parquet status exception

[bubbledep]

Hi everyone,

I got the error "parquet status exception" when I tried to create a new semantic model using data from a Data Warehouse which I've ingested with data from a Gen2 Dataflow merged table.

Previously I had success using that same logic to create a new semantic model using data from a Data Warehouse which I've ingested with data from a Gen2 Dataflow data, but instead of a merged table I was using four tables (same data, just not merged yet).

Today I also noticed that the Fabric changed his layout, in particular I noticed that they added navigation bars on top of the screen to change between artifacts from the workspace, so maybe they are changing something?

Any help would be appreciated.

[Nabha-Ahmed]

Hi @bubbledep 

The “Parquet status exception” usually appears when Fabric tries to load a Warehouse table whose underlying delta/parquet files are in an inconsistent state.
This often happens when the table was created or overwritten by a Gen2 Dataflow merge step.

✔ Why it happens

Merged Gen2 Dataflow outputs can temporarily leave:

• Corrupted parquet metadata
• Incomplete transaction logs
• Delta files with mismatched schema

Semantic models are more sensitive to these issues than Warehouses, so they fail when reading the table.

✔ Recommended steps to fix it

1. Refresh / re-run the Dataflow to force Fabric to rewrite the delta files cleanly.
2. In the Warehouse, run:

   VACUUM <table_name> RETAIN 0 HOURS;

   (this cleans orphan parquet files)
3. If you are using merge, try switching to:
   • Append-only
   • Or writing to a staging table, then doing a Warehouse SQL MERGE into the final table.
4. Try creating the semantic model from the original 4 tables again; if that works, then the issue is isolated to the merged table’s parquet state.

✔ About the new UI layout

Yes — Fabric rolled out a new top navigation bar today.
UI changes often accompany backend updates, and some users are seeing temporary instability in Warehouse–Semantic Model integration.
So it is likely related.

---

If this helps, please mark the answer as Accepted, and receive KUDO.

[v-sdhruv]

Hi @bubbledep ,

Hi hope the information provided above helped you with the query if you are facing any issues try to manually sync schema changes between Gen2 Dataflow and the semantic model:

Use the edit table option in Power BI Direct Lake to manually include new fields.

There is no automatic sync between Dataflow Gen2 and semantic models anymore.

If issue still persists, try raising a support ticket- 
How to create a Fabric and Power BI Support ticket - Power BI | Microsoft Learn

Hope this helps!

[BhaveshPatel]

Hi @bubbledep 

There is diffeence between parquet file ( Data Lake ) and parquet file with transaction-log ( _delta_log ) + Parquet file. I think so you were using Power BI Dataflow Gen 1 which is based on data lake, and this is the reason you see exception error in parquet file. Move your data to dataflow gen 2, and you see delta lake ( _delta_log + parquet file ) and you will not see any error. 

Thanks

Bhavesh

[tayloramy]

Hi @bubbledep, 

I have run into "ParquetStatusException" a few times in Fabric, especially right after writing tables from Dataflow Gen2. In most cases it has been stale or inconsistent metadata on the default semantic model, or a schema quirk introduced by a merged output. Here are the things that usually clear it up for me:

1. Quick reset of the default semantic model
   Open your Warehouse or Lakehouse, go to Reporting > Manage default semantic model, remove all tables, Confirm, then add them back and try again. Several folks have reported this resolves ParquetStatusException without any other changes: community thread.
2. Recheck the merged output’s schema
   Merges can quietly create duplicate or odd column names, mixed data types, or unsupported types that write fine but fail when read by Direct Lake. In your Dataflow Gen2 step that produces the merged table:
   • Make every column’s type explicit (no Any). Prefer text, whole number, fixed decimal, datetime, boolean.
   • Ensure every column name is unique and simple (no dots, slashes, or leading/trailing spaces).
   • Avoid very wide decimals or binary columns.
   If in doubt, write the four source tables again and do the merge in SQL as a Warehouse table or view to isolate the issue.
3. Force a clean table rewrite
   If the delta/parquet files got into a bad state during the previous write, create a brand-new output table name from the dataflow and point the semantic model at that fresh table. This often bypasses lingering file references that trigger the exception. Similar 404 BlobNotFound and parquet status cases have been reported and resolved with a fresh write: example report, another thread.
4. Keep Direct Lake expectations in mind
   Direct Lake reads physical tables from OneLake. Model tables sourced from views fall back to DirectQuery. If you are mixing modes, a problematic table can surface only when it is read via Direct Lake. More context here: Direct Lake notes and an overview of how Direct Lake reads delta/parquet under the hood: overview.
5. If it persists, capture details for support
   Grab the error time, workspace name, item ids, table name, and correlation id, and include the links above showing this as a known class of issue. That will help Support triage faster.

On the UI change you noticed: yes, the navigation has been updated recently, but the ParquetStatusException itself usually tracks back to metadata or file state, not the UI.

If you found this helpful, consider giving some Kudos. If I answered your question or solved your problem, mark this post as the solution.

[bubbledep]

Hi @tayloramy ,

Unfortunately, we aren't able anymore to access the step 1, since July'25 Update on Sunsetting Default Semantic Models.

[tayloramy]

Hi @bubbledep, 

Ah shoot, I knew I was forgetting something. 
Try refreshing the table in your non default semantic model or creating a new semantic model and see if it will refresh the metadata properly. 

You can also try using the REST API to refresh the endpoint metadata: 
Refresh SQL analytics endpoint Metadata REST API (Preview) | Microsoft Fabric Blog | Microsoft Fabri...

If you found this helpful, consider giving some Kudos. If I answered your question or solved your problem, mark this post as the solution.

[bubbledep]

Hi @tayloramy ,

The problem is that I cannot creat a new semantic model with that new merged table. Even creating a new Warehouse and re-merging the tables doesn't worked, unfortunately.

[tayloramy]

Hi @bubbledep, 

I see, and refreshing the SQL endpoint via the API still doesn't allow the table to show up? 

In this case, this seems to be a bug. I'm out of ideas, can you open a ticket to MS to see if they can provide any support? 

In the meantime, we can wait and see if a CST will reply to this thread with any other ideas. 

If you found this helpful, consider giving some Kudos. If I answered your question or solved your problem, mark this post as the solution.