#ci/cd

fr so i was stuck on this one part of our pipeline where the deployment was failing due to some perms issue.. and i tried everything rn changed the env vars, updated the docker image, even tried switching to a different deploy method lol but turns out it was just a simple misconfig in the yaml file.. like wut anyone else have those moments where u spend hrs on something and its just some stupid typo??

added an after_failure hook in the yaml and it just works now lol no more manual resets rn the pipeline now reverts the deploy if any test flakes tbh it saved me a ton of headache

Spent the last couple evenings tweaking our CI flow and managed to make the build automatically retry when a flaky test kills the job. Added a tiny wrapper script that checks the exit code, sleeps a bit, then re‑runs the step up to three times. The best part? No more manual reruns from the dashboard – the pipeline just recovers and pushes the artifact forward. It saved us at least an hour a day during the sprint and cut the noise in the Slack alerts. If anyone's dealing with flaky tests, give the retry‑wrapper a shot – it’s a few lines and works with any runner.

so ive been dealin with this ci/cd pipeline at work rn and its been drivin me crazy... basically the automated tests keep failin due to some dependency issues or somethin? tbh i havent had the time to properly look into it yet but its def got me frustrated lol anyone else have to deal with stuff like this??

i’ve been wrestling with argo cd for a while. when i push an app that spawns a pod with an initContainer, the Argo app’s “Sync” window shows success, the actual pod finishes, but the health status flashes RED/unknown for the rest of the rollout timeout (eg., 300 sec). i’m only using the built‑in health checks that rely on a kube API “PodReady” watch.

no clear docs mention initContainers effect? a bunch of people seem to argue they shouldn't matter if the main container reaches ready; another voice says kube doesn’t signal readyness till the final container exits.

1. if I add a “wait” init container that sleeps for a bit – the rollout seems to work if I tweak the timeout down.
2. what exactly does the health monitor observe during the whole timeout if any init container is present?
3. should I use a separate readinessProbe per init or tell Argo to ignore init Containers for health?

anybody already solved this or got a way to hack the settings? i'm missing the documentation and see varying patterns.

gl on who can answer and keep the fire burning.