TL;DR: Refreshing datasets (on demand and scheduled) in Power BI Service are only partially loading from the source without prompting any errors.
Similar to this post. I have commented there as well.
I have a collection of reports all suffering from the same issue. Here have been my reproduction steps:
- Refresh a local copy via Power BI Desktop. -- this will be 100% data complete.
- Publish to web. -- Data matches the local copy still.
- Refresh the data set (manually or via Scheduled Refresh). -- Data is now only partially loaded with 0 errors occuring according to Refresh History.
From this point, I download the now-incorrect .pbix and seach for common themes in the partially loaded data. I have noticed that some of the queries will be at exactly 1000 or 2000 rows of data (when the underlying SQL tables are definitely larger than this). Is there possible batching or chunking of data that is erroneously thinking it is complete?
Supplemental Information:
- Data source is Azure SQL Server.
- Data is not missing from our SQL DB.
- First noticed occurrence on 2018-07-25.
- I am on Power BI Pro.
- Publishing from Power BI Desktop from the Windows Store. -- I am going to download Power BI Desktop from the website and try this same test.
- I am not opted into Incremental Refresh.
Some of my tests and results:
- Azure SQL Server is too heavily taxed -- I would expect this to prompt a timeout or similar error.
- I ramped up our DTUs to the next tier but results are the same.
- Also, not all reports from the same source are failing.
- I tried publishing under a new workspace and a new name (suggestion from my initial support request).
Next tests:
- Downloading Power BI Desktop from the website and seeing if I achieve different results.
- Trying various Data Load settings under Options.
- Dive into using SQL Profiler on this.
Any further ideas or suggestions?