Forum Discussion
Skilled SharerSemantic Model Refresh activity gets disattached from the semantic model
I agree this looks like a product bug. I have also seen the Semantic model refresh activity lose its selected semantic model after a save and reopen, leaving the step valid but pointing at nothing and completing without error. The official docs cover how to configure the activity but do not mention this behavior, which suggests it is unintended.
This has also been reported here in other forum posts: Pipeline doesn't save changes to Semantic model refresh ActivityWhat I recommend now
Verify after saving
After you click Save, refresh the browser and reopen the activity to confirm the semantic model is still selected. This catches the silent no-op case early.Use an ID-based refresh instead of the UI binding
Trigger the refresh by workspaceId and datasetId using the Power BI REST API from a Web or Notebook activity. Docs: Datasets - Refresh and Get Refresh History. If you prefer a notebook approach, Semantic Link (SemPy) exposes helpers and a Fabric REST client you can call programmatically: SemPy FabricRestClient and Semantic Link overview.Add a post-run check
Immediately after the refresh step, call Get Refresh History for the target dataset and fail the pipeline if there is no recent entry.
If you found this useful, consider giving Kudos. If I solved your problem, mark this post as the solution.
Super UserTotal opinion based response here:
I, too, think that the Semantic Model Refresh task is not ready for prime-time. First, it asks for a list of tables (which is totally optional, but does not indicate such). And I have seen this behavior as well, or something similar.
Microsoft should downgrade the task back to "preview" and give it a thorough testing, including folks like amaaiia as front-line testers.