Add health and readiness probes to controller manager

[Okabe-Junya]

What

Add health and readiness probes to the controller manager

• Set HealthProbeBindAddress (:8081) on the controller-runtime manager
• Register /healthz (liveness) and /readyz (readiness) checks via healthz.Ping
• Add --health-probe-bind-address CLI flag to allow customization
• Add livenessProbe and readinessProbe to all Helm chart deployment templates and config/manager/manager.yaml

Why

Controller manager currently has no health or readiness probes configured. This means:

• Kubelet cannot detect a hung or deadlocked controller, so the pod is never restarted automatically
• During rolling updates, traffic may be routed to pods that are not yet ready
• There is no mechanism to remove a shutting-down controller from service endpoints

[Copilot]

Pull request overview

This PR adds health and readiness probe support for the controller manager, wiring controller-runtime’s health endpoints into the manager binary and enabling Kubernetes probe configuration in the shipped manifests/charts.

Changes:

• Add --health-probe-bind-address flag and configure controller-runtime HealthProbeBindAddress on the manager.
• Register /healthz and /readyz endpoints via controller-runtime health checks.
• Add livenessProbe and readinessProbe to the kustomize manager manifest and Helm controller deployments.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
main.go	Adds controller-runtime health/ready endpoints and CLI flag to configure probe bind address.
config/manager/manager.yaml	Adds Kubernetes liveness/readiness probes to the manager container.
charts/gha-runner-scale-set-controller/templates/deployment.yaml	Adds liveness/readiness probes to the scale-set controller deployment.
charts/gha-runner-scale-set-controller-experimental/templates/_controller_template.tpl	Adds liveness/readiness probes to the experimental controller container template.
charts/actions-runner-controller/templates/deployment.yaml	Adds liveness/readiness probes to the actions-runner-controller deployment.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

These probes hard-code port 8081, but the manager now supports --health-probe-bind-address and this chart allows users to append arbitrary extraArgs. If a user sets a different bind address/port via extraArgs, the probes will start failing and repeatedly restart the pod. Suggest adding a chart value for the health probe bind address/port, wiring it into both the container args and the probe port field (or at least deriving the probe port from the configured bind address).

Suggested change

	livenessProbe:
	httpGet:
	path: /healthz
	port: 8081
	initialDelaySeconds: 15
	periodSeconds: 20
	readinessProbe:
	httpGet:
	path: /readyz
	port: 8081
	{{- $healthProbeBindAddress := ":8081" -}}
	{{- with .Values.controller.manager.extraArgs }}
	{{- if kindIs "slice" . }}
	{{- range . }}
	{{- if and (kindIs "string" .) (hasPrefix "--health-probe-bind-address=" .) }}
	{{- $healthProbeBindAddress = trimPrefix "--health-probe-bind-address=" . -}}
	{{- end }}
	{{- end }}
	{{- end }}
	{{- end }}
	{{- $healthProbePort := 8081 -}}
	{{- $healthProbeBindParts := splitList ":" $healthProbeBindAddress -}}
	{{- if gt (len $healthProbeBindParts) 1 }}
	{{- $healthProbePort = (last $healthProbeBindParts | int) -}}
	{{- end }}
	livenessProbe:
	httpGet:
	path: /healthz
	port: {{ $healthProbePort }}
	initialDelaySeconds: 15
	periodSeconds: 20
	readinessProbe:
	httpGet:
	path: /readyz
	port: {{ $healthProbePort }}

[Okabe-Junya]

Since (technically) we can use extraArgs in experimental chart, this feeedback mighy be reasonable, but I think it would introduce unnecessary complexity and a lack of consistency with other charts.

But, if this change is necessary, I will create a fixup