Delta Sharing: Databricks-to-Databricks

Key insights

• Databricks-to-Databricks Delta Sharing: A native protocol that lets Unity Catalog–enabled Databricks workspaces share Delta tables and other assets securely across metastore boundaries.
  It keeps data in the provider’s cloud storage while recipients access it through their own workspace catalog view.
• No token management: The protocol removes manual credential exchange by using metastore identifiers and built-in server-to-server authentication.
  This reduces operational overhead and lowers the risk of leaked tokens.
• Share and Recipient workflow: Providers create a Share (collections of assets), register a Recipient using the recipient’s metastore identifier, and then grant the share.
  Recipients mount the shared catalog into their workspace to query data directly.
• Supported assets: Beyond tables, the protocol supports views, notebooks, volumes and AI models, enabling richer collaboration than simple file sharing.
  Changes to shared assets propagate quickly so recipients see near-real-time updates.
• Permissions model: Unity Catalog enforces granular access controls; metastore admins can grant broad rights but a minimal-permission approach limits risk.
  Be careful about allowing wide catalog creation or granting excessive privileges.
• Cross-cloud and cross-account use: Works across Azure, AWS, and GCP when both sides use Unity Catalog, making it suitable for enterprise partner collaborations.
  Sharing outside your organization may require toggling external-sharing settings and additional governance checks.

Keywords

Delta Sharing, Databricks data sharing, Databricks-to-Databricks data sharing, Delta Lake sharing, Secure data sharing Databricks, Cross-account data sharing, Open data sharing protocol, Real-time data sharing Databricks