Just noticed that Helm [deployment template](https://gitlab.com/gitlab-org/charts/gitlab-runner/-/blob/main/templates/deployment.yaml#L80-95) using the following commands for probes: * Liveness: `command: ["/bin/bash", "/configmaps/check-live", "{{ default 4 .Values.probeTimeoutSeconds | add -1 | max 1 }}"]` * Readiness: `command: ["/usr/bin/pgrep","gitlab.*runner"]` FWIW, you should provide HTTP probes here instead for more reliable healthz handling. Imagine process is running somehow but app doesn't expose any endpoint / port. So `/healthz` endpoint should return `200 OK`. And also the following above commands requires having `bash` and `pgrep` binaries in the image. You could just provide: `/livez` and `/readyz` endpoints in the gitlab-runner. Related doc: https://gitlab.com/gitlab-org/gitlab-runner/-/blob/main/docs/install/kubernetes-agent.md?plain=1#L214 See Kubernetes doc: https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/ <!-- template sourced from https://gitlab.com/gitlab-org/gitlab-runner/-/blob/main/.gitlab/issue_templates/Default.md -->