Catalog-level branching creates independent metadata histories or views across multiple catalog-managed tables without copying their underlying data. A branch can represent an entire environment, allowing coordinated changes while production remains isolated. section 7.5.6 section 11.2.2
The technique supports:
- Branching production for reproducible stress tests and experimentation. section 7.5.6
- Testing coordinated schema and data changes before merging them into the main environment. section 11.2.2
- Tagging historical states so reported metrics remain reproducible and referenced snapshots follow retention rules. section 11.2.2
Isolated stress-testing environment
A financial firm creates a daily Nessie branch from production, runs Spark-based simulations against the exact production state, and discards or merges results without duplicating large datasets. section 7.5.6
SELECT ref_name, type, snapshot_id, max_reference_age_in_ms FROM "db "."critical_fact_table.refs ";
Inspect reference metadata to distinguish branches and tags and track snapshot identifiers and reference-age policies. section 11.2.2
