Semantic model got corrupted

Hi powerbiexpert22​,

Yes, restoring the semantic model through Version History was a valid recovery approach. Microsoft now specifically documents Version History as the recommended first recovery option when a semantic model becomes corrupted and is loaded in metadata-only mode.

There are a few details I would keep in mind.

Version History is not limited to web edits. Once Version History has been initialized for the semantic model, publishing a PBIX from Power BI Desktop or uploading a PBIX also captures the semantic model state immediately before that publish.

Microsoft documents that each saved version contains the semantic model metadata and data.

However, I would not treat it as full source control. Power BI currently retains up to five versions, and restoring versions older than 14 days is officially unsupported. Microsoft recommends Git integration when you need longer-term source control and more complete version history.

For RLS, I would distinguish the role definition from the member assignments.

Microsoft documents that RLS roles and rules are defined in the semantic model, while users and groups are assigned to those roles separately in the Power BI Service.

Because Version History stores the semantic model metadata, the RLS role definitions from the restored version should be part of the model restoration. However, I do not see Microsoft documenting the Service-side RLS member assignments as versioned content, so I would explicitly open the semantic model Security page after the restore and verify the role memberships.

I would do the same for the Power BI app configuration. App audience mappings are configured at the app/service level rather than inside the semantic model itself, so I would not expect Version History to act as a restore mechanism for app audiences. Verify the app audience and any relevant report/semantic-model permissions separately after recovery.

One other important step is to refresh the semantic model after restoring it. Microsoft notes that the data contained in the restored version can be older than the latest source data.

Since your workspace is on Premium capacity, you may also want to configure Power BI Premium Backup and Restore for future disaster recovery. Microsoft supports XMLA-based semantic model backups to ADLS Gen2 specifically for scenarios such as corruption or loss.

For longer-term development protection, I would also keep the semantic model in PBIP/Git rather than relying only on Version History.

So in your current situation, I would:

• restore the known-good version
• refresh the semantic model
• verify RLS roles and Service-side role membership
• verify app audience and semantic-model permissions
• test the reports before considering the recovery complete
  
  

AI-assisted drafting: AI was used to help structure and phrase this response. I reviewed and validated the technical content before posting.