Setting up the DR cluster for WarehousePG Disaster Recovery

Prepare the DR cluster for recovery. Install WarehousePG and whpg-dr, create a restore configuration file, and run whpg-dr configure restore to prepare the cluster to receive restores from the storage repository. Run all commands as gpadmin on the DR coordinator. Once configured, proceed to choose a recovery method.

Whether to complete this step in advance depends on your DR mode. For warm DR, complete it before a failure occurs so the cluster can start replaying WAL immediately. For cold DR, you can do it at recovery time, though recording the cluster details in advance saves time when a failure occurs.

Prerequisites

Confirm that the DR cluster meets these requirements before starting:

• DR hosts provisioned and reachable from the DR coordinator over SSH as gpadmin. You need one coordinator host and enough segment hosts to hold all primary segments. The host count doesn't need to match the primary, but the segment count must. The DR hosts don't need to be the same hardware type as the primary — you can restore to different instance types or cloud providers, as long as the OS and WarehousePG version match.
• Access to the storage repository from all DR hosts. For NFS, the path must be mounted read-write. For S3, credentials must be available on all hosts. S3 requires WarehousePG 7.x.
• No standalone barman or barman-cli packages installed on any DR host. The edb-whpg-dr RPM conflicts with them.

Installing WarehousePG on the DR cluster

Install WarehousePG on all DR hosts at the exact same patch version and OS as the primary cluster. Follow the WarehousePG install guide through Creating the Data Storage Areas, then stop. When creating data directories, create only the coordinator and primary segment directories and leave them empty, as whpg-dr restore populates them from the base backup and rejects non-empty targets. Don't create mirror directories, and don't run gpinitsystem. Mirrors and standby coordinators aren't part of the restored cluster.

Install any extension software that the primary cluster uses. Extension data is restored automatically as part of the cluster state, but the software packages and libraries must already be present on the DR hosts before the restore starts. Anything listed in shared_preload_libraries is loaded at startup, and the cluster won't start if those libraries are missing.

Installing whpg-dr

Install the edb-whpg-dr RPM on the DR coordinator and all DR segment hosts. Run these steps as gpadmin on the DR coordinator.

1. Download the edb-whpg-dr package to the DR coordinator:

   → WrapCopy

   export EDB_SUBSCRIPTION_TOKEN=<your-token>
   curl -1sSLf "https://downloads.enterprisedb.com/$EDB_SUBSCRIPTION_TOKEN/gpsupp/setup.rpm.sh" | sudo -E bash
   sudo dnf download edb-whpg-dr

2. Create a file all_hosts listing every DR host:

   → WrapCopy

   cdw-dr
   sdw1-dr
   sdw2-dr
   sdw3-dr

3. Copy the RPM to all DR hosts:

   • WHPG 7
   • WHPG 6

   → WrapCopy

   gpsync -f all_hosts edb-whpg-dr-<version>.el9.x86_64.rpm =:/tmp/

   → WrapCopy

   gpscp -f all_hosts edb-whpg-dr-<version>.el9.x86_64.rpm =:/tmp/

4. Install the package on all DR hosts:

   → WrapCopy

   gpssh -f all_hosts -u gpadmin -e "sudo dnf install -y /tmp/edb-whpg-dr-<version>.el9.x86_64.rpm"

5. Add the whpg-dr bin directory to gpadmin's PATH:

   → WrapCopy

   echo 'export PATH=$PATH:/usr/edb/whpg-dr/bin' >> ~/.bashrc && source ~/.bashrc

6. Verify the installation on the DR coordinator:

   → WrapCopy

   whpg-dr --version

By default, whpg-dr stores its configuration, state, and logs under $HOME/.whpg-dr. To change this on the DR cluster, set WHPG_DR_HOME as described in Customizing the home directory.

Creating the restore configuration file

Create a YAML file that describes the storage repository and the DR cluster topology, then run whpg-dr configure restore to prepare the cluster. The storage block must point to the same repository the primary cluster writes to. The RPM installs ready-to-adapt starting points under /usr/edb/whpg-dr/examples/restore/.

