helm: add a health check and liveness probe

Author: maelvls

deploy/charts/venafi-kubernetes-agent/templates/deployment.yaml

@@ -94,6 +94,18 @@ spec:
             - containerPort: 8081
               name: http-metrics
           {{- end }}
+          livenessProbe:
+            httpGet:
+              path: /healthz
+              port: 8081
+            initialDelaySeconds: 15
+            periodSeconds: 20
+          readinessProbe:
+            httpGet:
+              path: /readyz
+              port: 8081
+            initialDelaySeconds: 5
+            periodSeconds: 10
       {{- with .Values.nodeSelector }}
       nodeSelector:
         {{- toYaml . | nindent 8 }}

pkg/agent/run.go

@@ -80,6 +80,18 @@ func Run(cmd *cobra.Command, args []string) {
 		go func() {
 			prometheus.MustRegister(metricPayloadSize)
 			metricsServer := http.NewServeMux()
+
+			// Health check endpoint. Since we haven't figured a good way of knowning
+			// what "ready" means for the agent, we just return 200 OK inconditionally.
+			// The goal is to satisfy some Kubernetes distributions, like OpenShift,
+			// that require a liveness and health probe to be present for each pod.
+			metricsServer.HandleFunc("/healthz", func(w http.ResponseWriter, r *http.Request) {
+				w.WriteHeader(http.StatusOK)
+			})
+			metricsServer.HandleFunc("/readyz", func(w http.ResponseWriter, r *http.Request) {
+				w.WriteHeader(http.StatusOK)
+			})
+
 			metricsServer.Handle("/metrics", promhttp.Handler())
 			err := http.ListenAndServe(":8081", metricsServer)
 			if err != nil && !errors.Is(err, http.ErrServerClosed) {