Refreshes from SQL over the Gateway taking longer and longer

[motoray]

We have a Power BI Premium Capacity (yep, still thanks to our EA) and I've noticed that over the past year, data refreshes for Power BI reports published to the service are taking an average of 200 seconds longer than they did at the same time last year. I've done some work to try and spread the load out so we don't have so many refreshes happening concurrently, but this still seems like a really big change over the course of one year. Most of our refreshes hit our on-prem SQL datamart, so that means the refreshes use our two-node Power BI gateway setup. The biggest change in refresh times seemed to happen around from September through November 2024, when the average time zoomed up to around 190 seconds. Has anyone else seen anything like this? Has anyone found a solution? Anyone have thoughts as to what might be going on?

[Poojara_D12]

Hi @motoray 

You're observing a significant increase in data refresh durations for Power BI reports published to your Premium Capacity environment—on average, about 200 seconds longer now compared to the same time last year. Despite efforts to balance the refresh load and avoid concurrency bottlenecks, performance hasn't improved much. Most of your reports pull data from an on-prem SQL datamart via a two-node Power BI gateway cluster, which adds a layer of complexity to troubleshooting. The most notable increase in refresh times occurred between September and November 2024, indicating a possible correlation with either a change in the Power BI service, gateway updates, or infrastructure changes in your data source environment. This kind of performance degradation could be due to several factors, including changes in Power BI’s internal refresh engine behavior (especially with Premium Gen2 optimization rollouts), increased dataset complexity, SQL server performance issues, gateway load contention, or even subtle changes like Windows patches or network latency that affect the gateway-to-database connection. It’s also possible that refresh performance telemetry or capacity metrics during peak times could reveal memory or CPU constraints on your Premium node or gateway machines. To investigate further, you might analyze metrics from the Power BI Premium Metrics App, Gateway Performance Monitoring logs, and refresh logs in the Service to identify bottlenecks. Solutions others have used include scaling gateway hardware, updating gateway software, isolating high-demand refreshes to separate time windows, or migrating the gateway closer to the database for lower latency. This is a common concern for Premium users with growing usage and complex on-prem data pipelines, and pinpointing the root cause usually involves coordinated monitoring across Power BI, the gateway, and the SQL datamart.

[motoray]

There's a lot to look at here. We're going to start with adding another node to our gateway cluster.

[ToddChitt]

For cases like this, you have to ask yourself, "What changed?" Some thoughts to get you started:

• Amount of data?
• Power Query steps such that it no longer 'folds'? (Meaning it has to do some steps at the client level, not the SQL Server)
• SQL Indexing strategy?
• Network bandwidth?

Can you isolate the performance issue to one table in a model? What happens if you extract the SQL statement from Power Query and run that against the server directly?

[motoray]

Appreciate the thoughts. This is across all reports in the tenant and we have over 300. They're all taking much longer to refresh than they did this time last year--from the ones doing the heavy lifting with lots of data, to the ones doing a few dozen rows. The change happened about the same time as we started getting lots of failure for refreshes going across the gateway around July 2024. We then reverted to the May 2024 gateway and didn't update again until October...and that's when we started getting longer and longer refresh times. 

One thing to note: we did add the server names to the config file.

https://powerbi.microsoft.com/en-us/blog/on-premises-data-gateway-september-2024-release/

[v-menakakota]

Hi @motoray ,

You saw some relief after rolling back to May 2024, so version differences matter. Try testing with the latest gateway version on a non-production node there may be performance improvements or fixes.

 Even small delays between the gateway and your SQL Server can add up across 300+ reports.It’s worth checking for increased network latency or DNS resolution issues

If I misunderstand your needs or you still have problems on it, please feel free to let us know.   

Best Regards, 
Menaka.
Community Support Team  

[motoray]

We've just installed the May 21, 2025 version recently, but I don't expect that to fix anything. Due to other issues with failed refreshes and open tickets with Microsoft, we've been on the bleeding edge of Gateways since October and refresh times continue to get longer, not shorter.

