Merge pull request kosmos-io#747 from OrangeBao/main

feat: add probe

Author: duanmengkk

cmd/kubenest/node-agent/app/serve/serve.go

@@ -69,6 +69,11 @@ func Start(addr, certFile, keyFile, user, password string) error {
 	passwordHash := sha256.Sum256([]byte(password))
 
 	http.HandleFunc("/", func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path == "/healthz" || r.URL.Path == "/readyz" {
+			w.WriteHeader(http.StatusOK)
+			return
+		}
+
 		auth := r.Header.Get("Authorization")
 		if auth == "" {
 			http.Error(w, "Unauthorized", http.StatusUnauthorized)

cmd/kubenest/operator/app/operator.go

@@ -15,6 +15,7 @@ import (
 	cliflag "k8s.io/component-base/cli/flag"
 	"k8s.io/klog/v2"
 	controllerruntime "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/healthz"
 	"sigs.k8s.io/controller-runtime/pkg/manager"
 
 	"github.com/kosmos.io/kosmos/cmd/kubenest/operator/app/config"
@@ -223,11 +224,24 @@ func run(ctx context.Context, config *config.Config) error {
 		LeaderElection:          config.LeaderElection.LeaderElect,
 		LeaderElectionID:        config.LeaderElection.ResourceName,
 		LeaderElectionNamespace: config.LeaderElection.ResourceNamespace,
+		LivenessEndpointName:    "/healthz",
+		ReadinessEndpointName:   "/readyz",
+		HealthProbeBindAddress:  ":8081",
 	})
 	if err != nil {
 		return fmt.Errorf("failed to build controller manager: %v", err)
 	}
 
+	err = mgr.AddHealthzCheck("healthz", healthz.Ping)
+	if err != nil {
+		return fmt.Errorf("failed to build healthz: %v", err)
+	}
+
+	err = mgr.AddReadyzCheck("readyz", healthz.Ping)
+	if err != nil {
+		return fmt.Errorf("failed to build readyz: %v", err)
+	}
+
 	hostKubeClient, err := kubernetes.NewForConfig(config.RestConfig)
 	if err != nil {
 		return fmt.Errorf("could not create clientset: %v", err)

deploy/virtual-cluster-operator.yml

@@ -103,11 +103,37 @@ spec:
                       values:
                         - virtual-cluster-operator
                 topologyKey: kubernetes.io/hostname
+      initContainers:
+      - name: init-container
+        image: ghcr.io/kosmos-io/virtual-cluster-operator:__VERSION__
+        command:
+          - sh
+          - "-c"
+          - |
+            mkdir -p /kosmos && \
+            tar -xzf /config/manifest-use.tar -C /kosmos
+        volumeMounts:
+          - name: manifest-volume
+            mountPath: /kosmos 
+          - name: manifest-tar
+            mountPath: /config 
       containers:
       - name: virtual-cluster-operator
         # Change a valid image address
         image: ghcr.io/kosmos-io/virtual-cluster-operator:__VERSION__
         imagePullPolicy: IfNotPresent
+        livenessProbe:
+          httpGet:
+            path: /healthz
+            port: 8081
+          initialDelaySeconds: 15
+          periodSeconds: 10
+        readinessProbe:
+          httpGet:
+            path: /readyz
+            port: 8081
+          initialDelaySeconds: 5
+          periodSeconds: 10
         env:
           - name: IMAGE_REPOSITIRY
             value: ghcr.io/kosmos-io
@@ -150,8 +176,8 @@ spec:
         - name: shellscript
           mountPath: /etc/vc-node-dir/config.yaml
           subPath: config.yaml
-        - mountPath: /kosmos/manifest
-          name: components-manifest
+        - mountPath: /kosmos
+          name: manifest-volume
         command:
         - virtual-cluster-operator
         - --kubeconfig=/etc/virtual-cluster-operator/kubeconfig
@@ -163,10 +189,11 @@ spec:
       - name: shellscript
         configMap:
           name: virtual-cluster-operator
-      - hostPath:
-          path: /home/vc-operator/manifest
-          type: DirectoryOrCreate
-        name: components-manifest
+      - name: manifest-tar
+        configMap:
+          name: manifest-tar
+      - name: manifest-volume
+        emptyDir: {} 
 ---
 apiVersion: apps/v1
 kind: DaemonSet
@@ -216,6 +243,20 @@ spec:
       containers:
         - name: install-agent
           image: cis-hub-huabei-3.cmecloud.cn/node-agent/node-agent:latest
+          livenessProbe:
+            httpGet:
+              path: /healthz
+              port: 5678
+              scheme: HTTPS
+            initialDelaySeconds: 15
+            periodSeconds: 10
+          readinessProbe:
+            httpGet:
+              path: /readyz
+              port: 5678
+              scheme: HTTPS
+            initialDelaySeconds: 5
+            periodSeconds: 10          
           securityContext:
             privileged: true # container privileged
           command: