Forum Discussion

Nikos83's avatar
Nikos83
Advocate I
1 year ago
Solved

Refreshing semantic model does not work

Hi Community, Since yesterday, we’re encountering major issues refreshing our semantic models in Microsoft Fabric. We consistently receive the following error: Failed to save modifications to the ...
admin
  • Nikos83's avatar
    Nikos83
    1 year ago

    The issues with our semantic models have been resolved. They were related to a problem that recently occurred in the French region. It turned out to be a synchronization issue in the metadata coming from the lakehouse.

    All our semantic models can now be updated again, and the dashboards are displaying correctly.

rohit1991's avatar
rohit1991
Super User
1 year ago

Hi Nikos83 

Root Cause: Even though the Delta table exists and is accessible via SQL Endpoint, Microsoft Fabric sometimes fails to sync the table's metadata properly with the semantic model especially after schema changes or table recreation.

 

 Steps :

  1. Go to the Lakehouse where your Delta table is stored.
  2. In the “Tables” section, delete the existing table reference (not the actual Delta file, just the reference).
  3. Click “New Table” >> “From Files”, and re-add the same Delta table again.
    >> This forces a fresh metadata sync with the Lakehouse.
  4. Then, go to your Semantic Model, remove the old table (if still linked), and add the newly re-linked table from the Lakehouse again.
  5. Now try refreshing the semantic model
Nikos83's avatar
Nikos83
Advocate I
1 year ago

Thanks for your quick repley.

I’ve already created an entirely new Lakehouse with a brand-new table in it, and I’m experiencing exactly the same issue there. I think microsoft is having some problems related to their meta data synchronisation 

  • rohit1991's avatar
    rohit1991
    Super User
    1 year ago

    Hi Nikos83 


    Thanks for the update and you're absolutely right, Microsoft Fabric has had ongoing intermittent issues with metadata sync, especially when:

     

    1. Tables are recreated (even with new names),
    2. Or when schema changes are involved,
    3. And even in brand-new Lakehouses as you mentioned.

    Try this forced metadata rebuild trick:

    1. Delete the semantic model completely from the workspace (not just refresh).
    2. In the Lakehouse:

              a.  Go to the Delta table folder >> Temporarily rename the *_delta_log folder (e.g., to                                               _delta_log_backup)

              b. Wait a few seconds, then rename it back to _delta_log.

                  This forces Fabric to re-evaluate the table metadata on next connection.

    1. Then recreate the semantic model from scratch using the re-added table.

     Optional Check:

    Open the SQL Endpoint and run:

    sqlCopyEditSELECT * FROM INFORMATION_SCHEMA.TABLES