Power BI Service - Dataset refresh failing - [Unable to combine data]

[Anonymous]

Hi all,

Hope you will have an idea regarding my issue.

Description

The report is combining 2 datasources :

- entities from Dataflows (from the same workspace)

- Multiple Queries on a Google Big Query datalake (one project_ID)

The results of some queries using dataflows and other using Google Big Query are appended as weel as merged in other cases.

Report refreshes on desktop with no issue.

Published on Power BI Service, replacing a report with the same name (with only dataflows as sources in the replaced report), the refresh is failing systematically.

Error details

[Unable to combine data] Section1/XXXXXXXXX/result references other queries or steps, so it may not directly access a data source. Please rebuild this data combination.. The exception was raised by the IDbCommand interface. XXXXXXXX

Clues

I tested many options.

Mainly :

- modifying my queries. Isolating collection steps from transformation steps (don't know if still relevant)

- changing privacy levels of data source connection IDs (aligned on Organizational as for the Gateway used for the dataflows. Then I tested None). Same error.

I am assuming the problem is coming from the appending or merging of tables coming from the two different datasources (the table specified in the end of the error message is one of the results of this "combinations").

Question

Do you have any idea on the reason that could explain the refresh fail please ?

I am still reading posts in the Community and on Chrome. I don't have any idea.

May republishing the same report under a new name on the Service be helpful ? 

Thank you to all for reading me and your ideas.

Regards,

[otravers · Accepted solution]

These privacy issues can be aggravating. I'd do the data source merge in a new dataflow rather than in the dataset.

[Anonymous]

Thank you @otravers for your answer,

Is this issue a known issue from the product team and accepted as it is right now waiting for a solution later. 

I'ld like to investigate your option however it would be the last possible one in production as I am told that I am not supposed to duplicate data from one datalake (Google Big Query elected as so) into an other one (Dataflows entities being considered as so. And not the dataset..... I know).

If there is no other solution I'll take it.

Regards,

[otravers]

It's known and working as intended. The alternative is to change the structure of your queries until the Power BI service is happy with it. This should help you understand what you have to solve:

https://www.excelguru.ca/blog/2015/03/11/power-query-errors-please-rebuild-this-data-combination/

[Anonymous]

Thank you otravers,

I know this article. I have read it this week to investigate my solutions that is why I wrote :

"modifying my queries. Isolating collection steps from transformation steps (don't know if still relevant)".

By the way, this article was written in 2015 regarding Power Query in Excel. Things may have changed.

For these two reasons I dont think the alternative you are suggesting will help me more than that.

Thank you again.

Regards, 

[otravers]

Why are you assuming this is no longer relevant? PQ is 99% the same between products where it's implemented, and the privacy framework between sources has not fundamentally changed in past years.

[Anonymous]

Thank you @otravers for the information.I did not know.

However I am not assuming anything regarding this option as I applied it and the result doesn't seem to solve the issue.

I look forward to reading other suggestions if you have.

Regards,