feat: use webhooks to prevent the pod of notready's kosmosnode from b…

…eing expelled

Signed-off-by: renxiangyu <[email protected]>

Makefile

@@ -18,6 +18,7 @@ TARGETS :=  clusterlink-controller-manager  \
 			clusterlink-proxy \
 			clustertree-cluster-manager \
 			scheduler \
+			webhook
 
 CTL_TARGETS := kosmosctl
 

cmd/webhook/app/options/options.go

@@ -0,0 +1,63 @@
+package options
+
+import (
+	"github.com/spf13/pflag"
+	"k8s.io/client-go/tools/leaderelection/resourcelock"
+	componentbaseconfig "k8s.io/component-base/config"
+
+	"github.com/kosmos.io/kosmos/pkg/utils"
+	"github.com/kosmos.io/kosmos/pkg/webhook"
+)
+
+type Options struct {
+	LeaderElection       componentbaseconfig.LeaderElectionConfiguration
+	KubernetesOptions    KubernetesOptions
+	WebhookServerOptions WebhookServerOptions
+	PodValidatorOptions  webhook.PodValidatorOptions
+}
+
+type KubernetesOptions struct {
+	KubeConfig string  `json:"kubeconfig" yaml:"kubeconfig"`
+	Master     string  `json:"master,omitempty" yaml:"master,omitempty"`
+	QPS        float32 `json:"qps,omitempty" yaml:"qps,omitempty"`
+	Burst      int     `json:"burst,omitempty" yaml:"burst,omitempty"`
+}
+
+type WebhookServerOptions struct {
+	Host     string
+	Port     int
+	CertDir  string
+	CertName string
+	KeyName  string
+}
+
+func NewOptions() *Options {
+	return &Options{
+		LeaderElection: componentbaseconfig.LeaderElectionConfiguration{
+			LeaderElect:       true,
+			ResourceLock:      resourcelock.LeasesResourceLock,
+			ResourceNamespace: utils.DefaultNamespace,
+			ResourceName:      "network-manager",
+		},
+	}
+}
+
+func (o *Options) AddFlags(flags *pflag.FlagSet) {
+	if o == nil {
+		return
+	}
+
+	flags.BoolVar(&o.LeaderElection.LeaderElect, "leader-elect", true, "Start a leader election client and gain leadership before executing the main loop. Enable this when running replicated components for high availability.")
+	flags.StringVar(&o.LeaderElection.ResourceName, "leader-elect-resource-name", "kosmos-webhook", "The name of resource object that is used for locking during leader election.")
+	flags.StringVar(&o.LeaderElection.ResourceNamespace, "leader-elect-resource-namespace", utils.DefaultNamespace, "The namespace of resource object that is used for locking during leader election.")
+	flags.Float32Var(&o.KubernetesOptions.QPS, "kube-qps", 40.0, "QPS to use while talking with kube-apiserver.")
+	flags.IntVar(&o.KubernetesOptions.Burst, "kube-burst", 60, "Burst to use while talking with kube-apiserver.")
+	flags.StringVar(&o.KubernetesOptions.KubeConfig, "kubeconfig", "", "Path for kubernetes kubeconfig file, if left blank, will use in cluster way.")
+	flags.StringVar(&o.KubernetesOptions.Master, "master", "", "Used to generate kubeconfig for downloading, if not specified, will use host in kubeconfig.")
+	flags.StringVar(&o.WebhookServerOptions.Host, "bind-address", "0.0.0.0", "The IP address on which to listen for the --secure-port port.")
+	flags.IntVar(&o.WebhookServerOptions.Port, "secure-port", 9443, "The secure port on which to serve HTTPS.")
+	flags.StringVar(&o.WebhookServerOptions.CertDir, "cert-dir", "/etc/certs", "The directory that contains the server key and certificate.")
+	flags.StringVar(&o.WebhookServerOptions.CertName, "tls-cert-file-name", "tls.crt", "The name of server certificate.")
+	flags.StringVar(&o.WebhookServerOptions.KeyName, "tls-private-key-file-name", "tls.key", "The name of server key.")
+	flags.StringArrayVar(&o.PodValidatorOptions.UsernamesNeedToPrevent, "usernames-need-to-prevent", []string{"system:serviceaccount:kube-system:node-controller"}, "Usernames that need to prevent deleting pods on NotReady kosmos nodes.")
+}

cmd/webhook/app/options/validation.go

@@ -0,0 +1,10 @@
+package options
+
+import "k8s.io/apimachinery/pkg/util/validation/field"
+
+// Validate checks Options and return a slice of found errs.
+func (o *Options) Validate() field.ErrorList {
+	errs := field.ErrorList{}
+
+	return errs
+}

