Managed Identity for Compute Instances

We need to connect Azure Data Lake Storage Gen2 to Azure Machine Learning by means of a datastore. For security reasons we do not want to provide the credential-based authentication credentials (service principal or SAS token). Instead we want to connect with identity based access.

The problem we face is that we are not able to assign a managed identity to a compute instance ('ComputeInstance' object has no attribute 'add_identity'), so we can connect from notebooks to the Data Lake. In the documentation is explained how to assign a managed identity to a cluster, but we need the same for the compute instance, as it is the only way to run commands directly from the notebook.

Is there a way to assign managed identity to an Azure Machine Learning Compute Instance? Otherwise, we would like to know the best approach to overcome this issue, considering that we do not want to introduce the credentials in the code.

No se mencionan archivos ni pruebas. Empieza revisando el flujo documentado de identidad administrada para los clústeres de Azure Machine Learning y compáralo con la ComputeInstance API descrita en el issue. Se considerará terminado cuando se haya establecido si se admite el acceso basado en identidad para las instancias de cómputo y se haya identificado el enfoque compatible sin credenciales.

Escrito por el modelo de indexación a partir del texto del issue.