Maintain Catalog Authority and Metadata Freshness

Preserve Systems of Records

Concern	Authoritative system
Enterprise discovery, business context, ownership, and governance metadata	AI Catalog
Operational table identity, current snapshot, and commit history	The source catalog
Physical metadata and data objects	The object store or external data provider
Database-local objects, SQL privileges, views, and credential mappings	The individual Oracle AI Database

Note - Do not create competing authoritative definitions for the same asset without an explicit ownership boundary. Duplicate local external tables, catalog mounts, and enterprise records can resolve different table versions and confuse lineage.

Manage Metadata Freshness at the Correct Layer

Enterprise harvesting and database-local catalog caching are separate lifecycles. Refresh AI Catalog when enterprise discovery metadata is stale. Refresh Lake Cache only when cached table data is stale; it does not repair catalog metadata.

Example: Refresh the layer that is stale

Observation	Action
The mounted-table query works, but the asset is absent from enterprise search.	Refresh AI Catalog through its product-specific harvesting mechanism.
A new commit exists in the source catalog, but the database mount resolves older metadata.	Use the provider-supported DBMS_CATALOG metadata-cache refresh or flush operation.
Metadata is current, but eligible repeated scans continue to read the same remote data.	Populate or refresh Lake Cache according to instructions mentioned in Section Lake Cache.

For details about database-local catalog refresh and metadata-cache procedures, see DBMS_CATALOG.