There are two primary reasons for restoring a cluster: corrupted data has caused your cluster to malfunction, or you want to downgrade to a previous version of HPE Machine Learning Data Management due to an upgrade issue.

How to Restore a Cluster

Depending on your scenario, you might pick all or a subset of the following steps:

  1. Populate new pachyderm and dex databases on your PostgreSQL instance.
    • You can use the pg_restore command to restore the databases from the backup files.
    • If you are using a cloud provider, you can use their tools to restore the databases.
  2. Populate a new bucket or directly use the backed-up object-store.
    • You can use the gsutil or aws CLI to copy the objects from the backed-up object store to your new bucket.
  3. Create a new empty Kubernetes cluster and give it access to your databases and bucket.
  4. Deploy HPE Machine Learning Data Management into your new cluster by updating the Helm values to point to the new databases and object store.
     helm upgrade --install <release-name> pachyderm/pachyderm -f <path-to-values-file>
  5. Connect pachctl to your restored cluster and check that it is up and running.