topic Direct Lake semantic model not picking up new columns after gold layer schema change in Data Engineering

Direct Lake semantic model not picking up new columns after gold layer schema change

robertozsr — Wed, 06 May 2026 11:05:59 GMT

Hi everyone,

I added a new column check_completed_timestamp and rename a column from check_timestamp to created_check_timestamp to my fact_check gold Delta table using PySpark with overwriteSchema = True.

Now when I refresh my Direct Lake semantic model I get:

Refreshing data
An error has occurred.
Show details
Underlying Error Delta protocol violation: the column '<oii>check_timestamp</oii>' is not found in delta table '<oii>fact_check</oii>'. Table: fact_check. Partition: fact_check.

The Delta table looks correct when I query it directly — the column is there.

I get that the issue is because I changed the column name in the delta table.

Anyone know how to force the semantic model to pick up the new schema?

I have tried by clicking edit tables, refreshing from there, then clicked refresh, but nothing. (normally this is what it works)

This is the first time this happening

Thanks

Re: Direct Lake semantic model not picking up new columns after gold layer schema change

deborshi_nag — Wed, 06 May 2026 15:14:38 GMT

Hello @robertozsr

Refreshing a semantic model in the usual way won't update the schema if you've made changes to the underlying tables, such as adding, removing, or renaming columns. To update the schema, you need to reload the table using the Edit table option.

Open your semantic model in Fabric or Power BI service. In the menubar, select "Edit tables." You'll see a refresh button next to the search bar—click it to reload the schema. Alternatively, you can uncheck and recheck the tables to trigger a schema update.

Re: Direct Lake semantic model not picking up new columns after gold layer schema change

robertozsr — Wed, 06 May 2026 15:34:39 GMT

Hello @deborshi_nag this is what I usually do: now it is working again. I think the related issue was that I was not able to query the end point, because the owner of the admin portal had to log in again to refresh th token. And indeed now it is working 🙂 thanks a lot!