An external data source is a reusable, named connection that lets Transformations read from storage outside Cognite Data Fusion (CDF). For example, you can read Microsoft Fabric OneLake without staging a full copy in CDF RAW. Configure credentials and location once, then reference the source in transformation SQL by externalId.
This API is enabled per project. If requests return that the endpoint isn’t enabled, contact Cognite Support. This is separate from the on-premises Cognite Fabric connector.
Concepts
- Source type — The external system the source connects to. The supported value is
one_lake for Microsoft Fabric OneLake.
- Credentials — Microsoft Entra ID application credentials:
clientId, tenantId, and clientSecret. The client secret is write-only; it is stored encrypted and never returned when you read a data source.
- Location — Where in OneLake the source points:
workspaceId and containerId, both as GUIDs.
- Data set scoping — Optional
dataSetId that scopes access to the data source through CDF data set permissions.
Usability checks
Before you use an external data source in a transformation, call the usability endpoint (POST /transformations/externaldata/usability) to verify that CDF can reach the storage location with the stored credentials. The response reports whether the connection and permissions succeed, so you can fix configuration issues before a transformation job fails.
Key capabilities
- List external data sources configured for the project
- Create or replace external data sources, matched on
externalId
- Delete external data sources by
id or externalId
- Verify usability of a data source before using it in a transformation
Rate and concurrency limits
External data source operations share the general CDF request budgets. For limit types and throttling behavior, see API rate limits. Last modified on August 10, 2026
