Home / Databricks on Azure

Databricks on AzureRunbookAdvanced

Transformation: credential passthrough and legacy table ACLs to Unity Catalog permissions

Replace Entra ID credential passthrough and Hive table access control with Unity Catalog grants, external locations and row/column controls. Map ADLS ACLs to UC privileges, test with real users and switch clusters to UC access modes.

Deepak Bhardwaj2 min read

Applies toAzure Databricks clusters with credential passthrough or table access control, ADLS Gen2 ACLs, Unity Catalog

Why it has to go

Credential passthrough is a legacy feature (no new support, limited runtimes) and table ACLs only cover the workspace Hive metastore. Unity Catalog gives one permission model across workspaces, audit logs, lineage and fine-grained controls.

1. Capture the current access model

  • Passthrough: permissions live in ADLS ACLs / RBAC on storage. Export them:
az storage fs access show --account-name stlake --file-system raw --path sales --auth-mode login
  • Table ACLs: export grants from the workspace metastore:
SHOW GRANTS ON SCHEMA hive_metastore.sales;
SHOW GRANTS ON TABLE hive_metastore.sales.orders;

Write a table: who (group) → what (path/table) → which right (read/write).

2. Map to Unity Catalog

Old Unity Catalog
ADLS ACL read on a folder SELECT on tables in that folder; READ FILES on the external location/volume
ADLS ACL write MODIFY on tables; WRITE FILES on volume/location
Table ACL SELECT / MODIFY Same privileges on the UC table
USAGE on database USE CATALOG + USE SCHEMA
Separate views per audience Row filters / column masks, or views

Grant to Entra-synced groups, not users:

GRANT USE CATALOG ON CATALOG main TO `grp-sales-readers`;
GRANT USE SCHEMA, SELECT ON SCHEMA main.sales TO `grp-sales-readers`;
GRANT MODIFY ON TABLE main.sales.orders TO `grp-sales-engineers`;

3. Storage access moves to the access connector

Users no longer need storage RBAC/ACLs: UC reads storage through the access connector's managed identity. Plan to remove user storage permissions afterwards, otherwise people can bypass UC with direct abfss:// reads.

4. Tables

Upgrade tables into UC (see the Hive metastore to Unity Catalog / UCX post). UCX can also translate legacy table ACLs and assess passthrough clusters.

5. Clusters and testing

Switch clusters to standard (shared) or dedicated access mode. Test with a person from each group: the queries they ran yesterday should work, and things they shouldn't see should fail.

6. Clean up

Disable passthrough on remaining clusters, remove storage ACLs/RBAC for user groups, and keep only the access connector identities on the storage accounts.

Caution

Passthrough enforced per-user ADLS ACLs. If you grant UC privileges more broadly than those ACLs, people may suddenly see more data. Compare old and new access before switching.

Checklist

  • Current ACLs and grants exported
  • Mapping table reviewed by data owners
  • Grants applied to groups; filters/masks where needed
  • Users tested per group
  • Passthrough disabled; user storage permissions removed