Obtaining profiler traces related to performance issues

  • Updated
Description

Performance issues frequently require profiler traces to diagnose a root cause. In DxP environments you can make use of Microsoft's Application Insights (AI) for this purpose. To avoid lost time from irrelevant traces you can use AI's role and time filtering to obtain a useful trace. 

Notes:

  • With Windows app services (.Net Framework CMS 11 DxP clients) profiler must be enabled via a new support request.
  • If Application Insights is not accessible, this can be requested by a new support request.
Steps

 

  1. Open the Application Insights (AI) resource. 


    For convenience you can make this resource a default. 



    Now AI will show in the top left menu. 
  2. Navigate to the Investigate> Performance resource.


    AI_Perf.png
  3. If an issue is app service or instance specific, click the roles filter and select accordingly. 
    role_selector.png
  4. Zoom in on a significant peak. For instance in the above graph you may select from 5 - 7AM. Besides operation time you may also be interested in CPU or memory. In that case set the Request count dropdown accordingly. 
    metric_types.png
  5. Use the distribution selector to filter on longer running requests.
    distribution.png
  6. You can also select a particular operation name to retrieve profiler traces for a particular request. 
  7. Click on profiler traces and examine and/or download any desired traces. 
    profiler_traces.png

 

Traces can be reviewed using PerfView a free tool from Microsoft or opened in Visual Studio. 

https://github.com/microsoft/perfview