cmd/webhook/app/webhook.go

@@ -0,0 +1,103 @@
+package app
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+
+	"github.com/spf13/cobra"
+	"k8s.io/client-go/tools/clientcmd"
+	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/webhook"
+
+	"github.com/kosmos.io/kosmos/cmd/webhook/app/options"
+	"github.com/kosmos.io/kosmos/pkg/scheme"
+	"github.com/kosmos.io/kosmos/pkg/sharedcli/klogflag"
+	kosmoswebhook "github.com/kosmos.io/kosmos/pkg/webhook"
+)
+
+func NewWebhookCommand(ctx context.Context) *cobra.Command {
+	opts := options.NewOptions()
+
+	cmd := &cobra.Command{
+		Use:  "kosmos-webhook",
+		Long: `TODO`,
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if errs := opts.Validate(); len(errs) != 0 {
+				return errs.ToAggregate()
+			}
+			if err := Run(ctx, opts); err != nil {
+				return err
+			}
+			return nil
+		},
+		Args: func(cmd *cobra.Command, args []string) error {
+			for _, arg := range args {
+				if len(arg) > 0 {
+					return fmt.Errorf("%q does not take any arguments, got %q", cmd.CommandPath(), args)
+				}
+			}
+			return nil
+		},
+	}
+
+	fss := cliflag.NamedFlagSets{}
+
+	genericFlagSet := fss.FlagSet("generic")
+	opts.AddFlags(genericFlagSet)
+
+	logsFlagSet := fss.FlagSet("logs")
+	klogflag.Add(logsFlagSet)
+
+	cmd.Flags().AddFlagSet(genericFlagSet)
+	cmd.Flags().AddFlagSet(logsFlagSet)
+
+	return cmd
+}
+
+func Run(ctx context.Context, opts *options.Options) error {
+	config, err := clientcmd.BuildConfigFromFlags(opts.KubernetesOptions.Master, opts.KubernetesOptions.KubeConfig)
+	if err != nil {
+		panic(err)
+	}
+	config.QPS, config.Burst = opts.KubernetesOptions.QPS, opts.KubernetesOptions.Burst
+
+	mgr, err := controllerruntime.NewManager(config, controllerruntime.Options{
+		Logger: klog.Background(),
+		Scheme: scheme.NewSchema(),
+		WebhookServer: &webhook.Server{
+			Host:     opts.WebhookServerOptions.Host,
+			Port:     opts.WebhookServerOptions.Port,
+			CertDir:  opts.WebhookServerOptions.CertDir,
+			CertName: opts.WebhookServerOptions.CertName,
+			KeyName:  opts.WebhookServerOptions.KeyName,
+		},
+		MetricsBindAddress:      "0",
+		HealthProbeBindAddress:  "0",
+		LeaderElection:          opts.LeaderElection.LeaderElect,
+		LeaderElectionID:        opts.LeaderElection.ResourceName,
+		LeaderElectionNamespace: opts.LeaderElection.ResourceNamespace,
+	})
+	if err != nil {
+		klog.Errorf("failed to build webhook server: %v", err)
+		return err
+	}
+
+	hookServer := mgr.GetWebhookServer()
+	// register an Admission Webhook
+	hookServer.Register("/validate-delete-pod", &webhook.Admission{Handler: &kosmoswebhook.PodValidator{
+		Client:  mgr.GetClient(),
+		Options: opts.PodValidatorOptions,
+	}})
+	hookServer.WebhookMux.Handle("/health", http.StripPrefix("/health", &healthz.Handler{}))
+
+	if err := mgr.Start(ctx); err != nil {
+		klog.Errorf("failed to start webhook manager: %v", err)
+		return err
+	}
+
+	return nil
+}

cmd/webhook/main.go

@@ -0,0 +1,17 @@
+package main
+
+import (
+	"os"
+
+	apiserver "k8s.io/apiserver/pkg/server"
+	"k8s.io/component-base/cli"
+
+	"github.com/kosmos.io/kosmos/cmd/webhook/app"
+)
+
+func main() {
+	ctx := apiserver.SetupSignalContext()
+	cmd := app.NewWebhookCommand(ctx)
+	code := cli.Run(cmd)
+	os.Exit(code)
+}

deploy/webhook/deployment.yaml

@@ -0,0 +1,53 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: kosmos-webhook
+  namespace: kosmos-system
+spec:
+  selector:
+    matchLabels:
+      app: kosmos-webhook
+  template:
+    metadata:
+      labels:
+        app: kosmos-webhook
+    spec:
+      serviceAccountName: kosmos-webhook
+      affinity:
+        nodeAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+            nodeSelectorTerms:
+              - matchExpressions:
+                  - key: kosmos.io/node
+                    operator: DoesNotExist
+        podAntiAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+            - labelSelector:
+                matchExpressions:
+                  - key: app
+                    operator: In
+                    values:
+                      - kosmos-webhook
+              namespaces:
+                - kosmos-system
+              topologyKey: kubernetes.io/hostname
+      containers:
+        - image: ghcr.io/kosmos-io/webhook:__VERSION__
+          name: kosmos-webhook
+          volumeMounts:
+            - name: tls
+              mountPath: "/etc/certs"
+          command:
+            - webhook
+            - --v=4
+          resources:
+            limits:
+              memory: 500Mi
+              cpu: 500m
+            requests:
+              cpu: 500m
+              memory: 500Mi
+      volumes:
+        - name: tls
+          secret:
+            secretName: kosmos-webhook-tls

