Recovery v1

In PostgreSQL terminology, recovery is the process of starting a PostgreSQL instance using an existing backup. The PostgreSQL recovery mechanism is very solid and rich. It also supports point-in-time recovery (PITR), which allows you to restore a given cluster up to any point in time, from the first available backup in your catalog to the last archived WAL. (The WAL archive is mandatory in this case.)

In EDB Postgres for Kubernetes, you can't perform recovery in place on an existing cluster. Recovery is instead a way to bootstrap a new Postgres cluster starting from an available physical backup.

The recovery bootstrap mode lets you create a cluster from an existing physical base backup. You then reapply the WAL files containing the REDO log from the archive.

WAL files are pulled from the defined recovery object store.

Base backups can be taken either on object stores or using volume snapshots.

You can achieve recovery from a recovery object store in two ways:

• We recommend using a recovery object store, that is, a backup of another cluster created by Barman Cloud and defined by way of the barmanObjectStore option in the externalClusters section.
• Alternatively, you can use an existing Backup object in the same namespace.

Both recovery methods enable either full recovery (up to the last available WAL) or up to a point in time. When performing a full recovery, you can also start the cluster in replica mode (see replica clusters for reference).

For recovery using volume snapshots:

• Use a consistent set of VolumeSnapshot objects that all belong to the same backup and are identified by the same k8s.enterprisedb.io/cluster and k8s.enterprisedb.io/backupName labels. Then, recover through the volumeSnapshots option in the .spec.bootstrap.recovery stanza, as described in Recovery from VolumeSnapshot objects.

Recovery from an object store

You can recover from a backup created by Barman Cloud and stored on a supported object store. After you define the external cluster, including all the required configuration in the barmanObjectStore section, you need to reference it in the .spec.recovery.source option.

This example defines a recovery object store in a blob container in Azure:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-restore
spec:
  [...]
  
  superuserSecret:
    name: superuser-secret

  bootstrap:
    recovery:
      source: clusterBackup

  externalClusters:
    - name: clusterBackup
      barmanObjectStore:
        destinationPath: https://STORAGEACCOUNTNAME.blob.core.windows.net/CONTAINERNAME/
        azureCredentials:
          storageAccount:
            name: recovery-object-store-secret
            key: storage_account_name
          storageKey:
            name: recovery-object-store-secret
            key: storage_account_key
        wal:
          maxParallel: 8

The previous example assumes that the application database and its owning user are named app by default. If the PostgreSQL cluster being restored uses different names, you must specify these names before exiting the recovery phase, as documented in "Configure the application database".

Recovery from VolumeSnapshot objects

EDB Postgres for Kubernetes can create a new cluster from a VolumeSnapshot of a PVC of an existing Cluster that's been taken using the declarative API for volume snapshot backups. You must specify the name of the snapshot, as in the following example:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-restore
spec:
  [...]

  bootstrap:
    recovery:
      volumeSnapshots:
        storage:
          name: <snapshot name>
          kind: VolumeSnapshot
          apiGroup: snapshot.storage.k8s.io

In case the backed-up cluster was using a separate PVC to store the WAL files, the recovery must include that too:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-restore
spec:
  [...]

  bootstrap:
    recovery:
      volumeSnapshots:
        storage:
          name: <snapshot name>
          kind: VolumeSnapshot
          apiGroup: snapshot.storage.k8s.io

        walStorage:
          name: <snapshot name>
          kind: VolumeSnapshot
          apiGroup: snapshot.storage.k8s.io

The previous example assumes that the application database and its owning user are named app by default. If the PostgreSQL cluster being restored uses different names, you must specify these names before exiting the recovery phase, as documented in "Configure the application database".

Recovery from a Backup object

If a Backup resource is already available in the namespace in which you need to create the cluster, you can specify the name using .spec.bootstrap.recovery.backup.name, as in the following example:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-example-initdb
spec:
  instances: 3

  bootstrap:
    recovery:
      backup:
        name: backup-example

  storage:
    size: 1Gi

This bootstrap method allows you to specify just a reference to the backup that needs to be restored.

The previous example assumes that the application database and its owning user are named app by default. If the PostgreSQL cluster being restored uses different names, you must specify these names before exiting the recovery phase, as documented in "Configure the application database".

