Hi, we need to perform load tests on Pega. The load test intends to simulate the full activity of various CSR operators respect the main use cases identified in the project, for this reason the perimeter of the stress test is the Pega frontend UI.
Test is based on a replay of the HTTP logs of the requests traced by the Browser (HAR file) during a normal user session (including login, to manage session cookies). The technology for replaying HTTP requests with the desired load is Gatling.
We have successfully create a loading ramp with the replay of HTTP requests, however we have found that the test is not repeatable: after an arbitrary time some resources no longer respond with the expected outcome but test receive a verbose and difficult response. Currently all the fields provided by the module parameters are not parameterized and therefore the same values are always resubmitted at each test execution. From empirical tests we have found that some form parameters change between one interaction and another, and one of these is the token 'pzuiactionzzz'. In particular, we found that by modifying the form request with an updated 'pzuiactionzzz' token, the response provides the desired positive outcome.
We would therefore like to understand what are the precautions to be adopted to perform a load test on the frontends, first of all to manage the 'pzuiactionzzz' token (or perhaps not to manage it, in order to facilitate the execution of the loading ramps) but also to anticipate any aspects not addressed so far with the tests.
Once you have recorder all your scripts and downloaded as HAR can you try to modify the file and replace all the values against the pzuiactionzzz token to "blank" (do not remove the token, just remove the recorded value) . Once you do that, save the updated HAR, and use that file as the source recording file for Gatling.