DirectLake Semantic Model failes to load due to supposed Disk Size Limit

[rvp_ordix]

We have a Semantic Model in our PowerBI-Service that connects to our Lakehouse in the same workspace via DirectLake. When I tried to open the data model today, though, the following error message is displayed:

Reports that connect to the model also fail to load, though the SQL Endpoint the model is building on works fine. The blacked out Database ID is the ID of the model itself, not the underlying lakehouse.

The claim that this model would "exceed the maximum size limit on disk" is very confusing, since according to the MS Learn Direct-Lake-page even an F2 capacity has a Size On Disk Limit of 10GB for DirectLake models. We use an F16, which means that our limit should be 20GB. 

Max Memory is the only limit we may plausibly be reaching, but even. Neither Size On Disk nor Max Memory will block you from working with a DirectLake Model if exceeded. According to the documentation they'd simply impact performance.

Usually I'd have tried to remove tables from the model to get it back under that "limit" and figure the cause out afterwards, but I can't even do that with the data model inaccessible. 

What is happening, and how can I fix this?

[Akash_Varuna]

Hi @rvp_ordix  , I think it is like a resource issue rather than exceeding size limits Please do try out these steps

• Check for Service Issues: Look for any ongoing issues on Power BI or Azure status pages.
• Clear Cache/Force Refresh: Try clearing the cache or refreshing the model to resolve metadata issues.
• Reduce Query Load: Temporarily disable heavy queries to ease the load.
• Check Connections: Ensure the connection between the DirectLake model and lakehouse is stable.
  If this post helped please do give a kudos and accept this as a solution 
  Thanks In Advance

[rvp_ordix · Accepted solution]

I've "solved" it now. Scaling the Capacity of that workspace up to F64 temporarily allowed me to access the data model again, and I've removed a few of the tables that were large but not very important. Afterwards it worked on F16 again. 

That being said, I still don't really understand, what exactly the problem was. To your points:

• There was no Service Issue
• I have attempted a regular refresh, though that failed. How can I manually clear a model's cache? There is no option for that in the settings.
• The problem was independent of any report, and there is no transformation in a DirectLake Model, so there were no queries to be deactivated.
• How do I check the connection? Both model and Lakehouse were in the same Fabric Workspace. I can't really affect anything there.