Take Full and Incremental Backups

Because CockroachDB is designed with high fault tolerance, backups are primarily needed for (i.e., if your cluster loses a majority of its nodes). Isolated issues (such as small-scale node outages) do not require any intervention. However, as an operational best practice, we recommend taking regular backups of your data. There are two main types of backups:

Full backups
Incremental backups

You can use the statement to efficiently back up your cluster’s schemas and data to popular cloud services such as AWS S3, Google Cloud Storage, or NFS, and the statement to efficiently restore schema and data as necessary. For more information, see .

The BACKUP ... TO and RESTORE ... FROM {storage_uri} syntax has been removed from CockroachDB v24.3 and later.For details on the syntax to run BACKUP and RESTORE, refer to the backup and restore examples.

We recommend using scheduled backups to automate daily backups of your cluster.

Supported products

The feature described on this page is available in CockroachDB Basic, CockroachDB Standard, CockroachDB Advanced, and CockroachDB self-hosted clusters when you are running . For a full list of features, refer to .

Backup collections

A backup collection defines a set of backups and their metadata. The collection can contain multiple full backups and their subsequent incremental backups. The path to a backup is created using a date-based naming scheme and stored at the passed with the BACKUP statement. For a . The backup collection will be stored according to the URIs passed with the BACKUP statement: BACKUP INTO LATEST IN {collectionURI}, {localityURI}, {localityURI}. Here, the collectionURI represents the default locality. By default, full backups are stored at the root of the collection’s URI in a date-based path, and incremental backups are stored in the /incrementals directory. The following example shows a backup collection created using these default values, where all backups reside in one storage bucket:

Collection URI:
|—— 2022
  |—— 02
    |—— 09-155340.13/
      |—— Full backup files
[...]
|—— incrementals
  |—— 2022
  |—— 02
    |—— 25-172907.21/
      |—— 20220325
        |—— 17921.23
          |—— incremental backup files

will display a list of the full backup subdirectories at the collection’s URI. In the examples on this page, is a placeholder for the storage location that will contain the example backup. You will encounter an error if you run multiple to the same storage URI. Backup collections can contain multiple full and incremental backups, but each collection’s URI must be unique. If you are using backup schedules, each schedule must have a unique URI. For example, if you have a backup schedule running backups for the database users the full backup and incremental backup should have the same storage URI for the full and incremental schedule. (CREATE SCHEDULE FOR BACKUP will automatically create two schedules for the full and incremental backup to the same storage URI.) If there is another backup schedule, for the database accounts, the full and incremental backups for accounts should have the same storage URI. However, the storage URI for the accounts backup collection should be different to the storage URI for the users backup collection.

Full backups

Full backups contain an un-replicated copy of your data and can always be used to restore your cluster. These files are roughly the size of your data and require greater resources to produce than incremental backups. You can take full backups as of a given timestamp. Optionally, you can include the available in the backup. In most cases, it’s recommended to take nightly full backups of your cluster. A cluster backup allows you to do the following:

Restore table(s) from the cluster
Restore database(s) from the cluster
Restore a full cluster

include . When you a full cluster backup that includes Enterprise license, the Enterprise license is also restored.

To set a target for the amount of backup data written to each backup file, use the bulkio.backup.file_size .See the page for more details on using cluster settings.

Take a full backup

To perform a full cluster backup, use the statement:

BACKUP INTO '{collectionURI}';

To restore a backup, use the statement, specifying what you want to restore as well as the collection’s URI:

To restore the latest backup of a table:

RESTORE TABLE bank.customers FROM LATEST IN '{collectionURI}';

To restore the latest backup of a database:

RESTORE DATABASE bank FROM LATEST IN '{collectionURI}';

To restore the latest backup of your full cluster:
RESTORE FROM LATEST IN '{collectionURI}';

A full cluster restore can only be run on a destination cluster that has never had user-created databases or tables.

To restore a backup from a specific subdirectory:

RESTORE DATABASE bank FROM {subdirectory} IN '{collectionURI}';

To view the available backup subdirectories, use .

Incremental backups

If your cluster grows too large for daily full backups, you can take less frequent full backups (e.g., weekly) with daily incremental backups. Incremental backups are storage efficient and faster than full backups for larger clusters. If you are taking backups on a regular cadence, we recommend for your backups.

