Google Big Query connection credentials

[Anonymous]

Hi all,

We have a workspace with two reports. Each of these reports should connect to the same Google Big Query account, but with different credentials (service accounts). This is because each of these service accounts have row level security defined for them within Google Big Query. Each of these service accounts has access to different subsets of the data and each of the individual reports should access the database via its unique service account to display different data.

The issue we're facing is that Power BI service seems to store Big Query credentials on tenant/account level. So; we change credentials in one of the report settings, this gets saved for the other report as well. This is bad as we don't want viewers from report A viewing data in report B.

Solutions tried:

- Clear all permissions on desktop app, upload again

- Upload each report to unique workspace

This did not work. It's clearly visible in impact analysis that connection is made to the same Google Big Query instance where this should be unique for each report/dataset (see screenshot below).

Please let me know if there's a solution to this, or maybe an alternative approach that solves the issue.

[v-yalanwu-msft]

Hi, @Anonymous ;

where are you implementing the RLS? In bigquery or power bi report?

https://cloud.google.com/bigquery/docs/row-level-security-intro

https://docs.microsoft.com/en-us/power-bi/enterprise/service-admin-rls

Best Regards,
Community Support Team _ Yalan Wu
If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[Anonymous]

Hi Yalan,

The RLS is implemented in BQ. The RLS is also not the issue, this is working properly. The issue is the way Power BI service handles credentials for Big Query. This seems to be on 'tenant' or 'account' level. So changing the data source credentials for one dataset also changes it for another. We are now solving it by creating multiple pro accounts for different reports, but I find it strange and inconsistent to other data source behavior that we can't specify details on a dataset level.