OpenSharing (formerly Delta Sharing)
Databricks · Unity Catalog & Governance
Theory
OpenSharing, formerly Delta Sharing, is Databricks' open protocol for sharing live data securely with other teams and companies without copying it. The provider creates a share, a read-only collection of tables, views, volumes, models or notebooks, and a recipient that represents the consumer. Granting the share to the recipient lets the consumer read the current data directly from the provider's storage, governed by Unity Catalog.
There are two flavours. Databricks-to-Databricks sharing connects two Unity Catalog metastores: the recipient is identified by its sharing identifier and mounts the share as a catalog. Open sharing (Databricks-to-Open) gives a bearer-token credential file so non-Databricks tools such as pandas, Power BI or Spark elsewhere can read the data. For example, TechTown shares main.sales.orders with a partner as share partner_sales; the partner queries it as a read-only catalog and always sees today's data.
Why we use it
Emailing CSV extracts or copying data to FTP creates stale, ungoverned copies. OpenSharing gives partners and other business units live, read-only access that you can audit and revoke, without moving data into another platform.
When to use it
Use it for sharing across organizations, clouds, regions or metastores. Inside one metastore, plain GRANT statements are simpler. Consumers who want to discover public or paid datasets use Databricks Marketplace, which is built on the same sharing technology.
How it works
- Provider creates a share
- Add tables, views or volumes to the share
- Create a recipient
- Does the recipient use Databricks with Unity Catalog? Yes: Databricks-to-Databricks: use their sharing identifier. No: Open sharing: send the token credential file.
- GRANT SELECT ON SHARE to the recipient
- Recipient queries live, read-only data
Syntax
CREATE SHARE name
ALTER SHARE name ADD TABLE main.sales.orders [WITH HISTORY]
CREATE RECIPIENT name [USING ID '<sharing-identifier>']
GRANT SELECT ON SHARE name TO RECIPIENT recipient
SHOW ALL IN SHARE name | DESCRIBE RECIPIENT name
Recipient side: CREATE CATALOG c USING SHARE provider.shareExample
-- Provider workspace
CREATE SHARE IF NOT EXISTS partner_sales
COMMENT 'Orders for Pune distributor';
ALTER SHARE partner_sales ADD TABLE main.sales.orders;
CREATE RECIPIENT IF NOT EXISTS pune_distributor
USING ID 'aws:ap-south-1:<metastore-uuid>'; -- Databricks-to-Databricks
GRANT SELECT ON SHARE partner_sales TO RECIPIENT pune_distributor;
-- Recipient workspace
CREATE CATALOG techtown_shared USING SHARE techtown.partner_sales;
SELECT region, SUM(amount) FROM techtown_shared.sales.orders GROUP BY region;Key points
- OpenSharing is the current name for Delta Sharing; the SQL still uses SHARE and RECIPIENT
- A share is read-only and live: recipients see the provider's current data
- Databricks-to-Databricks uses a sharing identifier; open sharing uses a token credential file
- Recipients mount a share as a catalog with CREATE CATALOG ... USING SHARE
- Revoke access with REVOKE SELECT ON SHARE ... FROM RECIPIENT