Forum Discussion
Dataflow Gen 1 keeps failing
Hi all
I have a Dataflow Gen 1 connected to an on-prem SQL Server DB through an On-Prem Gateway.
All looks well.
When I enter edit mode, I can see the data and refresh the preview.
The column quality details do not show any errors (Yes, I set it to check the entire dataset).
But it keeps failing.
The error message contains no useful information:
Requested on 23.10.2025 09:39
Dataflow name xxxxx TEST
Dataflow refresh status Failed
Table name AccountStructure
Partition name NA
Refresh status Failed
Start time 23.10.2025 09:39
End time 23.10.2025 09:40
Duration 00:00:30.9200
Rows processed NA
Bytes processed (KB) NA
Max commit (KB) NA
Processor Time NA
Wait time NA
Compute engine NA
Error Error: Request ID: ff465918-5525-7d3b-b431-6dc937b0cdea Activity ID: 981d442b-29dd-4c08-910a-99e559ab88fe
I created this one from scratch based on the same data.
I have the same error.
I loaded the same data into Power BI Desktop, published it, and then configured the same Gateway; the Refresh works.
One interesting detail from the Refresh history:
As you can see, the refreshes until yesterday have taken 7 - 8 minutes.
This morning, it failed after 1 minute and 31 seconds.
The refresh is now stopped after 00:31.
Any Idea?
I've found this one: https://community.fabric.microsoft.com/t5/Service/Dataflow-Gen-1-not-refreshing/m-p/4391001
It described the same issue, but no solution.
Kind regards
Salvatore
Hi all,
We have updated the Gateway, but it has had no effect.
We still have the same error (with no detailed error message).
After checking the lineage of the DataFlow, we noticed that its data is used only by one report.
Therefore, we moved the Load to the Power BI Report (Semantic Model), and we will remove the DataFlow in due course.
I would have preferred to solve this issue and post the solution, but this is the most pragmatic solution for us.
Kind regards
Salvatore
10 Replies
If query folding is happening, can you see anything strange in the SQL query?
Do all the columns have data types assigned? (no ABC 123 columns)
If query folding is not happening, can you make changes so that it does?
I'm just thinking of the troubleshooting steps I would start with.
Yes, all columns have the datatypes as defined in the database.
The source is an on-prem SQL Server database.
Interestingly, disabling the loading of the table listed in the error message causes the following table (in alphabetical order) to appear in the error message.
For this reason, I don't think it is a data issue, but rather a more fundamental one.
Is this available?
If you get the SQL, can you run it directly against the database?
Actually, if you add an index column to the last step of the table (guaranteed to break query folding), do you get a different result?
Yes, Query folding is available.
There are no further transformations during the import of the tables.
Adding an index column, as suggested, didn't change the outcome.
Seems like a global issue, I've been experiencing the same with my dataflows since yesterday.
Thankyou, KNP, and Automagical for your responses.
Hi SalvaC,
We appreciate your enquiry submitted via the Microsoft Fabric Community Forum.Based on my understanding, the failure may be caused by an intermittent connectivity or handshake issue between the Power BI service and the on-premises data gateway during dataflow execution. Possible reasons include an outdated gateway version, a transient service incident, gateway resource exhaustion, a timeout while establishing the connection, or a backend issue affecting Dataflow Gen1 refreshes.
Please follow the steps below, which may help resolve the issue:- Open the on-premises data gateway application and confirm its version and status. Ensure you are running the latest gateway build and that it is online.
- Re enter the data source credentials by navigating to Manage Gateways → select the SQL Server source → Edit credentials → Test connection.
- Create a new dataflow using a single small table through the same gateway. If this also fails, the issue is likely related to the gateway or the service.
Additionally, please refer to the link provided below:
Troubleshoot the on-premises data gateway | Microsoft LearnWe hope the information above helps to resolve the issue. If you have any further queries, please feel free to contact the Microsoft Fabric community.
Thank you.
Thank you for your advice.
I already did the first two things, without success.
I need to talk with the system admins for the update, as this is a client's system and it's a qualified environment.
This will take some time.
I will come back here as soon as I have an update.
Kind regards
Salvatore
Hi SalvaC,
Thank you for the update. Please keep us informed of any further developments. If you have any additional queries, please feel free to contact the Microsoft Fabric community.
Thank you.
Hi all,
We have updated the Gateway, but it has had no effect.
We still have the same error (with no detailed error message).
After checking the lineage of the DataFlow, we noticed that its data is used only by one report.
Therefore, we moved the Load to the Power BI Report (Semantic Model), and we will remove the DataFlow in due course.
I would have preferred to solve this issue and post the solution, but this is the most pragmatic solution for us.
Kind regards
Salvatore
Hi SalvaC,
Thank you for the update.
From your message, it appears you have implemented a practical workaround by moving the data refresh logic from Dataflow Gen1 to the Power BI report’s semantic model. This is an acceptable approach when Dataflow Gen1 refresh failures persist, because the Power BI service employs a different refresh engine for datasets that often avoids the transient gateway handshake issues affecting Gen1 dataflows.
If you wish to continue investigating the original issue at a later time, we recommend migrating your Gen1 dataflow to Dataflow Gen2, as the newer version offers improved reliability, performance and diagnostic capabilities compared with Gen1.
Should you have any further queries, please feel free to contact the Microsoft Fabric community.
Thank you.
