Premium P1 Capacity - Dataflows - Corrupt Model

[NorfolkDumpling]

Hi,
We've had some issues recently were some dataflow refreshes have failed and the history download file seems to suggest there is

corruption within the dataflow model.

Error: The requested model storage file 'https://wabineuopcdsa.blob.core.windows.net<ccon>/931e1a2b-3388-4224-98a4-db46c9678c3e/model.json?snapshot=2024-04-11T10:23:31.1806537Z</ccon>' cannot be found. This might indicate a corrupt model. Please rebuild the model and try again.. Param1 = https://wabineuopcdsa.blob.core.windows.net<ccon>/931e1a2b-3388-4224-98a4-db46c9678c3e/model.json?snapshot=2024-04-11T10:23:31.1806537Z</ccon> Request ID: 64e7c95e-f0be-7a08-b477-2cd775cc465a.

Not sure how to takeup the suggestion of "Please rebuild the model and try again" but we do know that if we export the JSON and create a new dataflow from it then it will recreate the Dataflow successfully, however, this will have a new DataflowID, which will break the Deployment pipelines and any Semantic models that have a dependency to the original DataflowID in their M-Code. 

Futhermore, the dataflow cannot be edited as it brings up an error, "There was a problem opening your dataflow" and it is also not possible to for another person to "Take Over" the dataflow (it has no effect).

We're just using the storage that's built in to our P1 Premuim Capacity and creating our dataflows from within the service (app.powerbi.com)

We've raised a case with Microsoft Technical support but its taking a while and we can't use our deployment pipepine at the moment. Wondered if anybody else has had something similar happen where they're using Dataflows as their main data source for reporting ? 

[v-yohua-msft]

Hi, @NorfolkDumpling 

The error message says that the model.json file cannot be found, which may indicate that the model is corrupted. As mentioned earlier, exporting JSON and creating a new dataflow from it can successfully recreate the dataflow.

However, this results in a new DataflowID, which can cause issues with deployment pipelines and semantic models that rely on the original DataflowID in their M code. You can check the following link:

Power BI Dataflows for Creation of Reusable Data Transformations (mssqltips.com)

There is a way to change the source of a dataflow in Power BI Desktop under Power Query Editor. You can change the source to the new dataflow details, or select a different dataflow option from the list. This may help to switch dependencies to the new DataflowID. You can check link:

Solved: How to Change report built from one dataflow to an... - Microsoft Fabric Community

If you're not the owner of the dataflow, many settings are disabled. To take ownership of your dataflow, you can select Take over to take control.You'll be prompted to provide your credentials to ensure you have the necessary level of access. You can check the following link:

Configure and consume a dataflow - Power BI | Microsoft Learn

How to Get Your Question Answered Quickly 

Best Regards

Yongkang Hua

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[NorfolkDumpling]

Thanks for your reply and information in the links. 

We're still working through the problem with Microsoft support trying to make sure that the DataflowIDs don't change on our Production dataflows as we have a lot of reports (semantic models) that depend on one particular dataflow that has many entities inside it. 

One thing I have learnt for new reports, that it is probably worth editing the M-Code to parameterise the Workflows and Dataflow IDs to make it easier to change. For example to switch between Test and Production areas.

Our "corrupted" dataflow is within the Test workspace inside a pipeline. We have tried to re-deploy the dataflow from Development to Test to no avail.

Our next thought is to detach the Test workspace from the pipeline, "rebuild" the broken dataflow by deleting and recreating it (from JSON). Then hoping the Test workspace will  successfully attach back into the pipeline. 

Would be good to know if anyone has faced a similar scenario before ?