Kind of related, but I spent about 5 weeks straight a couple months back working on our Playwright tests at work. Lots and lots of race conditions, many because of poor code since nobody knew the framework well, but many were just uncontrollable server side and network issues. I modified so much code and optimized it to the point that the entire pipeline was running in about half the original time, and I was getting successful consecutive pipeline runs. I found out that the test server is a VM with 1 CPU core and 8gb of RAM. Playwright by default will run all test files in parallel but each individual test in each file in serial, so we had about 10 tests trying to run frontend tests at the same time on that hardware. I think per Playwrights recommendations, it should be something like half as many cores as concurrent tests or similar, so we weren't close.
Anyway all of that text to say I was able to get that increased from 1 to 2 cores, and 8 to 16gb of RAM. It's still getting hammered, and it's still having issues, but it's no longer my problem, so the company gets what they're paying for.
and then the CI fails because the fucking test env collapsed since the billion dollar company won't cough up the cash to fix their CI/Testing infrastructure.
save