Been dealing with this more often lately. Tests pass on my machine, I push, and CI blows up. Usually it’s one of these:
- Different Node/Python/whatever version
- Missing env vars that exist in my .env but not in CI secrets
- File system case sensitivity (macOS vs Linux)
- Some flaky test that depends on timing
My current debugging flow is pretty basic: check the logs, compare versions, run the exact same Docker image locally if I can. But it still eats 20-30 minutes each time before I figure out the actual problem.
Anyone have a more systematic approach? Like a quick checklist you run through before you even look at the logs?
Also curious — do you replicate your CI environment locally with something like act (for GitHub Actions) or just trust the remote runner?
In QEMU all of our CI environments are replicable locally as docker/podman images. However usually flaky CI is due to races exposed on overloaded runners so I often also run make -j(nproc) at the same time to simulate that. A retry script is also useful to get an idea of how stable a test is. Having sanitizer builds can also help.