Power BI Service – BigQuery connector failing with ReadChunkWithRetries (Desktop works)

[nathansmith1]

Hi,

Since August 30th 2025, we’ve been seeing failures in Power BI Service when using the Google BigQuery connector.

Error shown:

Cannot execute <ReadChunkWithRetries>b__0 after 5 tries.

The exception was raised by the IDbCommand interface.

Key points

• Power BI Desktop works fine (Import and DirectQuery).
  
  
• In Service, both Import refreshes and DirectQuery report queries fail.
  
  
• Tested against:
  
  
• A raw BigQuery table
  
  
• A view with explicit schema
  
  
• A snapshot table with only DATE, STRING, INT64, FLOAT64 columns
  
  
• Tried UseStorageApi=true/false and both Legacy + Preview connectors.
  
  
• All BigQuery jobs complete successfully on the Google side.
  
  

Related issue

This appears identical to the bug logged in Apache Arrow ADBC:
https://github.com/apache/arrow-adbc/issues/3322

It looks like the Service connector (which now uses the ADBC driver) is failing during result streaming, even though queries succeed in BigQuery.

Question

Is this a known regression in the Power BI Service BigQuery connector, and if so, is a fix or rollback being planned?

Thanks!

[powerbi_user191]

I finally found some official documentation on this issue:

---

A recent fix was published for this issue and should be deployed in October 2025

---

The link has other relevant information regarding the issue:
https://learn.microsoft.com/en-us/power-query/connectors/google-bigquery#unable-to-refresh-partition...

[Ckg_PBI]

Hello i have experienced a similar error with the default selection of the location of the data in BQ , all my data is in EU and PowerBI now looks for it in US ,i haven´t found a way to change this behaviour .

By the moment i have appied the solution of powerbi_user191

In the query code, look for queries having the `[Implementation="2.0"]` parameter:

`= Value.NativeQuery(GoogleBigQuery.Database([Implementation="2.0"])`

and replace them with

`= Value.NativeQuery(GoogleBigQuery.Database()`

This is the error I have in PowerQuery (  ADBC framework seems to have some busg to solve already)  

DataSource.Error: ADBC: Cannot execute <ExecuteQueryInternalAsync>b__1 after 5 tries. Last exception: The service bigquery has thrown an exception. No HttpStatusCode was specified. Job analytics-ppc-ga4/US/job_605e8141_ff66_4e72_b771_a5cc6feef7b3 contained 2 error(s). First error message: Not found: Dataset analytics-ppc-ga4:reporting_page_path_level_funnel was not found in location US Detalles:     DataSourceKind=GoogleBigQuery     DataSourcePath=GoogleBigQuery     AdbcStatus=1     SQLState=     NativeError=0

[powerbi_user191]

Hello  Ckg_PBI!

In your FROM statement, are you specifying all three parts, like below?
project.dataset.table

If not, specify all three parts and see if it works. Project I believe is necessary to point BigQuery to the correct processing region.

Otherwise, the only thing I can think of right now is that
`DataSource.Error: ADBC: Cannot execute <ExecuteQueryInternalAsync>b__1 after 5 tries.`
error could be a permission issue: see
https://learn.microsoft.com/en-us/power-query/connectors/google-bigquery#executequeryinternalasync-f...

[Mauricio_]

Hi 

I was facing the exactly same issue, and solved by removing the parameter from the source connector

[Implementation="2.0"]

After removing this part, the model refresh worked normally.

The final connector information should look like this

= GoogleBigQuery.Database()

Discovered that this was being added since I had the preview feature active to use the new BigQuery connector.

[Alex_Dafnis]

Thank you so much!!

[powerbi_user191]

A potentially helpful note:
When updating an M / PowerQuery query's SQL code, I noticed `Implementation 2.0` is automatically inserted as a parameter.
After updating any BigQuery SQL, it's worthwhile to check if `Implementation 2.0` was inserted again: I didn't notice that at first and had some Service refreshes fail as a result.

[mikerenderco]

Replacing the Power Query code - Removing Implementation 2.0 worked for us as well.  But Does anyone have a link to a microsoft issues that we can follow to make sure the issue gets resolved?

[powerbi_user191]

I wasn't able to create a ticket in PowerBI because it apparently required a paid plan.
Nor have I found another way to reach support: does anyone know how to properly submit a ticket?

[nathansmith1]

Even though reverting to the older implementation or using the Simba ODBC driver with an explicitly set Billing Project ID does work as a temporary fix, this shouldn’t be the long-term solution. The [Implementation="2.0"] path and ADBC driver are clearly part of Power BI’s forward-looking architecture, and it's concerning that Service refreshes are failing while Desktop continues to work.

