Take and Restore Locality-aware Backups

On this page Carat arrow pointing down

This page provides information about how to take and restore locality-aware backups.

Note:

Locality-aware BACKUP is an Enterprise-only feature. However, you can take full backups without an Enterprise license.

You can create locality-aware backups such that each node writes files only to the backup destination that matches the node locality configured at node startup.

This is useful for:

  • Reducing cloud storage data transfer costs by keeping data within cloud regions.
  • Helping you comply with data domiciling requirements.

A locality-aware backup is specified by a list of URIs, each of which has a COCKROACH_LOCALITY URL parameter whose single value is either default or a single locality key-value pair such as region=us-east. At least one COCKROACH_LOCALITY must be the default. Given a list of URIs that together contain the locations of all of the files for a single locality-aware backup, RESTORE can read in that backup.

Note:

The locality query string parameters must be URL-encoded.

Every node involved in the backup is responsible for backing up the ranges for which it was the leaseholder at the time the distributed backup flow was planned. The locality of the node running the distributed backup flow determines where the backup files will be placed in a locality-aware backup. The node running the flow, and the leaseholder node of the range being backed up are usually the same, but can differ when lease transfers have occurred during the execution of the backup. The leaseholder node returns the files to the node running the backup flow (usually a local transfer), which then writes the file to the external storage location with a locality that matches its own localities (with an overall preference for more specific values in the locality hierarchy). If there is no match, the default locality is used.

Create a locality-aware backup

To create a locality-aware backup where nodes with the locality region=us-west write backup files to s3://us-west-bucket, and all other nodes write to s3://us-east-bucket by default, run:

icon/buttons/copy
> BACKUP INTO
      ('s3://us-east-bucket?COCKROACH_LOCALITY=default', 's3://us-west-bucket?COCKROACH_LOCALITY=region%3Dus-west');

You can restore this backup by running:

icon/buttons/copy
> RESTORE FROM LATEST IN ('s3://us-east-bucket', 's3://us-west-bucket');

Note that the first URI in the list has to be the URI specified as the default URI when the backup was created. If you have moved your backups to a different location since the backup was originally taken, the first URI must be the new location of the files originally written to the default location.

To restore from a specific backup, use RESTORE FROM {subdirectory} IN ....

For guidance on how to identify the locality of a node to pass in a backup query, see Show a node's locality.

Note:

For guidance on connecting to other storage options or using other authentication parameters, read Use Cloud Storage for Bulk Operations.

Show a node's locality

To determine the locality that a node was started with, run SHOW LOCALITY:

icon/buttons/copy
SHOW LOCALITY;
    locality
+---------------------+
region=us-east,az=az1
(1 row)

The output shows the locality to which the node will write backup data. One of the single locality key-value pairs can be passed to BACKUP with the COCKROACH_LOCALITY parameter (e.g., 's3://us-east-bucket?COCKROACH_LOCALITY=region%3Dus-east').

Note:

Specifying both locality tier pairs (e.g., region=us-east,az=az1) from the output will cause the backup query to fail with: tier must be in the form "key=value".

Restore from a locality-aware backup

Given a list of URIs that together contain the locations of all of the files for a single locality-aware backup, RESTORE can read in that backup. Note that the list of URIs passed to RESTORE may be different from the URIs originally passed to BACKUP. This is because it's possible to move the contents of one of the parts of a locality-aware backup (i.e., the files written to that destination) to a different location, or even to consolidate all the files for a locality-aware backup into a single location.

When restoring a full backup, the cluster data is restored first, then the system table data "as is." This means that the restored zone configurations can point to regions that do not have active nodes in the new cluster. For example, if your full backup has the following zone configurations:

> ALTER PARTITION europe_west OF INDEX movr.public.rides@primary \
        CONFIGURE ZONE USING constraints = '[+region=europe-west1]';

> ALTER PARTITION us_east OF INDEX movr.public.rides@primary \
        CONFIGURE ZONE USING constraints = '[+region=us-east1]';

> ALTER PARTITION us_west OF INDEX movr.public.rides@primary \
        CONFIGURE ZONE USING constraints = '[+region=us-west1]';

And the restored cluster does not have nodes with the locality region=us-west1, the restored cluster will still have a zone configuration for us-west1. This means that the cluster's data will not be reshuffled to us-west1 because the region does not exist. The data will be distributed as if the zone configuration does not exist. For the data to be distributed correctly, you can add node(s) with the missing region or remove the zone configuration.

