FIELD GUIDE 01
Open tables in practice
Test the engine pair, table features and catalog that your workload will use.
Test the format at the feature level.
Iceberg describes data files, snapshots and table metadata, so two readers can use the same underlying data when both support the table’s format version and features. Test partition and schema evolution, deletes and snapshot selection with the engines you will actually run.
Treat the catalog as a separate contract.
Record the REST catalog endpoint, table registration and credential flow. Specify whether the target only reads or also writes. For writes, test a conflicting commit and a retry in an isolated table before testing production cutover.
Measure the result as well as the connection.
Compare selected row counts, aggregates and checksums against a named snapshot. Include a representative predicate and record the query plan and performance threshold. Capture the engine and connector versions with the result.
Primary references
- SpecificationApache Iceberg table format specification
- ProtocolApache Iceberg REST catalog OpenAPI definition
- SnowflakeHorizon Catalog product reference
- DatabricksUnity Catalog and the open lakehouse