We really need Microsoft to investigate and resolve the regression in the new connector implementation. Stability in the Power BI Service is critical, especially for enterprise workflows relying on scheduled refreshes and DirectQuery.

Hoping this gets prioritised soon — thanks to everyone who’s shared workarounds!

[v-sshirivolu]

Hi @nathansmith1 ,

Thank you for providing the detailed background. Based on your observations, this appears to be a regression related to the new ADBC-based BigQuery connector within the Service. Since the issue does not occur in Desktop, it suggests the problem is specific to the Service environment.

Here are a few steps you can consider while Microsoft investigates:

Permissions – Please verify that the Power BI Service identity (service principal or user account) has the following BigQuery roles assigned:

roles/bigquery.user

roles/bigquery.jobUser (including bigquery.jobs.create)

roles/bigquery.dataViewer

Re-authenticate in Service – Navigate to Data source settings - Edit permissions and re-sign in using OAuth2 to address any potential token issues.

Support Ticket – Since the issue is isolated to Service, consider opening a Microsoft support ticket via your Power BI Admin Portal. Please include error details and reference the ADBC regression for escalation.

This approach should help maintain your workflow while awaiting a permanent fix from Microsoft.

[v-sshirivolu]

Hi @nathansmith1 ,

Could you please confirm if the issue has been resolved after raising a support case? If a solution has been found, it would be greatly appreciated if you could share your insights with the community. This would be helpful for other members who may encounter similar issues.

Thank you for your understanding and assistance.

[v-sshirivolu]

Hi @nathansmith1 ,
I wanted to check if you had the opportunity to review the information provided. Please feel free to contact us if you have any further questions.

[powerbi_user191]

I found a workaround that resolved the issue for me.

In the query code, look for queries having the `[Implementation="2.0"]` parameter:

`= Value.NativeQuery(GoogleBigQuery.Database([Implementation="2.0"])`

and replace them with

`= Value.NativeQuery(GoogleBigQuery.Database()`

by deleting `[Implementation="2.0"]`

It seems that `[Implementation="2.0"]` is the newer version that has recently stopped working in Service refreshes, and the older implementation is still working fine.

[AlexTG]

Your workaround worked perfectly for me. Thank you! 👏😊

[jfranklin]

This worked for me as well.

[AlexMerkkur]

This helped me as well. Thank you!!

[LuigiBros]

This workaround worked for me as well

[juniorBI]

I had the same issue as the OP and this actually solved my problem. Thanks a lot!

[powerbi_user191]

I've also been experiencing this issue, with this error message:

Data source errorCannot execute <ReadChunkWithRetries>b__0 after 5 tries. The exception was raised by the IDbCommand interface. Table: test_table.

Everything works fine on Desktop, as was mentioned.

I have no ability to use a gateway, nor can I submit a ticket to official support.
I hope to raise more awareness to this issue: Microsoft, please fix this!

[v-sshirivolu]

Hi @nathansmith1 ,
Thanks for reaching out to Microsoft Fabric Community Forum .
Try these Workarounds

Based on your testing with raw tables, views, and snapshots using DATE, STRING, INT64, FLOAT64, and different UseStorageApi settings, here are some solutions to address the ADBC driver issue:

Use ODBC with Simba Driver:

Install the Simba Google BigQuery ODBC driver from Google’s SDK.

Set up an ODBC connection in Power BI Service via a data gateway, making sure to include the Billing Project ID in the connection string to prevent routing issues.

This method bypasses the ADBC driver and is considered stable for both Import and DirectQuery modes. Test with a small dataset to ensure it works for your scenario.

Export Data to Intermediate Storage:

For Import mode, export BigQuery data to CSV or Parquet in Google Cloud Storage or Azure Blob Storage. Then, connect Power BI Service to these files using a suitable connector. This method avoids the streaming issue but is less ideal for DirectQuery.

Since your snapshot table testing was successful, this approach should work well for static data refreshes.

Check Authentication and Permissions:

Make sure the service account for Power BI Service has the bigquery.jobs.create and bigquery.tables.getData roles in BigQuery. Incorrect permissions can also cause streaming errors.

Re-authenticate the BigQuery connection in Power BI Service to eliminate token-related issues.

Escalate and Monitor:

Submit a support ticket through the Power BI admin portal, referencing the ADBC driver bug and related GitHub issue. Keep an eye on the Microsoft Fabric Community for updates from Microsoft support.