Additional Considerations

Whether you recover from an object store, a volume snapshot, or an existing Backup resource, no changes to the database, including the catalog, are permitted until the Cluster is fully promoted to primary and accepts write operations. This restriction includes any role overrides, which are deferred until the Cluster transitions to primary. As a result, the following considerations apply:

• The application database name and user are copied from the backup being restored. The operator does not currently back up the underlying secrets, as this is part of the usual maintenance activity of the Kubernetes cluster.
• To preserve the original postgres user password, configure enableSuperuserAccess and supply a superuserSecret.

By default, recovery continues up to the latest available WAL on the default target timeline (latest). You can optionally specify a recoveryTarget to perform a point-in-time recovery (see Point in Time Recovery (PITR)).

Point in time recovery (PITR)

Instead of replaying all the WALs up to the latest one, after extracting a base backup, you can ask PostgreSQL to stop replaying WALs at any given point in time. PostgreSQL uses this technique to achieve PITR. The presence of a WAL archive is mandatory.

The operator generates the configuration parameters required for this feature to work if you specify a recovery target.

PITR from an object store

This example uses a recovery object store in Azure that contains both the base backups and the WAL archive. The recovery target is based on a requested timestamp.

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-restore-pitr
spec:
  instances: 3

  storage:
    size: 5Gi

  bootstrap:
    recovery:
      # Recovery object store containing WAL archive and base backups
      source: clusterBackup
      recoveryTarget:
        # Time base target for the recovery
        targetTime: "2023-08-11 11:14:21.00000+02"

  externalClusters:
    - name: clusterBackup
      barmanObjectStore:
        destinationPath: https://STORAGEACCOUNTNAME.blob.core.windows.net/CONTAINERNAME/
        azureCredentials:
          storageAccount:
            name: recovery-object-store-secret
            key: storage_account_name
          storageKey:
            name: recovery-object-store-secret
            key: storage_account_key
        wal:
          maxParallel: 8

In this example, you had to specify only the targetTime in the form of a timestamp. You didn't have to specify the base backup from which to start the recovery.

The backupID option is the one that allows you to specify the base backup from which to initiate the recovery process. By default, this value is empty.

If you assign a value to it (in the form of a Barman backup ID), the operator uses that backup as the base for the recovery.

If you don't specify the backup ID, the operator detects the base backup for the recovery as follows:

• When you use targetTime or targetLSN, the operator selects the closest backup that was completed before that target.
• Otherwise, the operator selects the last available backup, in chronological order.

PITR from VolumeSnapshot objects

The example that follows uses:

• A Kubernetes volume snapshot for the PGDATA containing the base backup from which to start the recovery process. This snapshot is identified in the recovery.volumeSnapshots section and called test-snapshot-1.
• A recovery object store in MinIO containing the WAL archive. The object store is identified by the recovery.source option in the form of an external cluster definition.

The recovery target is based on a requested timestamp.

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-example-snapshot
spec:
  # ...
  bootstrap:
    recovery:
      source: cluster-example-with-backup
      volumeSnapshots:
        storage:
          name: test-snapshot-1
          kind: VolumeSnapshot
          apiGroup: snapshot.storage.k8s.io
      recoveryTarget:
        targetTime: "2023-07-06T08:00:39"
  externalClusters:
    - name: cluster-example-with-backup
      barmanObjectStore:
        destinationPath: s3://backups/
        endpointURL: http://minio:9000
        s3Credentials:
          accessKeyId:
            name: minio
            key: ACCESS_KEY_ID
          secretAccessKey:
            name: minio
            key: ACCESS_SECRET_KEY

Recovery targets

Here are the recovery target criteria you can use:

targetTime : Time stamp up to which recovery proceeds, expressed in RFC 3339 format. (The precise stopping point is also influenced by the exclusive option.)

targetXID : Transaction ID up to which recovery proceeds. (The precise stopping point is also influenced by the exclusive option.) Keep in mind that while transaction IDs are assigned sequentially at transaction start, transactions can complete in a different numeric order. The transactions that are recovered are those that committed before (and optionally including) the specified one.

targetName : Named restore point (created with pg_create_restore_point()) to which recovery proceeds.

targetLSN : LSN of the write-ahead log location up to which recovery proceeds. (The precise stopping point is also influenced by the exclusive option.)

targetImmediate : Recovery ends as soon as a consistent state is reached, that is, as early as possible. When restoring from an online backup, this means the point where taking the backup ended.

This example uses a targetName-based recovery target:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
[...]
  bootstrap:
    recovery:
      source: clusterBackup
      recoveryTarget:
        backupID: 20220616T142236
        targetName: 'restore_point_1'
[...]

You can choose only a single one among the targets in each recoveryTarget configuration.

