Deployment Pipeline - Binding / Autobinding of lakehouse to a DirectLake semantic model not working

[aprenti67]

In the following scenario I have the lakehouse "microsoft_graph_api" in one workspace "Microsoft Graph - DEV" and the semantic model + report in another "Business_Unit-DEV".

The workspaces are in deployment pipelines and set up like this:

First I deploy the lakehouse from the Development to the Test stage in the "Microsoft Graph ETL" pipeline
Then I deploy the semantic model + report from Development to Test stage in the "Content Pipeline"

The problem is that the semantic model in the Test stage is still connected to the lakehouse in the Development stage

I have seen and tried the solution https://github.com/microsoft/semantic-link-labs/blob/main/src/sempy_labs/directlake/_update_directlake_model_lakehouse_connection.py but was unsuccessful in getting it to work. Am I on the right path?

[aprenti67]

The first run of directlake.update_direct_lake_model_connection resulted in the message "Multiple expressions found in the model. Please use the update_direct_lake_partition_entity function to update specific tables." I ran update_direct_lake_partition_entity function, then tried directlake.update_direct_lake_model_connection, but the result was the same

[aprenti67]

@m-kovalsky, any insight here would be appreciated or redirect to the right resources, as I am stuck here.

[m-kovalsky]

The update_direct_lake_model_connection function is intended for Direct Lake models with a single connection. It appears that your semantic model has multiple connections (different tables in your model point to different sources). If that's the case, the question is: how do you want to remap your table connections? Do you want all tables to source from a single source or do you want some tables to source from connection A and some from connection B? @SheilaHart

[aprenti67]

We would require the later: some tables from A remap to A' and some tables from B remap to B'. I can also envision a scenario (such as more static tables like dim_date) where A could remain A but B would remap to B'.

[SheilaHart]

@m-kovalsky My model only contains one connection to the warehouse established using DirectLake on OneLake. I see how DirectLake on OneLake opens up the possibility of using multiple connections to other lakehouses and warehouses within a single model. However, that is not the case in this model. I was able to update the connection by updating the connection string using tabular editor. I also successfully tested using the TMDL view to update the connection.

To answer your question, it seems like on models with multiple connections, we would want to be able to update some tables from connection A and some from connection B.

[m-kovalsky]

Fixed in 0.9.11. Use the new 'tables' parameter to specify which tables will be associated with the specified connection.

[aprenti67]

Excellent! I've successfully tested all the permutations I can imagine, as well as observing the expected behavior after redeployment from DEV to QA. Thank you for the expeditious and straight-forward solution!

"mixed" scenario example results:

[SheilaHart]

Fantastic! I was able to update the OneLake connection from Test to Prod using update_directlake_model_connection. Thank you for taking care of this so quickly. We appreciate your hard work. Thank you!!!