How-to guideApplies to: Unified engine deployments
Use this Google identity method for one BigLake catalog connection.
Before you start
Provide application default credentials for the source identity on the
credential-service host. Allow that identity to impersonate the target service
account. Grant the target access to the catalog and its data.
Use an existing catalog with at least one readable Iceberg table. The engine and
workers must reach the catalog and its table storage. Have the remote identity
and permissions ready before creating the Unified registration. Replace the
namespace and table with one you know has data. The service must return usable
temporary storage credentials, or the table location must already have
configured storage access in Unified. Requesting delegation alone does not grant
access.
Values to collect
- Identity-specific values below; BigLake URI, warehouse, billing project,
namespace, and table.
Run the SQL
Replace every <placeholder> before running. Use unused names for the resources
below. Run the statements in order in Workbench.
Check the result
The SELECT should return up to five rows from the chosen table. Use a nonempty
table for this check. This verifies both metadata and table-file access;
creating the registration alone does not verify either.
Clean up
This removes only the Unified registrations. It leaves the remote catalog and
its tables in place.
Last modified on November 4, 2026