Recommendations for incremental backup frequency

Incremental backups form chains between full backups. Each incremental backup contains only the data that has changed since a base set of backups. This base set of backups must include one full backup and can include multiple incremental backups, which are typically smaller and faster to run than full backups. You can take incremental backups either as of a given timestamp or with full . Cockroach Labs supports up to 48 incremental backups between full backups. Taking more than 48 incremental backups between full backups increases the risk of performance issues due to the size of the backup chain. The maximum supported frequency for incremental backups is every five minutes.

Garbage collection and backups

Incremental backups with are created by finding what data has been created, deleted, or modified since the timestamp of the last backup in the chain of backups. For the first incremental backup in a chain, this timestamp corresponds to the timestamp of the base (full) backup. For subsequent incremental backups, this timestamp is the timestamp of the previous incremental backup in the chain. determines the period for which CockroachDB retains revisions of a key. If the GC TTL of the is shorter than the frequency at which you take incremental backups with revision history, then the revisions become susceptible to garbage collection before you have backed them up. This will cause the incremental backup with revision history to fail. We recommend configuring the garbage collection period to be at least the frequency of incremental backups and ideally with a buffer to account for slowdowns. You can configure garbage collection periods using the ttlseconds . If an incremental backup is created outside of the garbage collection period, you will receive a protected ts verification error…. To resolve this issue, see the page.

If you are creating incremental backups as part of a , will ensure the backup revision data is not garbage collected, which allows you to lower the GC TTL. See for more detail.

Take an incremental backup

Periodically run the BACKUP command to take a full backup of your cluster:

> BACKUP INTO '{collectionURI}';

Then, create nightly incremental backups based off of the full backups you’ve already created. To append an incremental backup to the most recent full backup created at the collection’s URI, use the LATEST syntax:

> BACKUP INTO LATEST IN '{collectionURI}';

This will add the incremental backup to the default /incrementals directory at the root of the backup collection’s directory. With incremental backups in the /incrementals directory, you can apply different lifecycle/retention policies from cloud storage providers to the /incrementals directory as needed. If it’s ever necessary, you can then use the RESTORE statement to restore your cluster, database(s), and/or table(s). Restoring from incremental backups requires previous full and incremental backups. To restore from the latest backup in the collection, stored in the default /incrementals collection subdirectory, run:

RESTORE FROM LATEST IN '{collectionURI}';

To restore from a specific backup in the collection:

RESTORE FROM '{subdirectory}' IN '{collectionURI}';

For example:

RESTORE FROM '2023/03/23-213101.37' IN 's3://bucket/path?AUTH=implicit';

When you restore from an incremental backup, you’re restoring the entire table, database, or cluster. CockroachDB uses both the latest (or a ) incremental backup and the full backup during this process. You cannot restore an incremental backup without a full backup. Furthermore, it is not possible to restore over a , , or with existing data. Refer to for detail on the types of backups you can restore.

RESTORE will re-validate when are created from an older version (v20.2.2 and earlier or v20.1.4 and earlier), but restored by a newer version (v21.1.0+). These earlier releases may have included incomplete data for indexes that were in the process of being created.

Examples

The examples in this section use one of the following storage URIs:

External connections, which allow you to represent an external storage or sink URI. You can then specify the external connection’s name in statements rather than the provider-specific URI. For detail on using external connections, see the page.
Amazon S3 connection strings with the default AUTH=specified parameter. For guidance on using AUTH=implicit authentication with Amazon S3 buckets instead, read .

For guidance on connecting to other storage options or using other authentication parameters instead, read .

Scheduled backups

You can use to set a recurring schedule for full and incremental backups. Include the FULL BACKUP ALWAYS clause for a schedule to take only full backups. For example, to create a schedule for taking full cluster backups:

CREATE SCHEDULE core_schedule_label
    FOR BACKUP INTO 's3://{BUCKET NAME}/{PATH}?AWS_ACCESS_KEY_ID={KEY ID}&AWS_SECRET_ACCESS_KEY={SECRET ACCESS KEY}'
    RECURRING '@daily'
    FULL BACKUP ALWAYS
    WITH SCHEDULE OPTIONS first_run = 'now';

     schedule_id     |        name         | status |         first_run         | schedule |                                                                                       backup_stmt
---------------------+---------------------+--------+---------------------------+----------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  588799238330220545 | core_schedule_label | ACTIVE | 2020-09-11 00:00:00+00:00 | @daily   | BACKUP INTO 's3://{BUCKET NAME}/{PATH}?AWS_ACCESS_KEY_ID={KEY ID}&AWS_SECRET_ACCESS_KEY={SECRET ACCESS KEY}' WITH detached
(1 row)

For more examples on how to schedule backups that take full and incremental backups, refer to .