deploy/webhook/rbac.yaml

@@ -0,0 +1,33 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: kosmos-webhook
+  namespace: kosmos-system
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: kosmos-webhook
+rules:
+  - apiGroups: ['*']
+    resources: ["nodes", "pods"]
+    verbs: ["*"]
+  - apiGroups: [ "coordination.k8s.io" ]
+    resources: [ "leases" ]
+    verbs: [ "get", "list", "watch", "create", "update", "patch", "delete" ]
+  - apiGroups: [""]
+    resources: ["events"]
+    verbs: ["create"]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: kosmos-webhook
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: kosmos-webhook
+subjects:
+  - kind: ServiceAccount
+    name: kosmos-webhook
+    namespace: kosmos-system

deploy/webhook/service.yaml

@@ -0,0 +1,12 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: kosmos-webhook
+  namespace: kosmos-system
+spec:
+  ports:
+    - port: 9443
+      protocol: TCP
+      targetPort: 9443
+  selector:
+    app: kosmos-webhook

deploy/webhook/tls-secret.yaml

@@ -0,0 +1,9 @@
+apiVersion: v1
+data:
+  tls.crt: __BASE64_SERVER_CRT__
+  tls.key: __BASE64_SERVER_KEY__
+kind: Secret
+metadata:
+  name: kosmos-webhook-tls
+  namespace: kosmos-system
+type: kubernetes.io/tls

deploy/webhook/validate-delete-pod-config.yaml

@@ -0,0 +1,26 @@
+apiVersion: admissionregistration.k8s.io/v1
+kind: ValidatingWebhookConfiguration
+metadata:
+  name: "validate-delete-pod.kosmos.io"
+webhooks:
+  - name: "validate-delete-pod.kosmos.io"
+    rules:
+      - apiGroups: [""]
+        apiVersions: ["v1"]
+        operations: ["DELETE"]
+        resources: ["pods"]
+        scope: "*"
+    admissionReviewVersions: ["v1"]
+    # FailurePolicy defines how unrecognized errors from the admission endpoint are handled - allowed values are
+    # Ignore or Fail. Defaults to Fail.
+    failurePolicy: Ignore
+    sideEffects: None
+    timeoutSeconds: 3
+    clientConfig:
+      service:
+        namespace: kosmos-system
+        name: kosmos-webhook
+        path: /validate-delete-pod
+        port: 9443
+      caBundle: |
+        __BASE64_CA_CRT__

hack/gen-webhook-certs.sh

@@ -0,0 +1,42 @@
+#!/bin/bash
+
+WEBHOOK_NAME="kosmos-webhook"
+NAMESPACE="kosmos-system"
+# 有效期
+DAYS="36500"
+
+openssl genrsa -out ca.key 2048
+
+# generate CA private keys and self-signed certificates:
+openssl req -new -x509 -days ${DAYS} -key ca.key \
+  -subj "/C=CN/CN=${WEBHOOK_NAME}"\
+  -out ca.crt
+
+# fenerating Mutating Webhook Server private key and certificate signing request (csr)
+openssl req -newkey rsa:2048 -nodes -keyout server.key \
+  -subj "/C=CN/CN=${WEBHOOK_NAME}" \
+  -out server.csr
+
+printf "subjectAltName=DNS:${WEBHOOK_NAME}.${NAMESPACE}.svc" > extfile.txt
+
+# sign the mutating webhook server certificate using the CA
+openssl x509 -req \
+  -extfile extfile.txt \
+  -days ${DAYS} \
+  -in server.csr \
+  -CA ca.crt -CAkey ca.key -CAcreateserial \
+  -out server.crt
+
+echo
+echo ">> Generating kube secrets..."
+kubectl create secret tls ${WEBHOOK_NAME}-tls \
+  --cert=server.crt \
+  --key=server.key \
+  --dry-run=client -o yaml \
+  > tls-secret.yaml
+
+echo
+echo ">> MutatingWebhookConfiguration caBundle:"
+cat ca.crt | base64 | fold
+
+rm ca.crt ca.key ca.srl server.crt server.csr server.key extfile.txt

hack/util.sh

@@ -23,6 +23,7 @@ CLUSTERLINK_TARGET_SOURCE=(
   clusterlink-controller-manager=cmd/clusterlink/controller-manager
   clustertree-cluster-manager=cmd/clustertree/cluster-manager
   kosmosctl=cmd/kosmosctl
+  webhook=cmd/webhook
 )
 
 #https://textkool.com/en/ascii-art-generator?hl=default&vl=default&font=DOS%20Rebel&text=KOSMOS