The cluster name and storage configuration must match your primary cluster. The restore YAML also includes fields that don't exist in the backup configuration: coordinator_host, coordinator_data_directory, segment_hosts, data_directory, and data_directory_prefix. These fields describe your DR environment and can be set to match your infrastructure. The DR cluster doesn't need to match the primary's host layout, hostname scheme, or directory structure. data_directory accepts either a single path or a list of mount-point directories to spread segments across multiple disks; use data_directory_overrides to set a different path for specific hosts.

Using POSIX storage

Create a restore YAML file with storage.type: posix when the DR hosts can mount the same NFS path as the primary:

source_cluster_name: my_cluster

storage:
  type: posix
  path: /data/backups/whpg-dr

coordinator_host: cdw-dr
coordinator_data_directory: /data/coordinator

segment_hosts:
  - sdw1-dr
  - sdw2-dr
  - sdw3-dr

data_directory: /data/primary
data_directory_prefix: gpseg

For a full description of all available fields, see Restore configuration YAML.

Using S3 storage

Create a restore YAML file with storage.type: s3 if all DR hosts have network access to the S3 endpoint. S3 requires WarehousePG 7.x.

source_cluster_name: my_cluster

storage:
  type: s3
  bucket: my-dr-bucket
  prefix: whpg/my_cluster
  region: us-east-1
  access_key_id: your_aws_access_key_id
  secret_access_key: your_aws_secret_access_key

coordinator_host: cdw-dr
coordinator_data_directory: /data/coordinator

segment_hosts:
  - sdw1-dr
  - sdw2-dr
  - sdw3-dr

data_directory: /data/primary
data_directory_prefix: gpseg

To improve restore performance, set jobs to parallelize the base-backup copy within each segment, and wal_prefetch to fetch WAL files in parallel during replay. Changes to barman_options take effect on the next restore without requiring a new full restore. Re-run whpg-dr configure restore after updating the YAML.

For a full description of all fields including other credential methods, see Restore configuration YAML.

Overriding data directories per host

Use data_directory_overrides when one or more DR hosts have a different disk layout from the rest. Each value takes the same shape as data_directory, either a single path or a list of mount-point directories.

To point a specific host to a different base path:

segment_hosts:
  - sdw1-dr
  - sdw2-dr
  - sdw3-dr

data_directory: /data/primary
data_directory_prefix: gpseg

data_directory_overrides:
  sdw3-dr: /scratch/primary

To spread a host's segments across multiple mount points:

data_directory: /data/primary
data_directory_prefix: gpseg

data_directory_overrides:
  sdw3-dr:
    - /data1/primary
    - /data2/primary

Preparing the DR cluster for restore

Run whpg-dr configure restore to set up the DR cluster for restore operations. The command doesn't restore any data, but generates the necessary configuration files to run a restore:

whpg-dr configure restore /home/gpadmin/restore.yaml

The command:

• Reads segment_configuration.csv from the storage repository to map source segments to DR hosts. The segment count always matches the source, but the DR host count can differ, with content IDs distributed round-robin across segment_hosts.
• Generates Barman configuration files under $WHPG_DR_HOME/<cluster_name>/restore/barman_conf/ and distributes them to the coordinator and all DR segment hosts. For S3 storage, also writes AWS credential files to all hosts.
• Saves a copy of the restore YAML to $WHPG_DR_HOME/<cluster_name>/restore/config.yaml.
• Saves the segment mapping to $WHPG_DR_HOME/<cluster_name>/restore/restore.json.

Use --dry-run to preview what files will be generated without writing anything:

whpg-dr configure restore /home/gpadmin/restore.yaml --dry-run

After the command completes, check what backups and restore points are available in the repository:

whpg-dr list-backup my_cluster

By default, configure restore targets the backup associated with the latest restore point. To target a different backup, re-run the command with --backup-name using one of the names returned by list-backup:

whpg-dr configure restore /home/gpadmin/restore.yaml --backup-name 20260601T120000_base_backup

For next steps, see Cold DR, Warm DR, or Point-in-time recovery.