KEY_ID={Access Key ID}&AWS_SECRET_ACCESS_KEY={Secret Access Key}" AS OF SYSTEM TIME "-10s";
  2. To find a specific backup to validate in the storage location, show the stored backups in the storage location:

    SHOW BACKUPS IN "s3://bucket?AWS_ACCESS_KEY_ID={Access Key ID}&AWS_SECRET_ACCESS_KEY={Secret Access Key}";
    (2 rows)
  3. Use SHOW BACKUP ... check_files with a backup for validation:

    SHOW BACKUP "2022/09/19-134123.64" IN "s3://bucket?AWS_ACCESS_KEY_ID={Access Key ID}&AWS_SECRET_ACCESS_KEY={Secret Access Key}" WITH check_files;

    This will return the following output after validating that the backup files are correct and present:

      database_name | parent_schema_name |        object_name         | object_type | backup_type | start_time |          end_time          | size_bytes | rows  | is_full_cluster | file_bytes
      NULL          | NULL               | movr                       | database    | full        | NULL       | 2022-09-19 13:41:23.645189 |       NULL |  NULL |        f        |       NULL
      movr          | NULL               | public                     | schema      | full        | NULL       | 2022-09-19 13:41:23.645189 |       NULL |  NULL |        f        |       NULL
      movr          | public             | users                      | table       | full        | NULL       | 2022-09-19 13:41:23.645189 |      31155 |   340 |        f        |      16598
      movr          | public             | vehicles                   | table       | full        | NULL       | 2022-09-19 13:41:23.645189 |      22282 |   113 |        f        |      12459
      movr          | public             | rides                      | table       | full        | NULL       | 2022-09-19 13:41:23.645189 |     261950 |   902 |        f        |     135831
      movr          | public             | vehicle_location_histories | table       | full        | NULL       | 2022-09-19 13:41:23.645189 |     742557 | 10850 |        f        |     318583
      movr          | public             | promo_codes                | table       | full        | NULL       | 2022-09-19 13:41:23.645189 |     228320 |  1034 |        f        |     118376
      movr          | public             | user_promo_codes           | table       | full        | NULL       | 2022-09-19 13:41:23.645189 |       9320 |   111 |        f        |       4832

    The output will return file_bytes along with the columns you receive from SHOW BACKUP without check_files. The file_bytes column indicates the estimated bytes in external storage for a particular table object. For more detail on the output columns, see the SHOW BACKUP Response table.

  4. If SHOW BACKUP ... check_files cannot read from a file, it will return an error message similar to the following:

    ERROR: The following files are missing from the backup:

    SHOW BACKUP ... check_files will return up to ten file paths for incorrect or missing files.

Validate a backup is restorable

To validate that a backup is restorable, you can run RESTORE with the schema_only option, which will complete a restore without restoring any rows. This process is significantly faster than running a regular restore for the purposes of validation.

A schema_only restore produces close to complete validation coverage on backups. However, this restore type does not read or write from any of the SST files, which store the backed-up rows. You can use SHOW BACKUP ... WITH check_files in addition to a schema_only restore to check that these SST files are present for a restore operation. Or, you can use schema_only in combination with verify_backup_table_data. Refer to Validate backup table data is restorable.

Run RESTORE with the schema_only option, specifying either LATEST or the specific backup you would like to restore:

RESTORE DATABASE movr FROM "2022/09/19-134123.64" IN "s3://bucket?AWS_ACCESS_KEY_ID={Access Key ID}&AWS_SECRET_ACCESS_KEY={Secret Access Key}" WITH schema_only;
        job_id       |  status   | fraction_completed | rows | index_entries | bytes
  797982663104856065 | succeeded |                  1 |    0 |             0 |     0
(1 row)

You can also use the new_db_name option to restore a database to a different name. For example, new_db_name = test_movr.

Verify that the table schemas have been restored, but that the tables contain no rows:

  schema_name |         table_name         | type  | owner | estimated_row_count | locality
  public      | promo_codes                | table | root  |                   0 | NULL
  public      | rides                      | table | root  |                   0 | NULL
  public      | user_promo_codes           | table | root  |                   0 | NULL
  public      | users                      | table | root  |                   0 | NULL
  public      | vehicle_location_histories | table | root  |                   0 | NULL
  public      | vehicles                   | table | root  |                   0 | NULL
(6 rows)

You'll find the tables in place with no rows.

SELECT * FROM rides;
  id | city | vehicle_city | rider_id | vehicle_id | start_address | end_address | start_time | end_time | revenue
(0 rows)

Once you have successfully validated the restore, you can revert the cluster to its pre-restore state by dropping the table or database:


Cluster-level backup validation

It is important to note that full cluster restores with schema_only will write the system tables to disk. This provides important coverage for validation at the cluster level. Writing the system tables should not have a notable impact on the runtime of this process.

Once you have successfully validated a cluster-level restore, the restored system tables cannot be reverted. However, you can drop the databases and tables, as shown in the previous command.

Validate backup table data is restorable

A restore with the verify_backup_table_data option will perform a schema_only restore, and:

  1. Read all rows from the storage location.
  2. Checksum the rows to ensure they are not corrupt.
  3. Discard the rows before they are written to disk.

Similarly, to just schema_only restores, you'll find the table schemas restored. If a file is not present or unreadable in the backup, you'll receive an error.

Unlike a schema_only restore, a verify_backup_table_data restore also reads and checksums the rows to validate the backup.

It is necessary to include schema_only when you run a restore with verify_backup_table_data:

RESTORE DATABASE movr FROM LATEST IN "s3://bucket?AWS_ACCESS_KEY_ID={Access Key ID}&AWS_SECRET_ACCESS_KEY={Secret Access Key}" WITH schema_only, verify_backup_table_data;

See also

Yes No
On this page

Yes No