[v-menakakota]

Hi @motoray 

Thanks for the update I completely understand how frustrating this must be, especially after staying up to date with the latest gateway versions and still seeing performance get worse. Since the issue seems persistent across all reports and gateway versions, it may be worth looking into other potential bottlenecks.

• Reviewing gateway diagnostics logs can help identify whether delays are happening during authentication, query execution, or data transfer.
• Running the same queries directly against SQL Server, outside of Power BI, may reveal issues like blocking, slow execution plans, or other performance constraints.
• Monitoring for network latency or DNS resolution issues between the gateway and SQL Server could help, as even small delays can add up significantly across many refreshes.
  
  If this post was helpful, please give us Kudos and consider marking Accept as solution to assist other members in finding it more easily.
  Thank you for reaching out to us on the Microsoft Fabric Community Forum.
   
  Best Regards, 
  Menaka. 
  Community Support Team.
   

[v-menakakota]

Hi  @motoray ,

May I ask if you have resolved this issue? If so, please mark the helpful reply and accept it as the solution. This will be helpful for other community members who have similar problems to solve it faster. 

Thank you. 

[motoray]

We're currently testing, but still having issues.  If we find a solution, I'll come back here and let everyone know. And if someone's suggestion solves the problem, I'll likewise come back and mark that suggestion as the answer. As of right now the problem is still there.

[motoray]

Haven't really made any progress on this still. Testing continues. Long refresh times persist.

[v-menakakota]

Hi @motoray ,

Please go through the below documentation link:
Monitor and optimize on-premises data gateway performance | Microsoft Learn

Best Regards, 
Menaka.
Community Support Team

[v-menakakota]

Hi  @motoray ,

May I ask if you have resolved this issue? If so, please mark the helpful reply and accept it as the solution. This will be helpful for other community members who have similar problems to solve it faster. 

Thank you

[v-menakakota]

Hi  @motoray ,

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. If my response has addressed your query, please accept it as a solution so that other community members can find it easily. 

 
Thank you. 

[motoray]

I've been out of the office for a while--back today and catching up...will report back.

[v-menakakota]

Hi @motoray ,
 

May I ask if you have resolved this issue? If so, please mark the helpful reply and accept it as the solution. This will be helpful for other community members who have similar problems to solve it faster. 

Thank you. 

[motoray]

We have not solved the issue. I went over the gateway event logs and although I can see the trend of the refreshes taking longer there, I was unable to find any one thing that was an overall cause. It seems to be random in some ways--we some bigger models that are fairly consistent, and then we have smaller models that some days seem to take forever to refresh. I don't know that there is an answer to this. It's just sad that it's gotten so much worse--and continues to get worse.

[v-menakakota]

Hi @motoray ,

If you are still unable to solve the issue you can reach out to Microsoft Support by raising a ticket with Microsoft Support. 

Please refer below link on how to raise a contact support or support ticket. 
How to create a Fabric and Power BI Support ticket - Power BI | Microsoft Learn 

Best Regards, 
Menaka. 
Community Support Team

[v-menakakota]

Hi @motoray ,

We are following up once again regarding your query. Could you please confirm if the issue has been resolved through the support ticket with Microsoft?

If the issue has been resolved, we kindly request you to share the resolution or key insights here to help others in the community. If we don’t hear back, we’ll go ahead and close this thread.

Should you need further assistance in the future, we encourage you to reach out via the Microsoft Fabric Community Forum and create a new thread. We’ll be happy to help.

Thank you for your understanding and participation.

[motoray]

Still not resolved. We're in the process now of adding another gateway server node to our cluster to see if that helps. We've begun getting refresh failures several times a week now where the refreshes don't complete due to the them taking too long. This is new behavior.

I've been working to refactor longer running models, but it feels like a losing battle. For every few seconds I scrape back on an individual model, the overall trend continues to go up. I finally found a good visual to show it and there's a drastic jump in November 2024--and that's when we fully moved off the June 2024 version of the gateway. Something changed with the gateways then and has only gotten worse since that change.

[v-menakakota]

Hi @motoray ,

Did you get a chance to raise a support ticket? The support team can investigate the issue further on the backend. If you’ve already submitted one, please wait for their response.

Thank you.