MLOps + DLT

[PassionateDBD]

What are the best practices for using MLOps and DLT together?

This page https://learn.microsoft.com/en-us/azure/databricks/compute/access-mode-limitations states that 

You cannot use a single user cluster to query tables created by a Unity Catalog-enabled Delta Live Tables pipeline, including streaming tables and materialized views created in Databricks SQL. To query tables created by a Delta Live Tables pipeline, you must use a shared cluster using Databricks Runtime 13.1 and above.

MLOps cluster only supports single user access mode as stated on this page https://learn.microsoft.com/en-us/azure/databricks/machine-learning/ 

To access data in Unity Catalog for machine learning workflows, the access mode for the cluster must be single user (assigned). Shared clusters are not compatible with Databricks Runtime for Machine Learning. 

Is it possible to use an ML cluster with dlt + unity catalog somehow? If not then will it become possible in near future and is there a workaround meanwhile?

[Kaniz]

Hi @PassionateDBD ,

• As per the provided links, DLT tables created by a Unity Catalog-enabled pipeline cannot be queried using a single-user cluster. Instead, a shared cluster with Databricks Runtime 13.1 and above is required.
• MLOps clusters support only single-user access mode, which is not compatible with shared clusters used for DLT.
• Unfortunately, there is no direct workaround to use an ML cluster with DLT and Unity Catalog simultaneously.

For now, consider using separate clusters for MLOps and DLT workflows.