DuckDB and Apache Iceberg: OLake Go guides

Q: How does DuckDB attach AWS Glue?

With ATTACH and ENDPOINT_TYPE glue. The reader also needs Lake Formation grants.

DuckDB attaches Iceberg REST (representational state transfer) catalogs from a single process. It is a quick way to query the tables OLake Go writes.

Short answer. To query a replicated database in DuckDB, run OLake Go with an Iceberg REST catalog such as Polaris, Lakekeeper, Unity Catalog or Horizon, or with AWS Glue. Then attach that catalog in DuckDB with the iceberg extension. Add DuckDB as a target query engine and OLake Go writes positional deletes, which DuckDB reads.

Key takeaways

DuckDB attaches Iceberg REST catalogs, and AWS Glue, from a single process with the iceberg extension.
With DuckDB as a target query engine, OLake Go writes positional deletes, which DuckDB reads.
Two guides end in DuckDB, such as Postgres to DuckDB, and many more use it as the cross-check reader.

Last reviewed October 2026.

How OLake Go gets data into DuckDB

OLake Go uses change data capture (CDC) where the source has it. It commits Iceberg tables to the catalog. DuckDB reads Polaris, Lakekeeper and the vendor REST catalogs, and Glue through its own endpoint type.

Two guides end in DuckDB, and many more use it as the cross-check reader. The MySQL query engines post compares DuckDB with other readers.

When to use it

You want to inspect replicated tables from a laptop or a notebook.
The catalog speaks the Iceberg REST protocol.

When not to use it

The catalog is Hive Metastore or JDBC. DuckDB attaches REST catalogs only.
DuckDB is the main reader and the job uses equality deletes. The DuckDB docs do not cover equality delete reads, so prefer positional deletes.

Catalogs DuckDB reads

Catalog	DuckDB reads	How
AWS Glue	Yes	ATTACH with ENDPOINT_TYPE glue (needs Lake Formation grants)
Apache Polaris	Yes	ATTACH to the REST endpoint
Lakekeeper	Yes	ATTACH to the REST endpoint
Snowflake Horizon	Yes	ATTACH to the Horizon REST endpoint with a Snowflake token
Unity Catalog	Yes	ATTACH to the Unity REST endpoint (AWS-hosted workspaces)
BigLake (Google Cloud Lakehouse)	Yes	ATTACH with a Google access token
Amazon S3 Tables	Yes	ATTACH with ENDPOINT_TYPE s3_tables
Hive Metastore	No	DuckDB attaches REST catalogs only
JDBC	No	DuckDB attaches REST catalogs only

Delete formats DuckDB reads

Equality deletes (v2)	Positional deletes (v2)	Deletion vectors (v3)
Partial	Yes	Yes

DuckDB reads positional deletes and, since duckdb-iceberg 1.5.3, deletion vectors. Equality delete reads are undocumented; they returned correct rows on the runs on this site (duckdb-iceberg 1.5.x, checked 2026-10-29), but prefer positional deletes when DuckDB is the main reader. Set DuckDB as a target query engine on the OLake Go job and OLake Go picks a format it reads; with DuckDB as the only reader that is positional deletes (v2).

Sources: DuckDB: Iceberg REST catalogs (checked 2026-10-12); DuckDB: new Iceberg features (v3 support) (checked 2026-10-12).

Tested guides for DuckDB

Guides where DuckDB reads the same tables

These pages show DuckDB reading the same tables through the same catalog, in their cross-engine section. No second copy of the data is made.

Frequently asked questions

Does DuckDB read Iceberg deletion vectors?

Yes, since duckdb-iceberg 1.5.3.

Does DuckDB read equality deletes?

The DuckDB docs do not list it. On the runs on this site they returned correct rows with duckdb-iceberg 1.5.x, checked 2026-10-29, but positional deletes are the safer choice.

How does DuckDB attach AWS Glue?

With ATTACH and ENDPOINT_TYPE glue. The reader also needs Lake Formation grants.

How OLake Go gets data into DuckDB​

When to use it​

When not to use it​