How to create a Foreign Table in a Lakehouse on Microsoft Fabric

João Barros 05 de August de 2026 4 min read

This tutorial shows how to create a Foreign Table in the Lakehouse of Microsoft Fabric to query external data (for example, a Parquet file or an external table). The Foreign Table lets you extend the Lakehouse with external sources without copying data, useful for integrating systems and reducing duplication.

Prerequisites

Account with access to Microsoft Fabric and permissions to edit a Lakehouse.
An existing Lakehouse in OneLake or permission to create a new one.
Accessible external source (e.g., Parquet file in OneLake or external endpoint with credentials).
Basic T-SQL knowledge to run commands in the Notebook or Warehouse.

Step 1: Prepare the external source and get the path

Identify the external source you want to use. If it is a Parquet file in OneLake, copy the relative path in the Lakehouse. If it is an external endpoint, confirm the URL and credentials. The Foreign Table creates a logical pointer to that location.

Step 2: Open a Notebook or Warehouse to run T-SQL

Choose where you will run the commands: a Notebook with T-SQL language or a Warehouse (T-SQL). Open the resource and connect it to the Lakehouse where you will create the Foreign Table. Using a T-SQL session allows you to create objects and run immediate queries.

Step 3: Create an external data source (optional as needed)

Some sources require first defining an external data source with credentials or access settings. For OneLake/simple paths, this may not be necessary. Example of creating a data source for a supported endpoint:

CREATE EXTERNAL DATA SOURCE MyExternalSource
WITH (
    LOCATION = 'abfss://@.dfs.core.windows.net/folder/',
    CREDENTIAL = MyCredential
);

Replace MyCredential with a credential previously created in the environment, if required.

Step 4: Create the Foreign Table with T-SQL

Use CREATE EXTERNAL TABLE or CREATE FOREIGN TABLE according to Fabric support to map the source structure. Here is a generic example that points to a Parquet file in a Lakehouse path.

CREATE EXTERNAL TABLE dbo.ForeignSales (
  SaleID bigint,
  SaleDate datetime2,
  Amount decimal(18,2)
)
WITH (
  LOCATION = '/lakehouse/path/to/folder/sales.parquet',
  DATA_SOURCE = MyExternalSource,
  FILE_FORMAT = 'Parquet'
);

If your source is a folder with multiple Parquet files, the LOCATION can point to the folder. If you did not use DATA_SOURCE, specify only the absolute LOCATION in OneLake.

Step 5: Define compatible schemas and types

Confirm that the T-SQL types you defined match the actual data. Common errors: incompatible types (e.g., datetime vs string) or null fields without allowing NULL. Adjust the table definition as needed.

Step 6: Test queries and performance

After the Foreign Table is created, run queries to validate. Avoid SELECT * on large volumes; filter and limit to test. If necessary, create views or materializations in the Lakehouse to improve performance.

-- Simple test
SELECT TOP 10 SaleID, SaleDate, Amount
FROM dbo.ForeignSales
ORDER BY SaleDate DESC;

-- Count to validate volume
SELECT COUNT(*) AS TotalRows FROM dbo.ForeignSales;

Verify the result

Confirm that queries return the expected data and that observed types make sense. Check the existence of the object with a catalog lookup and inspect metadata. Common errors include incorrect paths, permissions and file formats. If there are no results, validate the LOCATION and the data source permissions.

Conclusion

Creating a Foreign Table in the Lakehouse of Microsoft Fabric allows you to query external data without duplicating files, facilitating integration and saving space. Next steps: create views that join external data with internal tables, or materialize results to optimize performance. Tip: when testing, always start with LIMIT/TOP to avoid cost and performance surprises — which source do you want to connect next?