Additionally, you can specify targetTLI to force recovery to a specific timeline.

By default, the previous parameters are considered to be inclusive, stopping just after the recovery target, matching the behavior in PostgreSQL.

You can request exclusive behavior, stopping right before the recovery target, by setting the exclusive parameter to true. The following example shows this behavior, relying on a blob container in Azure for both base backups and the WAL archive:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-restore-pitr
spec:
  instances: 3

  storage:
    size: 5Gi

  bootstrap:
    recovery:
      source: clusterBackup
      recoveryTarget:
        backupID: 20220616T142236
        targetName: "maintenance-activity"
        exclusive: true

  externalClusters:
    - name: clusterBackup
      barmanObjectStore:
        destinationPath: https://STORAGEACCOUNTNAME.blob.core.windows.net/CONTAINERNAME/
        azureCredentials:
          storageAccount:
            name: recovery-object-store-secret
            key: storage_account_name
          storageKey:
            name: recovery-object-store-secret
            key: storage_account_key
        wal:
          maxParallel: 8

Configure the application database

For the recovered cluster, you can configure the application database name and credentials with additional configuration. To update application database credentials, you can generate your own passwords, store them as secrets, and update the database to use the secrets. Or you can also let the operator generate a secret with a randomly secure password for use. See Bootstrap an empty cluster for more information about secrets.

The following example configures the app database with the owner app and the password stored in the provided secret app-secret, following the bootstrap from a live cluster.

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
[...]
spec:
  bootstrap:
    recovery:
      database: app
      owner: app
      secret:
        name: app-secret
      [...]

With the above configuration, the following will happen only after recovery is completed:

1. If the app database does not exist, it will be created.
2. If the app user does not exist, it will be created.
3. If the app user is not the owner of the app database, ownership will be granted to the app user.
4. If the username value matches the owner value in the secret, the password for the application user (the app user in this case) will be updated to the password value in the secret.

How recovery works under the hood

You can use the data uploaded to the object storage to bootstrap a new cluster from an existing backup. The operator orchestrates the recovery process using the barman-cloud-restore tool (for the base backup) and the barman-cloud-wal-restore tool (for WAL files, including parallel support, if requested).

For details and instructions on the recovery bootstrap method, see Bootstrap from a backup.

The way it works is that the operator injects an init container in the first instance of the new cluster, and the init container starts recovering the backup from the object storage.

When the base backup recovery process is complete, the operator starts the Postgres instance in recovery mode. In this phase, PostgreSQL is up, though not able to accept connections, and the pod is healthy according to the liveness probe. By way of the restore_command, PostgreSQL starts fetching WAL files from the archive. (You can speed up this phase by setting the maxParallel option and enabling the parallel WAL restore capability.)

This phase terminates when PostgreSQL reaches the target (either the end of the WAL or the required target in case of PITR. You can optionally specify a recoveryTarget to perform a PITR. If left unspecified, the recovery continues up to the latest available WAL on the default target timeline (latest).

Once the recovery is complete, the operator sets the required superuser password into the instance. The new primary instance starts as usual, and the remaining instances join the cluster as replicas.

The process is transparent for the user and is managed by the instance manager running in the pods.

Restoring into a cluster with a backup section

A manifest for a cluster restore might include a backup section. This means that,after recovery, the new cluster starts archiving WALs and taking backups if configured to do so.

For example, this section is part of a manifest for a cluster bootstrapping from the cluster cluster-example-backup. In the storage bucket, it creates a folder named recoveredCluster, where the base backups and WALs of the recovered cluster are stored.

  backup:
    barmanObjectStore:
      destinationPath: s3://backups/
      endpointURL: http://minio:9000
      serverName: "recoveredCluster"
      s3Credentials:
        accessKeyId:
          name: minio
          key: ACCESS_KEY_ID
        secretAccessKey:
          name: minio
          key: ACCESS_SECRET_KEY
    retentionPolicy: "30d"

  externalClusters:
  - name: cluster-example-backup
    barmanObjectStore:
      destinationPath: s3://backups/
      endpointURL: http://minio:9000
      s3Credentials:

Don't reuse the same barmanObjectStore configuration for different clusters. There might be cases where the existing information in the storage buckets could be overwritten by the new cluster.

• On this page
• Recovery from an object store
• Recovery from VolumeSnapshot objects
• Recovery from a Backup object
• Additional Considerations
• Point in time recovery (PITR)
• Configure the application database
• How recovery works under the hood
• Restoring into a cluster with a backup section