On this page Carat arrow pointing down
CockroachDB v20.2 is no longer supported as of May 10, 2022. For more details, refer to the Release Support Policy.

New in v20.2: RESTORE no longer requires an Enterprise license, regardless of the options passed to it or to the backup it is restoring.

The RESTORE statement restores your cluster's schemas and data from a BACKUP stored on a services such as AWS S3, Google Cloud Storage, NFS, or HTTP storage.

Because CockroachDB is designed with high fault tolerance, restores are designed primarily for disaster recovery, i.e., restarting your cluster if it loses a majority of its nodes. Isolated issues (such as small-scale node outages) do not require any intervention.

You can restore:


RESTORE is a blocking statement. To run a restore job asynchronously, use the DETACHED option. See the options below.

Required privileges

  • Full cluster restores can only be run by members of the admin role. By default, the root user belongs to the admin role.
  • For all other restores, the user must have write access (CREATE or INSERT) on all objects affected.

Source privileges

The source file URL does not require the ADMIN role in the following scenarios:

  • S3 and GS using SPECIFIED (and not IMPLICIT) credentials. Azure is always SPECIFIED by default.
  • Userfile

The source file URL does require the ADMIN role in the following scenarios:

We recommend using cloud storage for bulk operations.


RESTORE FROM TABLE table_pattern , DATABASE database_name , FROM subdirectory IN full_backup_location partitioned_backup_location , AS OF SYSTEM TIME timestamp WITH restore_options_list OPTIONS ( restore_options_list )


Parameter Description
table_pattern The table or view you want to restore.
database_name The name of the database you want to restore (i.e., restore all tables and views in the database). You can restore an entire database only if you had backed up the entire database.
full_backup_location The URL where the full backup is stored.

For information about this URL structure, see Backup File URLs.
incremental_backup_location The URL where an incremental backup is stored.

Lists of incremental backups must be sorted from oldest to newest. The newest incremental backup's timestamp must be within the table's garbage collection period.

For information about this URL structure, see Backup File URLs.

For more information about garbage collection, see Configure Replication Zones.
AS OF SYSTEM TIME timestamp Restore data as it existed as of timestamp. You can restore point-in-time data only if you had taken full or incremental backup with revision history.
restore_options_list Control your backup's behavior with these options.


You can control RESTORE behavior using any of the following in the restore_options_list. To set multiple RESTORE options, use a comma-separated list:

into_db Database name Use to change the target database for table restores. (Does not apply to database or cluster restores.)

Example: WITH into_db = 'newdb'
skip_missing_foreign_keys N/A Use to remove the missing foreign key constraints before restoring.

Example: WITH skip_missing_foreign_keys
skip_missing_sequences N/A Use to ignore sequence dependencies (i.e., the DEFAULT expression that uses the sequence).

Example: WITH skip_missing_sequences
skip_missing_sequence_owners N/A New in v20.2: Must be used when restoring either a table that was previously a sequence owner or a sequence that was previously owned by a table.

Example: WITH skip_missing_sequence_owners
skip_missing_views N/A Use to skip restoring views that cannot be restored because their dependencies are not being restored at the same time.

Example: WITH skip_missing_views
encryption_passphrase Passphrase used to create the encrypted backup The passphrase used to decrypt the file(s) that were encrypted by the BACKUP statement.
DETACHED N/A New in v20.2: When RESTORE runs with DETACHED, the job will execute asynchronously. The job ID is returned after the restore job creation completes. Note that with DETACHED specified, further job information and the job completion status will not be returned. For more on the differences between the returned job data, see the example below. To check on the job status, use the SHOW JOBS statement.

To run a restore within a transaction, use the DETACHED option.

Backup file URLs

CockroachDB uses the URL provided to construct a secure API call to the service you specify. The URL structure depends on the type of file storage you are using. For more information, see the following:

Functional details

You can restore:

Full cluster

A full cluster restore can only be run on a target cluster with no user-created databases or tables. Restoring a full cluster includes:


When you restore a full cluster with an Enterprise license, it will restore the Enterprise license of the cluster you are restoring from. If you want to use a different license in the new cluster, make sure to update the license after the restore is complete.


