Skip to content

Iceberg Integration

Ontul integrates with Apache Iceberg as a first-class, format-version 2 native catalog: every write is a v2 snapshot, with hidden partitioning, position-delete DELETE, RowDelta UPDATE/MERGE, schema and partition evolution, snapshot rollback, branches/tags, time travel, and Iceberg-spec views. The data path (Parquet read/write, field-id resolution, partition transforms, predicate evaluation, snapshot commits) is Ontul code distributed across workers; the Iceberg JAR is used only as a metadata library, and Polaris-side credential vending is bypassed so Ontul IAM stays the policy boundary.

Full documentation now lives in the dedicated Iceberg Integration section:

  • Overview — REST catalog setup (Polaris / AWS Glue), DDL/DML, file formats, time travel, views, federation, and access control.
  • Table Maintenance — on-demand ALTER TABLE … EXECUTE procedures (optimize, expire_snapshots, remove_orphan_files, rollback_to_timestamp, …) and the cron-schedulable auto-maintenance service.
  • Write-Audit-Publish (WAP) — stage writes on a branch, audit, then publish to main.