Report Refresh Failures

Your refresh failures are likely due to P1 capacity strain (too many concurrent refreshes, heavy models, or schedule collisions).

Fixes:

Stagger refresh times

Use incremental refresh

Optimize models (smaller, fewer columns)

Monitor with Premium Metrics App

Reduce refresh concurrency

Scale up or enable autoscale if needed

Hi mjoseph111 

Many possible reasons:

• Complex queries
• You're making Power Query all or almost all transformations instead of doing it upstream
• High cardinality columns such as datetime. Split them into date and time columns. If you don't need the seconds, remove the as well.
• Merge - fine with small data but can contribute to slow refresh with large data
• Wide tables - do you need all the columns?
• Expensive DAX queries - you have a dedicated capacity which is shared across data refreshes and rendering of visuals

Or maybe you'll need a higher capacity because it is currently being stretched to its limits with the number of users and reports.

Hi mjoseph111 ,

Thanks vivien57  Shahid12523  and danextian  for the valuable inputs here.

As per my knowledge , I just want to add that the refresh failures you’re seeing.
(DM_GWPipeline_Gateway_DataSourceAccessError) are generally tied to capacity strain on P1 in the Gov cloud. The suggestions already shared (staggering refresh times, using incremental refresh, optimizing Power Query/M models, and monitoring via Premium Capacity Metrics App) are the recommended approaches to reduce these errors.

mjoseph111  did you already try staggering your refresh schedule and/or enabling incremental refresh on the larger models? That usually provides the most immediate relief in environments where scaling up (e.g. moving from P1 → higher capacity) isn’t an option.

If these don’t help, then unfortunately there isn’t a separate “fix” specific to sovereign cloud beyond optimization and capacity management, since the service operates with those constraints.

Regards,
Akhil.

Hi mjoseph111 ,

Thanks for the update and great to hear you were able to trace this back to the service interruption on Sept 3rd.

Also big thanks to super users for sharing valuable troubleshooting guidance here.

Yes, for future reference, checking the [Microsoft Fabric service status page] is the right step whenever you suspect a cloud-side issue. That way you can quickly confirm whether it’s a platform disruption vs. a tenant/model-specific refresh issue. Marking this one as resolved.

Regards,
Akhil.

Hello mjoseph111 

Can you send us some screenshots of the error messages?

By the way, if you are on P1, you will need to consider switching to F64.

https://powerbi.microsoft.com/en-us/blog/grace-period-for-transitioning-from-power-bi-premium-to-microsoft-fabric/

Feel free to give me a kudo if my answer helps you.

Vivien 

We are seeing the same issue in our environment.

The refresh fails sporadically with DM_GWPipeline_Gateway_DataSourceAccessError. There is no clear pattern. Sometimes the error happens in one semantic model, other times in a different one. Our gateway is currently on version 3000.298.8.

If we run the refresh again, it usually succeeds. Sometimes we can go several days without any error at all. Because of that, this does not look like a permanent dataset or query issue, but rather an intermittent gateway, network, or service-side problem.

We would also like to know whether anyone has identified the real root cause, since the issue is random and difficult to reproduce consistently.