An orphaned file is a data or metadata file written to storage but never committed to a valid Iceberg snapshot. It does not affect query correctness, but it increases storage costs and complicates audits. section 11.3.2
Manage orphaned files as part of table maintenance:
- Run Iceberg’s remove_orphan_files procedure to reclaim storage from failed or incomplete writes. section 11.3.2
- Retain a configurable age threshold; the default is commonly 10 days, with a 24-hour minimum, to avoid deleting files from in-flight or recently failed operations. section 11.3.2
- Coordinate orphan cleanup with compaction and snapshot expiration when implementing end-to-end deletion maintenance. section 11.2.5
Figure 11.4 With object versioning, if an object is corrupted or lost, you can revert to the previous version of the object. Here, the object is a data or metadata file in storage.
Object versioning provides a recovery layer for data or metadata files if storage objects are corrupted or lost. section 11.3.2
