| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
fix: exit non-zero when server run fails A non-nil return from httpx.Run is a genuine serve failure (bind conflict, TLS cert error) or shutdown timeout — it filters http.ErrServerClosed and returns nil on clean signal shutdown. Logging it at Error and returning left the process exiting 0, so k8s exit-code alerting, crashloop-backoff, and CI status checks saw success on a webhook that failed to serve. Use Fatalf to match the setup/keypair/ clientset fatal paths above. Bump cocoon-common to v0.2.5 (KUBECONFIG list handling; no logic impact on webhook).
fix(deploy): fail closed on the bare-pod gate pods.mutate exists to deny bare cocoon pods, so failurePolicy Ignore defeated it whenever the webhook, its CA, or TLS was unavailable. Now Fail, with a namespaceSelector excluding system namespaces so a webhook outage cannot wedge cluster-critical pod creation.
fix(metrics): unify fail-open decode as skipped; audit follow-ups Mutate's fail-open decode recorded result=error while the four identical fail-open validator decodes record skipped — dashboards on error would fire for one and miss the others. Rule now encoded in the const block: fail-open decode = skipped, fail-closed = error. parent-fetch error gains a reason (contract: "" only for allow/deny). Comment fixes: contract wording, Reloader/GetCertificate godocs (drop package-doc duplicate and the false lock-free claim), test helper restatement.
| Back | FazBrowse Home | New Git URL |