The database cannot already exist in the target cluster. Restoring a database will create a new database and restore all of its tables and views. The created database will have the name of the database in the backup.

RESTORE DATABASE backup_database_name FROM 'your_backup_location';

If dropping or renaming an existing database is not an option, you can use table restore to restore all tables into a specific database by using the WITH into_db option.


You can also restore individual tables (which automatically includes their indexes) or views from a backup. This process uses the data stored in the backup to create entirely new tables or views in the target database.

By default, tables and views are restored into a target database matching the name of the database from which they were backed up. If the target database does not exist, you must create it. You can choose to change the target database with the into_db option.

The target database must not have tables or views with the same name as the tables or views you're restoring. If any of the restore target's names are being used, you can:


RESTORE only offers table-level granularity; it does not support restoring subsets of a table.

New in v20.2: When restoring an individual table that references a user-defined type (e.g., ENUM), CockroachDB will first check to see if the type already exists. The restore will attempt the following for each user-defined type within a table backup:

  • If there is not an existing type in the cluster with the same name, CockroachDB will create the user-defined type as it exists in the backup.
  • If there is an existing type in the cluster with the same name that is compatible with the type in the backup, CockroachDB will map the type in the backup to the type in the cluster.
  • If there is an existing type in the cluster with the same name but it is not compatible with the type in the backup, the restore will not succeed and you will be asked to resolve the naming conflict. You can do this by either dropping or renaming the existing user-defined type.

In general, two types are compatible if they are the same kind (e.g., an enum is only compatible with other enums). Additionally, enums are only compatible if they have the same ordered set of elements that have also been created in the same way. For example:

  • CREATE TYPE t1 AS ENUM ('yes', 'no') and CREATE TYPE t2 AS ENUM ('yes', 'no') are compatible.
  • CREATE TYPE t1 AS ENUM ('yes', 'no') and CREATE TYPE t2 AS ENUM ('no', 'yes') are not compatible.
  • CREATE TYPE t1 AS ENUM ('yes', 'no') and CREATE TYPE t2 AS ENUM ('yes'); ALTER TYPE t2 ADD VALUE ('no') are not compatible because they were not created in the same way.

Object dependencies

Dependent objects must be restored at the same time as the objects they depend on.

Object Depends On
Table with foreign key constraints The table it REFERENCES (however, this dependency can be removed during the restore).
Table with a sequence The sequence.
Views The tables used in the view's SELECT statement.
Interleaved tables The parent table in the interleaved hierarchy.

Users and privileges

To restore your users and privilege grants, you can do a cluster backup and restore the cluster to a fresh cluster with no user data.

If you are not doing a full cluster restore, the table-level privileges need to be granted to the users after the restore is complete. To do this, backup the system.users table, restore users and their passwords, and then grant the table-level privileges.

Restore types

You can either restore from a full backup or from a full backup with incremental backups, based on the backup files you include:

Restore Type Parameters
Full backup Include only the path to the full backup.
Full backup +
incremental backups
If the full backup and incremental backups were sent to the same destination, include only the path to the full backup (e.g., RESTORE FROM 'full_backup_location';).

If the incremental backups were sent to a different destination from the full backup, include the path to the full backup as the first argument and the subsequent incremental backups from oldest to newest as the following arguments (e.g., RESTORE FROM 'full_backup_location', 'incremental_location_1', 'incremental_location_2';).


The RESTORE process minimizes its impact to the cluster's performance by distributing work to all nodes. Subsets of the restored data (known as ranges) are evenly distributed among randomly selected nodes, with each range initially restored to only one node. Once the range is restored, the node begins replicating it others.


When a RESTORE fails or is canceled, partially restored data is properly cleaned up. This can have a minor, temporary impact on cluster performance.

Viewing and controlling restore jobs

After CockroachDB successfully initiates a restore, it registers the restore as a job, which you can view with SHOW JOBS.

After the restore has been initiated, you can control it with PAUSE JOB, RESUME JOB, and CANCEL JOB.


