Skip to main content
Showing results for 
Search instead for 
Did you mean: 

Grow your Fabric skills and prepare for the DP-600 certification exam by completing the latest Microsoft Fabric challenge.

Can't remove data source from Dataflow

Given I have a Dataflow with one or more queries to data source A

When I change the queries to use data source B

And I remove data source A from the gateway

And I go to the Gateway connection for Dataflow

And I look at "Data sources included in this dataset"

Then I should see that data source B is there (yes, no problem)

And I should see that data source A is gone

But I see data source A is still there

And it has a green check mark

When I look at the "Status" column for the selected gateway

Then I should see the status is good

But I see "Gateway or datasources not reachable"

When I look at "Data source credentials"

Then I should see a single line saying

ODBC   (admin has granted access, credentials are not required)

But I see two lines, and the first line has an X in a circle

ODBC (x) (admin has granted access, credentials are not required)

When I hover on the circle, I see a message saying the data source may have been removed (which is the case)

When I click on the Settings gear ("Manage Gateways" on tool tip) icon under "Actions" for the gateway, the browser goes into an infinite loop loading two pages, where one of them has this URL:  (I replaced my GUIDs with XXXXX - this is not the actual URL)


My data sources are 64-bit ODBC, but I am doubtful that should matter.

Status: Accepted
Community Support

Hi @tlaresch


Based on my test, the result is a little different from yours. Create a dataflow connects to a ODBC DSN, save dataflow and configure gateway correctly for this dataflow. Then Edit dataflow entities, replace DSN with a new one, save the dataflow. Go to Manage Gateways remove the old ODBC data source, add a new ODBC data source points to the new DSN. When we go to the Dataflow Settings, the Gateway Connections has two data source, but it prompts us to add the previous ODBC in gateway: 



Once we add the previous ODBC data source in gateway, the result is below: 




I know it's a little strange we need to create two ODBC data sources in gateway though dataflow only has one ODBC data source. Before reporting the issue internally, I would like to ensure my test result is what you are concerning. Please run on-premises data gateway version 3000.31.3 (March 2020) as ours then test again to see if you get the same result as ours.  


Best Regards,
Qiuyun Yu

Advocate II


Thanks for researching my issue.

While I don't believe this issue is causing a failure in the processing of the Dataflow, I think it should be considered a bug and fixed.  It cost me a lot of time investigating this issue, and likely will cost many others a lot of time as well.

This issue as reported yesterday was running on the latest gateway version, the same as you.  I believe that I also saw the same issue with an older gateway some weeks back, but I don't have that firmly documented and can't even tell you which gateway version that was.

I want to confirm that the issue you are seeing is essentially the same as mine.  The Dataflow insists on keeping the old data source, even though it is no longer needed, and there is no way to clean that up.

The slight difference you and I see is that my test case starts from the ODBC.  With the starting environment being a dataflow with data source A, I went to the source server and modified data source A in ODBC so that it pointed to data source B and renamed it to "Data source B" as the first step.  From Power BI perspective, this should be the same as deleting ODBC dsn=A and adding dsn=B.  I am sorry I did not initially report this specific detail.  I believe this detail accounts for the minor differences in our results.

Here is a screenshot of my gateway version information. 

Our gateway is running the latest version, March 2020Our gateway is running the latest version, March 2020

Just now I deleted the old dara source A from the gateway, and this is what the dataflow connection looks like now:

Dataflow with old data source 2.PNG

Thanks and regards, and stay safe,


Community Support

Hi @tlaresch


Thank you for your confirmation. I have reported this issue internally: CRI 183834189. Will update here once I get any information. 


Best Regards,
Qiuyun Yu

Community Support
Status changed to: Accepted
Community Support

Hi @tlaresch


I got information that "This issue is by design and we created a low priority bug on our side to handle it in the future". 


Best Regards,
Qiuyun Yu

Not applicable

is this bug resolved, i see the issue in my dataflow, i wont be able to delete and recreate them as my datasets are already using this Data

DF bug.png

Frequent Visitor

This issue is still occurring today.  Is there a workaround whereby we can stop a data flow from looking for a data source that it no longer references?


We renamed an ODBC data source, and now the data flow is looking for both the old name and the new name.


Also, a thing cannot be both a bug and "by design".  This should be called a bug.

Helper II

@tlaresch2 I am experiencing the same. I added an OData source during testing. It is not referenced in the .json for the dataflow any longer, but the credential remains in the Dataflow settings. I have clicked "skip test connection" to bypass errors, but I'd rather not have it included any longer.


Guess it's time to open a support request. Will keep posted if anything develops.

Frequent Visitor

Hello, I am experiencing the same issue here.


I implemented a dataflow and for some reason, a Web data source, which isn't used anywhere in the dataflow's tables, appeared and I need to delete it in order for the dataflow's scheduled refreshes to run properly:



Each time I try to enter credentials for this useless data source, I get an error message, although it is written that adding it to the gateway is "facultative". It is a major bug as far as I am concerned. There should be a way to remove unused or unwanted datasources from the list, especially if they are marked as facultative by the system.


Thanks for reading me.

Frequent Visitor

I have unfortunately had to adapt our deployment processes to account for this.  When we change a data source for a data flow, we export the data flow, delete the existing one, re-import that same data flow (just to get rid of that obsolete data source).  Then, we have to set up the new data flow (credentials, etc.), andadjust anything that has a reference to that data flow to use its new GUID.


So, @DavBI, you have my sympathies, at least.