Until now I did run my pipeline using a dedicated docker server with a shared runner inside. Test cases pass nearly every time, fail only for my very own errors.
Now I tried to switch to a k8s cloud instead (just changing the global variable for tagging the targets, that was easy!) But I have the impression that formerly passing testcases now fail “sometimes” which makes me nervous. I now have the following questions rasing up:
- Is k8s runner more sensible to yet undiscovered timing problems? So I do have possible errors?
- Is it just unstable when using services within a job?
- Would it be a good idea to run tests in parallel on docker and k8s to be sure everything is even more solid?
gitlab is the latest self hosting version 14.10.2
k8s is using k3s setup with version 1.23.6