Forum Discussion
BUG::semantic model inaccessible when creating a report with Power BI persona
- 2 years ago
That wasn't it. Was on a call with MS Support, it's internal bugs they are working on and hopefully, everyone cross fingers now, they will be fixed at some point in future.
Have you received any update on this? I'm seeing the same error so curious if you had it resolved
Not yet. For now, since we prefer to have more control over reports by using PBI Desktop, we use as a workaround the SQL analytics endpoint to connect to the lakehouse and will postpone using PBI Service with DirectLake connection to the lakehouse.
- Kryt0n2 years agoAdvocate I
Thanks for the response... as someone who knows little about PowerBI, the "PBI Service with DirectLake connection" is what the fabric environment uses, is that right?
I've given the desktop a try, I choose the "Power BI semantic models", I can see my lakehouse but when selecting it, it says "Working on it" but then the "Data" tab is empty and all links become disabled, did I do the same as you did or go wrong somewhere?
- Element1152 years agoMemorable Member
1__There are 3 different data access modes available to your reports (well, there's a 4th one, live connection, but I am not familiar with it, so I'll refer you to SQLBI website for a thorough explanation): Import, DirectQuery, and now, DirectLake.
Apparently, any report created in Fabric, that is, you don't start creating it with PBI Desktop, but with the PBI Service instead, any such report if it connects to a Fabric datasource, be it a lakehouse or a data warehouse, the connection will default to type DirectLake, which offers the advantages of both Import and DirectQuery modes without the disadvantages. This means fast as Import mode, but no duplication or copy of the data into the PBI model as with Import, and data changes at the source are immediately reflected as with DirectQuery mode. In terms of performance, DirectQuery is as fast as the data source, so if your source is slow, then the user will wait longer to see the data.
So in summary, DirectLake is supposed to give the user fast data refresh, and also access to huge volume of data that would be impractical to import (ie copy) into the PBI model. (Ask Bing Copilot about this.)
2__Nope, you didn't do anything wrong. I ran into the same issue and was bashing my head against the wall about this too. It's a PBI Desktop or Fabric or both (who the hell knows at this point?!) bug.
From the lakehouse menu ribbon, you can create a 'New semantic model', which will be in DirectLake mode and when you try to connect to that from Power BI Desktop, you'll get an error message because DirectLake mode is not supported right now in PBI Desktop. This means that the only way to create a report from that semantic model is to start with 'New report' from within Fabric.
To start your report from PBI Desktop, the workaround is to select as your data source the SQL analytics endpoint of your lakehouse. That will behave like when you connect to a SQL Server DB and you will be able to choose your lakehouse tables. The connections you will be offered will be twofold: Import or DirectQuery.
Here is the thread I already started on this a week or two ago: Solved: BUG or FEATURE???::DirectLake mode not supported - Microsoft Fabric Community