Is it possible to run automatic performance regression tests based on load and performance monitoring?



  • Baseline:

    In our team, we want to emulate certain high loads in production over and over again as a kind of load and performance regression, so to speak. Within one test stage.

    We use Datadog for monitoring and can therefore detect the high load and where the corresponding peaks occur.

    The setup we want to build:

    • Datadog to detect high load and peaks.
    • Jmeter for the emulation of high load and performance.
    • A CI integration of Jmeter and Jenkins to have continuity here.

    Questions:

    • Is it possible to use automated monitoring results to use them via CI directly automated in Jmeter for example? And thus create an emulation?
    • How often should you do performance regression tests? What is your opinion here?
    • Are there other solutions for a setup in which you can emulate a correspondingly high load in Jmeter or another tool via monitoring?


    • Is it possible to use automated monitoring results to use them via CI directly automated in Jmeter for example? And thus create an emulation?

      If you want to create a JMeter test plan using information from DataDog take a look at https://jmeter.apache.org/usermanual/component_reference.html#Access_Log_Sampler Sampler. It's designed to read https://tomcat.apache.org/tomcat-8.0-doc/config/valve.html#Access_Logging but it's possible to come up with your own implementation which will be reading the data from DataDog and generate relevant https://jmeter.apache.org/usermanual/component_reference.html#HTTP_Request samplers. Take a look at the following materials:

      • https://www.blazemeter.com/blog/stop-making-assumptions-learn-how-replay-your-production-traffic-jmeter
      • https://jmeter.apache.org/usermanual/jmeter_accesslog_sampler_step_by_step.html#access-log-sampler
    • How often should you do performance regression tests? What is your opinion here?

      normally people run regression tests when there is something which might cause performance degradation, i.e. adding a new feature or fixing a bug so I would recommend running the tests after the major commit to the https://www.atlassian.com/git/tutorials/comparing-workflows/gitflow-workflow

    • Are there other solutions for a setup in which you can emulate a correspondingly high load in Jmeter or another tool via monitoring?

      the are dozens of them, you need to be more specific when it comes to technologies and requirements




Suggested Topics

  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2