Our automation scripts are developed using Cucumber and Selenium (4.6 version). These scripts are run first in multi-threaded mode (4 threads) in the Jenkins pipeline (EC2 instance, Amazon Linux2 OS) with headful chrome. Then the failed scripts are rerun (as UI tests can be flaky) within the same pipeline but in a single thread.
A few months back, we started upgrading Chrome to 108 and suddenly we noticed all our pipelines were taking longer duration. In the multi-threaded run, we noticed some scripts failing (due to CSS not displayed within time, not visible (or not loading) until some wait time), and also passed scripts taking longer duration compared to Chrome 80 run. However, the failed scripts are running fine in the rerun which is single-threaded. So, we concluded that the issue is happening only in multi-threaded run. We also tried running using Chrome 107, 109, 111, 113 and 114), but we see the same issue.
We ran our test suite on Windows and Mac in multi-threaded mode and observed that the tests are running fine without taking longer duration. It is almost the same as the Chrome 80 run on AL2 Linux. Then we concluded that this issue is happening on AL2 Linux and on the latest chrome versions (started somewhere from > 80 and < 108 version) in multi-threaded mode only. Not exactly sure if the latest Chrome has performance issues on AL2 Linux OR the latest Linux Chrome driver has performance issues OR issues with the combination of Selenium, the latest Chrome driver, and the latest Chrome on AL2 Linux in Multi-Threaded mode.
Did anyone observe this or a similar issue and be able to resolve it? Can someone please suggest how to resolve OR overcome this issue? (We have a very huge number of test cases and so many pipelines to run the scripts. So, running tests in a single thread is not an effective solution for us.)
Top comments (0)