Iceberg considerations
Key considerations for optimizing performance and efficiency when targeting the Iceberg engine.
When working with Iceberg, keep these points in mind to ensure proper configuration and reliable
data handling:
- Ensure Iceberg tables exist at the time of mapping.
- Avoid using tables with the property
OBJECT_STORE_ENABLED. - Prevent direct writes to tables that define Iceberg partitions.
- Supports mapping for standard mirror, audit, and adaptive apply.
- Exclude row-level and subscription exits from Iceberg mappings.
- Disallow MBCS mappings, since Parquet requires UTF-8 strings.
- Define a key in Standard Apply mappings; otherwise, a default key is chosen.
- Utilize the Iceberg Version 2 specification in the client library.
- Write all data files in Parquet format, with Standard and Adaptive Apply mappings using equality deletes.
- Apply data with eventual consistency and guarantee delivery.