A cross-table transaction coordinates changes across related tables so they become visible together as one consistent, atomic commit. In a lakehouse, a catalog coordination layer such as Project Nessie delegates table metadata and data-file management to Apache Iceberg while tracking their evolution. section 7.4.5
A practical implementation pattern uses catalog-level versioning:
- Create a dedicated branch for an ingestion pipeline or analytics job. section 7.4.5
- Write coordinated changes across multiple tables without exposing incomplete results. section 7.4.5
- Merge the completed branch into the mainline atomically, or roll it back when validation fails. section 7.4.5
- Verify that the catalog and connected engines support REST-compliant writes before depending on the capability. section 4.1.3 section 4.3.3
Coordinated ingestion
An ingestion pipeline updates related sales and customer datasets on a work branch. After both writes succeed, merging the branch publishes the changes together; a failed pipeline leaves the mainline unchanged. section 7.4.5
