An Apache Spark-based analytics platform optimized for Azure.
Deleted Azure Databricks clusters themselves cannot be restored. Clusters are compute resources; when deleted, they must be recreated (for example, by reapplying the previous configuration or IaC templates such as ARM/Bicep/Terraform, if available).
For data and metadata protection, rely on the underlying storage and database recovery mechanisms rather than the cluster:
- Persist all important data to durable storage (such as Azure Storage, Delta tables, or Lakehouse/Lakebase) so it is not lost with the cluster.
- For OLTP/Lakebase databases, use features like point-in-time restore, snapshots, or child instances to recover data states if needed.
- For MLflow experiments and runs, deleted runs can be restored from the UI or via
mlflow.restore_runs.
If the concern is disaster recovery of the overall Databricks environment, use a regional disaster recovery architecture where data is stored in resilient storage and Databricks workspaces/clusters can be recreated in another region.
References: