Model connection error with Object Level security

[ischo]

I tried to apply OLS to one of the collective tables of measurements in a cloud environment using the Tabular Editor.

Then, the following error phrase occurred. (※Since I am using the Korean version of Power BI Desktop, it is difficult to find the exact translation of the error, so I translated the phrase by myself. I wrote my username randomly.)

"The DirectQuery data source or internal model cannot be connected. Make sure the server and database names are correct and that you have permission to access them. The 'user (desktop account)' does not have access to the '(data source name) database' or the database does not exist."

Obviously, why did the model connections work properly before the OLS was applied
"Display Method:" (The function name is also translated by itself, so it may not be accurate.) If you see it in advance through the function, the above error phrase occurs, and if you post it on the power service as it is, you could not look up the document at the time of connection due to the role designated as ols.

The other reference columns in this document were filtered and it worked fine, but I don't understand why the model connection error is happening only when you specify the measurement set table...

Does anyone know about these cases or can help?

[v-saisrao-msft · Accepted solution]

Hi @ischo,

The issue happens because setting OLS None on a measure table hides all its measures for the role, making the table seem nonexistent. This can lead to confusing errors such as “DirectQuery data source cannot be connected” or “no measures available” in Excel. Avoid applying OLS to entire measure tables. Instead, secure only the sensitive columns or tables, or create explicit measures in Power BI Desktop for the role. This approach keeps visuals and PivotTables working while still protecting sensitive data.

Analysis Services tabular model object-level security | Microsoft Learn

Connect to Analysis Services data in Power BI Desktop - Power BI | Microsoft Learn

Object-Level Security (OLS) with Power BI - Microsoft Fabric | Microsoft Learn

Troubleshoot the connection from Excel to Power BI data - Power BI | Microsoft Learn

Thank you.

[ischo]

Thank you for your detailed answer.

I'd like to ask you an additional question.

As you said, in order to avoid applying OLS to the entire measurement table, I took some of the measurements that I want to apply OLS to, and created a new measurement set table and applied OLS, which is also causing the same error.

In this case, the reason seems to be that the entire measurement value & column & table referenced when filtering the measurement value is filtered, so the error appears, is my expectation correct?

Thank you.

[v-saisrao-msft]

Hi @ischo,

Your understanding aligns with how OLS behaves. Setting OLS to None on a measure table hides the table and its measures, which can cause dependent visuals or calculations to break. To avoid this, consider securing the underlying fact or dimension tables, or use DAX conditional logic (such as USERPRINCIPALNAME() ) to control measure visibility while keeping the model functional.

Object-Level Security (OLS) with Power BI - Microsoft Fabric | Microsoft Learn

Thank you.