Restore a PostgreSQL backup

Reviewed on October 1, 2026

Restoration lets you recreate a PostgreSQL cluster from an existing backup or from the WAL (Write-Ahead Logging) archive of a source cluster.

Restoration never happens in place: it always creates a new cluster. The source cluster is not modified. If data is corrupted, restoring a recent backup is therefore the recommended recovery method.

To identify the backup to restore, see List the backups of a PostgreSQL cluster ↗.

Permissions

info

This action requires the following IAM (Identity and Access Management) permissions:

• postgresql.cluster.create

API

API

The POST /postgresql/spaces/{spaceId}/clusters request restores a PostgreSQL cluster by creating a new cluster from a backup or from the WAL archive of a source cluster.

Parameters

Parameter	Type	Required	Description
spaceId	UUID (Universally Unique Identifier)	Yes	Identifier of the space

Request body

{
  "name": "my-restored-cluster",
  "nodeConfiguration": {
    "vcpuCount": 2,
    "memorySizeGiB": 4
  },
  "volume": {
    "type": "PERFORMANCE",
    "sizeGiB": 20
  },
  "visibility": "INTERNAL",
  "restoreConfiguration": {
    "sourceType": "BACKUP",
    "clusterId": "e592a4f6-2e49-46e7-b6d0-248a7711197a",
    "backupId": "a1b2c3d4-e5f6-7890-abcd-ef1234567890"
  }
}

The name, nodeConfiguration, volume and visibility fields remain required and apply to the new cluster. They can differ from those of the source cluster.

Restore configuration

The restore configuration is set through the restoreConfiguration field of the request body, which takes one of the two following modes.

Restore from a backup

Field	Type	Required	Description
sourceType	string	Yes	Restore source type. Value: BACKUP
clusterId	UUID	Yes	Identifier of the source cluster
backupId	UUID	Yes	Identifier of the backup to restore

Restore from a cluster

Field	Type	Required	Description
sourceType	string	Yes	Restore source type. Value: CLUSTER
clusterId	UUID	Yes	Identifier of the source cluster
pointInTime	date-time	No	Timestamp of the point-in-time recovery (PITR, Point-In-Time Recovery). If omitted, the restoration is performed to the latest available state

Example of a restore with point-in-time recovery:

{
  "restoreConfiguration": {
    "sourceType": "CLUSTER",
    "clusterId": "e592a4f6-2e49-46e7-b6d0-248a7711197a",
    "pointInTime": "2026-09-30T14:30:00Z"
  }
}

Restore behavior

• the restoration creates a new cluster; the source cluster is not modified;
• the databaseName, user and extensions fields of the request body are ignored and replaced with the values from the source cluster;
• the new cluster is created with the CREATING state, then moves to the RUNNING state once ready.

Response example

{
  "id": "550e8400-e29b-41d4-a716-446655440001",
  "name": "my-restored-cluster",
  "status": {
    "state": "CREATING",
    "message": "cluster is being created"
  }
}

The response corresponds to the created cluster: its complete structure is identical to a cluster creation.

See the full specification