Exclude a table’s data from backups

In some situations, you may want to exclude a table’s row data from a . For example, you have a table that contains high-churn data that you would like to more quickly than the incremental backup schedule for the database or cluster holding the table. You can use the exclude_data_from_backup = true parameter with a or statement to mark a table’s row data for exclusion from a backup. It is important to note that the backup produced contains an empty table with the data excluded. The backup process does not attempt to contact the ranges to retrieve data when exclude_data_from_backup = true. Thus, backups will continue to succeed even when some ranges or regions are unavailable, if and only if, the unavailable ranges are in tables excluded from the backups. Setting this parameter prevents the cluster or database backup from delaying on the key span for this table, and it also respects the configured GC TTL. This is useful when you want to set a shorter garbage collection window for tables containing high-churn data to avoid an accumulation of unnecessary data and improves the reliability in partially unavailable clusters. Using the movr database as an example:

SHOW TABLES;

schema_name |         table_name         | type  | owner | estimated_row_count | locality
--------------+----------------------------+-------+-------+---------------------+-----------
public      | promo_codes                | table | root  |                1021 | NULL
public      | rides                      | table | root  |                 730 | NULL
public      | user_promo_codes           | table | root  |                  58 | NULL
public      | users                      | table | root  |                 211 | NULL
public      | vehicle_location_histories | table | root  |               10722 | NULL
public      | vehicles                   | table | root  |                  69 | NULL

If the user_promo_codes table’s data does not need to be included in future backups, you can run the following to exclude the table’s row data:

ALTER TABLE movr.user_promo_codes SET (exclude_data_from_backup = true);

Then back up the movr database:

BACKUP DATABASE movr INTO 's3://{BUCKET NAME}/{PATH}?AWS_ACCESS_KEY_ID={KEY ID}&AWS_SECRET_ACCESS_KEY={SECRET ACCESS KEY}' AS OF SYSTEM TIME '-10s';

Restore the database with a :

RESTORE DATABASE movr FROM LATEST IN 's3://{BUCKET NAME}/{PATH}?AWS_ACCESS_KEY_ID={KEY ID}&AWS_SECRET_ACCESS_KEY={SECRET ACCESS KEY}' WITH new_db_name = new_movr;

Move to the new database:

USE new_movr;

You’ll find that the table schema is restored:

SHOW CREATE user_promo_codes;

table_name    |                                                create_statement
-------------------+------------------------------------------------------------------------------------------------------------------
user_promo_codes | CREATE TABLE public.user_promo_codes (
              |     city VARCHAR NOT NULL,
              |     user_id UUID NOT NULL,
              |     code VARCHAR NOT NULL,
              |     "timestamp" TIMESTAMP NULL,
              |     usage_count INT8 NULL,
              |     CONSTRAINT user_promo_codes_pkey PRIMARY KEY (city ASC, user_id ASC, code ASC),
              |     CONSTRAINT user_promo_codes_city_user_id_fkey FOREIGN KEY (city, user_id) REFERENCES public.users(city, id)
              | ) WITH (exclude_data_from_backup = true)

However, the user_promo_codes table has no row data:

SELECT * FROM user_promo_codes;

city | user_id | code | timestamp | usage_count
-----+---------+------+-----------+--------------
(0 rows)

To create a table with exclude_data_from_backup, see .

Advanced examples

For examples of advanced BACKUP and RESTORE use cases, see:

​Supported products

​Backup collections

​Full backups

​Take a full backup

​Incremental backups

​Recommendations for incremental backup frequency

​Garbage collection and backups

​Take an incremental backup

​Examples

​Scheduled backups

​Exclude a table’s data from backups

​Advanced examples

​See also

Supported products

Backup collections

Full backups

Take a full backup

Incremental backups

Recommendations for incremental backup frequency

Garbage collection and backups

Take an incremental backup

Examples

Scheduled backups

Exclude a table’s data from backups

Advanced examples

See also