We are looking for a Selenium replacement and have started using Cypress. And we faced with the fact that the tests are extremely unstable on Azure. We use the default agents on Ubuntu with 2 cores and 7 GB of RAM. Is this enough? What advice can you give to keep the tests stable?

I haven't worked with Azure, but want to get the conversation started.. What are the reason(s) for the unstability you are seeing in your tests?

That seems like a pretty weak VM. You definitely wont be able to do any parallel runs with that.

What would you suggest ?

Min specs are here https://docs.cypress.io/guides/getting-started/installing-cypress Hard to say if that's what your issue is since you don't have any error logs in here, but the specs your on are pretty much the minimum.

@stocky-salesclerk-42224 Can you elaborate on what you mean by instability? Are your tests hanging? Flaky? Are you getting an error of some kind? As @echoing-tent-95037 mentioned, specifically with CI, the requirements are: CPU - 2 CPUs minimum to run Cypress - 1 additional CPU if video recording is enabled - 1 additional CPU per process you run outside of Cypress, such as: --- App server (frontend) --- App server (backend) --- App database --- Any additional infrastructure (Redis, Kafka, etc..) Memory - 4GB minimum, 8GB+ for longer test runs

@adventurous-dream-20049 The tests are flaky. Cypress does not completely clear the fields. Tests with a detach error regularly fall, tests are executing very slowly

bump the machine specs

Always happy to hear from you, @gray-kilobyte-89541. I truly miss working with you. I'd like to share your feedback, but want to make sure I fully understand what is meant by "bump the machine specs". Would you mind elaborating on what ideally you'd like to see and how not having this is affecting your testing/debugging experience?

unfortunately it depends on the web app (is it being served locally? is it remote?) and how heavy it is. If I notice video skipping or hanging, that's a sure sign to go with a more powerful machine. On Circle we just bump the machine class and observe

Gotcha, gotcha. I completely misunderstood 😅 That is actually one of the reasons I enjoy using CircleCI, to increase the available RAM. But GitHub Actions is so much easier to get started with. Iirc, there was a beta released a while back to increase the GHA available specs.