Search Shortcut cmd + k | ctrl + k

DuckLake is a lakehouse format built on SQL

DuckLake delivers advanced data lake features without traditional lakehouse complexity by using Parquet files and a SQL database

DuckLake 1.0 is out Read the docs

Deployed in production

Austrian Supply Chain Intelligence Institute

PostgreSQL
SQLite
DuckDB, in-process
DuckDB + Quack

Client

Multiple clients
Works locally or in the cloud
No vendor lock-in

DuckLake

Catalog:

The catalog is served by an ACID-compliant SQL database

Storage: Parquet files

Parquet files can be stored on local disk or in object storage

DuckLake’s key features

SQL as a lakehouse format

DuckLake stores all metadata in a SQL database – PostgreSQL, MySQL, SQLite, or DuckDB. No custom catalog server required.

Data lake operations

DuckLake supports snapshots, time travel queries, schema evolution, and partitioning.

Snapshots

You can have as many snapshots as you want without expensive compaction steps.

ACID transactions

DuckLake allows concurrent access with ACID transactional guarantees over multi-table operations.

Open Parquet storage

Your data lives in plain Parquet files on local disk or object storage. The Parquet files are compatible with Iceberg.

Fast queries

DuckLake uses statistics for filter pushdown, enabling fast queries even on large datasets.

In conversation: DuckLake 1.0

Listen to Mark Raasveldt and Pedro Holanda discuss the road that led
to the DuckLake specification and explain the features of DuckLake 1.0.

Create your first DuckLake

DuckDB provides first-class support for DuckLake and can use PostgreSQL, SQLite, or DuckDB as the catalog database. With Quack, a DuckDB catalog database can be shared with multiple DuckDB clients over the network.

Documentation Live demo

INSTALL ducklake;

ATTACH 'ducklake:metadata.ducklake'
    AS my_ducklake
    (DATA_PATH 'data/');

USE my_ducklake;

INSTALL ducklake;

ATTACH 'ducklake:metadata.ducklake'
    AS my_ducklake
    (DATA_PATH 'data/');

USE my_ducklake;

INSTALL ducklake;
INSTALL postgres;

ATTACH 'ducklake:postgres:dbname=your_db host=your_host'
    AS my_ducklake (DATA_PATH 'data/');

USE my_ducklake;

INSTALL ducklake;
INSTALL sqlite;

ATTACH 'ducklake:sqlite:metadata.sqlite'
    AS my_ducklake (DATA_PATH 'data/');

USE my_ducklake;

-- Server: open the catalog database file
-- (e.g., duckdb metadata.ducklake) and run:
INSTALL quack;
CALL quack_serve('quack:localhost');

-- Client: use the token returned by quack_serve
INSTALL ducklake;
INSTALL quack;
CREATE SECRET (TYPE quack, TOKEN 'your_token',
    SCOPE 'quack:localhost');

ATTACH 'ducklake:quack:localhost'
    AS my_ducklake (DATA_PATH '/path/to/shared/data/');

USE my_ducklake;

Frequently asked questions

Do you have any questions about DuckLake? We got you covered.

Why should I use DuckLake?

DuckLake provides a lightweight one-stop solution if you need a data lake and catalog.

If you use DuckLake from DuckDB, it can unlock a “multiplayer DuckDB” experience with multiple DuckDB instances reading and writing the same dataset – a concurrency model not supported by vanilla DuckDB.

What is “DuckLake”?

“DuckLake” can refer to a number of things:

The DuckLake format that uses a catalog database and a Parquet storage to store data.
A DuckLake instance storing a dataset with the DuckLake lakehouse format.
The ducklake DuckDB extension, which supports reading/writing datasets using the DuckLake format.

Is DuckLake production-ready?

Yes! We published DuckLake 1.0 in April 2026, which is a production-ready release with guaranteed backward-compatibility.

What is the license of DuckLake?

The DuckLake specification and the ducklake DuckDB extension are released under the MIT license.

More FAQs