Embedded Service CPU Issue with Power Apps Cloud Flow Requests

[mikeyoungCO]

We have a Power BI Embedded instance (A1), and the CPU climbs throughout the day as Power Apps Cloud Flows request paginated reports. The only way I can get the CPU to drop is to restart the instance.

If I run the same paginated reports from the Power BI web service or from an API call from a web app, the CPU returns to its baseline immediately after the report is rendered.

The screenshot below shows the pattern:

• In between the instance restarts, the CPU climbs with each request from Power Apps Cloud Flows
• The green spike shows a report request from Power BI web service
• The pink spike shows a report request from a web app

Has anyone else seen this?

[Anonymous]

Hi @mikeyoungCO 

You can monitor API calls and latency using Application Insights, check the following link to learn more:

Overview of integration with Application Insights - Power Platform | Microsoft Learn

Best Regards,
Jarvis Tang
If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[Poojara_D12]

Hi @mikeyoungCO 

Key Points:

1. Issue:

   • Power BI Embedded (A1) CPU usage climbs continuously when paginated reports are triggered via Power Apps Cloud Flows.
   • CPU returns to baseline immediately after running the same reports from Power BI Web Service or a web app.

2. Behavior Observed:

   • Green spike: Report requests from Power BI Web Service quickly return CPU usage to baseline.
   • Pink spike: Report requests from a web app also stabilize CPU.
   • Requests from Power Apps Cloud Flows cause persistent CPU increase until the instance is restarted.

3. Possible Causes:

   • Inefficient handling of paginated reports by Power Apps Cloud Flows.
   • Potentially unoptimized API calls or connections between Power Apps and Power BI Embedded.
   • Resource contention or memory leaks in the Power BI Embedded instance.

4. Recommendations:

   • Investigate how Power Apps Cloud Flows execute API calls (e.g., frequency, caching, or parallel requests).
   • Optimize the Power Apps integration with Power BI by reviewing query and resource usage patterns.
   • Scale up the Power BI Embedded instance temporarily to test whether the issue persists.
   • Report the issue to Microsoft Support if the behavior is reproducible.

Did I answer your question? Mark my post as a solution, this will help others!

If my response(s) assisted you in any way, don't forget to drop me a "Kudos" 🙂

Kind Regards,
Poojara
Data Analyst | MSBI Developer | Power BI Consultant
Please Subscribe my YouTube for Beginners/Advance Concepts: https://youtube.com/@biconcepts?si=04iw9SYI2HN80HKS 

[mikeyoungCO]

What tools would we use to analyze the Power Apps Cloud Flows API calls?

[rajendraongole1]

Hi @mikeyoungCO -This behavior of the Power BI Embedded instance (A1) where the CPU usage climbs and doesn't drop after processing paginated report requests from Power Apps Cloud Flows could indicate a resource management or session handling issue specific to the requests originating from Power Apps.

Analyze the Load Pattern
Observation: Requests from Power Apps Cloud Flows appear to cause CPU usage to climb and not release resources.
Contrast: Requests from the Power BI web service or web apps appear to release resources as expected.
This suggests that requests from Power Apps might be keeping sessions alive or causing a resource leak.

2. Investigate the Cloud Flow Integration
Check how the Power Apps Cloud Flows are making API requests to Power BI:

Paginated Report API Call:

Verify if the flow is making multiple requests unnecessarily or if there's a lack of proper cleanup after the report is rendered.
Ensure the API calls use appropriate filters and are optimized to reduce report complexity.
Session or Connection Handling:

Some workflows or requests might inadvertently leave sessions open, causing resources to accumulate on the instance.
Confirm whether Cloud Flows close sessions properly after execution.
3. Review Report Complexity
The paginated reports may be causing the embedded instance to hold resources for too long:

Heavy Data Processing: Paginated reports with high data volume or complex queries may consume excessive CPU and memory.
Caching Issues: Check if caching is properly utilized. Reports with excessive on-the-fly computations may consume more resources.
Test simplifying the reports (reduce queries, pre-aggregate data).
4. Monitor Instance Performance
Use Azure Monitor or Application Insights to track resource utilization of the Power BI Embedded instance over time.
Specifically, monitor:
CPU usage
Memory utilization
Request patterns
Throttling or timeouts
5. Scale or Optimize the Instance
Scale the Capacity:
If the A1 instance is consistently maxing out, consider scaling to a higher SKU (e.g., A2 or A3) to accommodate the load.
Use the Autoscale feature to handle spikes dynamically.
Optimize Scheduling:
Schedule intensive Cloud Flows during non-peak hours to distribute the load.
6. Optimize Power Apps Cloud Flows
Batch Requests: If the flow sends multiple API requests in quick succession, consider batching them or introducing delays.
Use REST API Efficiently: Ensure that the flow leverages the Power BI REST API correctly, minimizing redundant requests.
Connection Throttling: Implement retry policies to avoid overloading the Embedded instance.
7. Test in Isolation
To confirm whether the issue lies in Power Apps Cloud Flows:

Create an identical flow but run it independently of Power Apps.
Test the API calls for paginated reports using tools like Postman or custom scripts.
Compare CPU behavior with and without the Cloud Flows.
8. Open a Microsoft Support Ticket
If none of the above resolves the issue, the problem may be related to how Power Apps interacts with Power BI Embedded resources. You can:

Provide Microsoft with the performance logs and explain the differing CPU behaviors.
Attach details of the Cloud Flows, paginated report specifications, and Embedded instance configuration.

[mikeyoungCO]

What tools would we use to analyze the Power Apps Cloud Flow API calls?