Merge pull request #28 from jelmersnoeck/add-healthz-check

Add health endpoint.

docs/kube/with-rbac.yaml

@@ -145,6 +145,26 @@ spec:
           imagePullPolicy: IfNotPresent
           args:
           - operator
+          livenessProbe:
+            httpGet:
+              path: /_healthz
+              port: 9090
+              scheme: HTTP
+            failureThreshold: 3
+            initialDelaySeconds: 10
+            periodSeconds: 50
+            successThreshold: 1
+            timeoutSeconds: 1
+          readinessProbe:
+            httpGet:
+              path: /_healthz
+              port: 9090
+              scheme: HTTP
+            failureThreshold: 3
+            initialDelaySeconds: 5
+            periodSeconds: 5
+            successThreshold: 1
+            timeoutSeconds: 1
           resources:
             requests:
               cpu: 1m

internal/httpsvc/health.go

@@ -0,0 +1,25 @@
+package httpsvc
+
+import (
+	"fmt"
+	"net/http"
+)
+
+// Health represents a Health server which can be used to expose a health check.
+type Health struct {
+	Server
+}
+
+// Start starts the Health Server.
+func (s *Health) Start(stopCh <-chan struct{}) error {
+	registerHealthCheck(&s.ServeMux)
+
+	return s.Server.Start(stopCh)
+}
+
+func registerHealthCheck(mux *http.ServeMux) {
+	mux.HandleFunc("/_healthz", func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusOK)
+		fmt.Fprintln(w, "OK")
+	})
+}

internal/httpsvc/server.go

@@ -0,0 +1,38 @@
+package httpsvc
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+	"time"
+)
+
+// Server is a wrapper around a http.ServeMux which sets up necessary endpoints
+// to run our operator.
+type Server struct {
+	http.ServeMux
+
+	Addr string
+	Port int
+}
+
+// Start starts the HTTP Server.
+func (s *Server) Start(stopCh <-chan struct{}) error {
+	srv := http.Server{
+		Addr:         fmt.Sprintf("%s:%d", s.Addr, s.Port),
+		Handler:      &s.ServeMux,
+		ReadTimeout:  10 * time.Second,
+		WriteTimeout: 10 * time.Second,
+	}
+
+	go func() {
+		<-stopCh
+
+		ctx := context.Background()
+		ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
+		defer cancel()
+		srv.Shutdown(ctx)
+	}()
+
+	return srv.ListenAndServe()
+}

internal/ingressmonitor/cmd/operator.go

@@ -4,6 +4,7 @@ import (
 	"log"
 	"time"
 
+	"github.com/jelmersnoeck/ingress-monitor/internal/httpsvc"
 	"github.com/jelmersnoeck/ingress-monitor/internal/ingressmonitor"
 	"github.com/jelmersnoeck/ingress-monitor/internal/provider"
 	"github.com/jelmersnoeck/ingress-monitor/internal/provider/logger"
@@ -22,6 +23,9 @@ var operatorFlags struct {
 	MasterURL    string
 	KubeConfig   string
 	ResyncPeriod string
+
+	HTTPAddr string
+	HTTPPort int
 }
 
 // operatorCmd represents the operator command
@@ -54,10 +58,20 @@ func runOperator(cmd *cobra.Command, args []string) {
 		log.Fatalf("Error building IngressMonitor clientset: %s", err)
 	}
 
+	// register the available providers
 	fact := provider.NewFactory(kubeClient)
 	statuscake.Register(fact)
 	logger.Register(fact)
 
+	// register the health server
+	healthSrv := &httpsvc.Health{
+		Server: httpsvc.Server{
+			Addr: operatorFlags.HTTPAddr,
+			Port: operatorFlags.HTTPPort,
+		},
+	}
+	go healthSrv.Start(stopCh)
+
 	op, err := ingressmonitor.NewOperator(
 		kubeClient, imClient, operatorFlags.Namespace,
 		resync, fact,
@@ -78,4 +92,7 @@ func init() {
 	operatorCmd.PersistentFlags().StringVar(&operatorFlags.MasterURL, "master-url", "", "The URL of the master API.")
 	operatorCmd.PersistentFlags().StringVar(&operatorFlags.KubeConfig, "kubeconfig", "", "Kubeconfig which should be used to talk to the API.")
 	operatorCmd.PersistentFlags().StringVar(&operatorFlags.ResyncPeriod, "resync-period", "30s", "Resyncing period to ensure all monitors are up to date.")
+
+	operatorCmd.PersistentFlags().StringVar(&operatorFlags.HTTPAddr, "http-addr", "0.0.0.0", "address the health server will bind to")
+	operatorCmd.PersistentFlags().IntVar(&operatorFlags.HTTPPort, "http-port", 9090, "port on which the health server is available")
 }