For example, use the following to create a locality-aware backup:

icon/buttons/copy
> BACKUP INTO
      ('s3://us-east-bucket?COCKROACH_LOCALITY=default', 's3://us-west-bucket?COCKROACH_LOCALITY=region%3Dus-west')

Restore a locality-aware backup with:

icon/buttons/copy
> RESTORE FROM LATEST IN ('s3://us-east-bucket/', 's3://us-west-bucket/');

To restore from a specific backup, use RESTORE FROM {subdirectory} IN ....

Note:

RESTORE is not truly locality-aware; while restoring from backups, a node may read from a store that does not match its locality. This can happen in the cases that either the BACKUP or RESTORE was not of a full cluster. Note that during a locality-aware restore, some data may be temporarily located on another node before it is eventually relocated to the appropriate node. To avoid this, you can manually restore zone configurations from a locality-aware backup.

Create an incremental locality-aware backup

If you backup to a destination already containing a full backup, an incremental backup will be appended to the full backup in a subdirectory. For example:

icon/buttons/copy
> BACKUP INTO LATEST IN
      ('s3://us-east-bucket?COCKROACH_LOCALITY=default', 's3://us-west-bucket?COCKROACH_LOCALITY=region%3Dus-west');
Note:

When restoring from an incremental locality-aware backup, you need to include every locality ever used, even if it was only used once.

It is recommended that the same localities be included for every incremental backup in the series of backups; however, only the default locality is required.

And if you want to explicitly control where your incremental backups go, specify the subdirectory in the BACKUP statement:

icon/buttons/copy
> BACKUP INTO {subdirectory} IN
        ('s3://us-east-bucket?COCKROACH_LOCALITY=default', 's3://us-west-bucket?COCKROACH_LOCALITY=region%3Dus-west');

To view the available subdirectories, use SHOW BACKUPS.

Restore from an incremental locality-aware backup

A locality-aware backup URI can also be used in place of any incremental backup URI in RESTORE.

For example, an incremental locality-aware backup created with:

icon/buttons/copy
> BACKUP INTO LATEST IN
      ('s3://us-east-bucket?COCKROACH_LOCALITY=default', 's3://us-west-bucket?COCKROACH_LOCALITY=region%3Dus-west')

can be restored by running:

icon/buttons/copy
> RESTORE FROM LATEST IN ('s3://us-east-bucket/', 's3://us-west-bucket/');

To restore from a specific backup, use RESTORE FROM {subdirectory} IN ....

Note:

When restoring from an incremental locality-aware backup, you need to include every locality ever used, even if it was only used once.

Manually restore zone configurations from a locality-aware backup

During a locality-aware restore, some data may be temporarily located on another node before it is eventually relocated to the appropriate node. To avoid this, you need to manually restore zone configurations first:

Once the locality-aware restore has started, pause the restore:

icon/buttons/copy
> PAUSE JOB 27536791415282;

The system.zones table stores your cluster's zone configurations, which will prevent the data from rebalancing. To restore them, you must restore the system.zones table into a new database because you cannot drop the existing system.zones table:

icon/buttons/copy
> RESTORE TABLE system.zones FROM '2021/03/23-213101.37' IN
    'azure://acme-co-backup?AZURE_ACCOUNT_KEY=hash&AZURE_ACCOUNT_NAME=acme-co'
    WITH into_db = 'newdb';

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

icon/buttons/copy
> INSERT INTO system.zones SELECT * FROM newdb.zones;

Then drop the temporary table you created:

icon/buttons/copy
> DROP TABLE newdb.zones;

Then, resume the restore:

icon/buttons/copy
> RESUME JOB 27536791415282;

See also


Yes No
On this page

Yes No