If initiated correctly, the statement returns when the restore is finished or if it encounters an error. In some cases, the restore can continue after an error has been returned (the error message will tell you that the restore has resumed in background).

Known limitations

An object's historical data is not preserved upon RESTORE. This means that if an AS OF SYSTEM TIME query is issued on a restored object, the query will fail or the response will be incorrect because there is no historical data to query.


Restore a cluster

To restore a full cluster:

> RESTORE FROM 'gs://acme-co-backup/test-cluster';

Restore a database

> RESTORE DATABASE bank FROM 'gs://acme-co-backup/database-bank-2017-03-27-weekly';

RESTORE DATABASE can only be used if the entire database was backed up.

Restore a table

To restore a single table:

> RESTORE bank.customers FROM 'gs://acme-co-backup/database-bank-2017-03-27-weekly';

To restore multiple tables:

> RESTORE bank.customers, bank.accounts FROM 'gs://acme-co-backup/database-bank-2017-03-27-weekly';

Restore from incremental backups

Restoring from incremental backups requires previous full and incremental backups. To restore from a destination containing the full backup, as well as the incremental backups (stored as subdirectories):

> RESTORE FROM 'gs://acme-co-backup/test-cluster';

To explicitly point to where your incremental backups are, provide the previous full and incremental backup locations in a comma-separated list. In this example, -weekly is the full backup and the two -nightly are incremental backups.

> RESTORE bank.customers \
FROM 'gs://acme-co-backup/database-bank-2017-03-27-weekly', 'gs://acme-co-backup/database-bank-2017-03-28-nightly', 'gs://acme-co-backup/database-bank-2017-03-29-nightly';

If you are restoring from HTTP storage, provide the previous full and incremental backup locations in a comma-separated list. You cannot use the simplified syntax.


RESTORE will re-validate indexes when incremental backups are created from an older version, but restored from a newer version.

Incremental backups created by v20.2.2 and prior v20.2.x releases or v20.1.4 and prior v20.1.x releases may include incomplete data for indexes that were in the process of being created. Therefore, when incremental backups taken by these versions are restored by v20.2.8+, any indexes created during those incremental backups will be re-validated by RESTORE.

Restore a backup asynchronously

New in v20.2: Use the DETACHED option to execute the restore job asynchronously:

'gs://acme-co-backup/test-cluster' \

The job ID is returned after the restore job creation completes:

(1 row)

Without the DETACHED option, RESTORE will block the SQL connection until the job completes. Once finished, the job status and more detailed job data is returned:

job_id             |  status   | fraction_completed | rows | index_entries | bytes
652471804772712449 | succeeded |                  1 |   50 |             0 |  4911
(1 row)

Other restore usages

Restore tables into a different database

By default, tables and views are restored to the database they originally belonged to. However, using the into_db option, you can control the target database.

> RESTORE bank.customers \
FROM 'gs://acme-co-backup/database-bank-2017-03-27-weekly' \
WITH into_db = 'newdb';

Remove the foreign key before restore

By default, tables with Foreign Key constraints must be restored at the same time as the tables they reference. However, using the skip_missing_foreign_keys option you can remove the Foreign Key constraint from the table and then restore it.

> RESTORE bank.accounts \
FROM 'gs://acme-co-backup/database-bank-2017-03-27-weekly' \
WITH skip_missing_foreign_keys;

Restoring users from system.users backup

The system.users table stores your cluster's usernames and their hashed passwords. To restore them, you must restore the system.users table into a new database because you cannot drop the existing system.users table.

After it's restored into a new database, you can write the restored users table data to the cluster's existing system.users table.

> RESTORE system.users \
FROM 'azure://acme-co-backup/table-users-2017-03-27-full?AZURE_ACCOUNT_KEY=hash&AZURE_ACCOUNT_NAME=acme-co' \
WITH into_db = 'newdb';
> INSERT INTO system.users SELECT * FROM newdb.users;
> DROP TABLE newdb.users;

See also

Yes No
On this page

Yes No