feat: check CoreDNS

go.mod

@@ -4,6 +4,8 @@ go 1.23
 
 toolchain go1.23.3
 
+replace github.com/longhorn/go-common-libs => github.com/c3y1huang/go-common-libs v0.0.0-20241114031240-e1117130cd3b
+
 require (
 	github.com/longhorn/go-common-libs v0.0.0-20241109101504-614ae31d2212
 	github.com/longhorn/longhorn-manager v1.7.2

go.sum

@@ -6,6 +6,8 @@ github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5 h1:0CwZNZbxp69SHPd
 github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5/go.mod h1:wHh0iHkYZB8zMSxRWpUBQtwG5a7fFgvEO+odwuTv2gs=
 github.com/blang/semver/v4 v4.0.0 h1:1PFHFE6yCCTv8C1TeyNNarDzntLi7wMI5i/pzqYIsAM=
 github.com/blang/semver/v4 v4.0.0/go.mod h1:IbckMUScFkM3pff0VJDNKRiT6TG/YpiHIM2yvyW5YoQ=
+github.com/c3y1huang/go-common-libs v0.0.0-20241114031240-e1117130cd3b h1:mXbX1SYCZnxrIO/s5gfKSqMMbJMHnwlQwNDUKCyBnTM=
+github.com/c3y1huang/go-common-libs v0.0.0-20241114031240-e1117130cd3b/go.mod h1:MXrEgkGm+fI2dHRUuBCiKOG+UHySQXYC8ecjTFBvvho=
 github.com/c9s/goprocinfo v0.0.0-20210130143923-c95fcf8c64a8 h1:SjZ2GvvOononHOpK84APFuMvxqsk3tEIaKH/z4Rpu3g=
 github.com/c9s/goprocinfo v0.0.0-20210130143923-c95fcf8c64a8/go.mod h1:uEyr4WpAH4hio6LFriaPkL938XnrvLpNPmQHBdrmbIE=
 github.com/chai2010/gettext-go v1.0.2 h1:1Lwwip6Q2QGsAdl/ZKPCwTe9fe0CjlUbqj5bFNSjIRk=
@@ -102,8 +104,6 @@ github.com/liggitt/tabwriter v0.0.0-20181228230101-89fcab3d43de h1:9TO3cAIGXtEhn
 github.com/liggitt/tabwriter v0.0.0-20181228230101-89fcab3d43de/go.mod h1:zAbeS9B/r2mtpb6U+EI2rYA5OAXxsYw6wTamcNW+zcE=
 github.com/lithammer/dedent v1.1.0 h1:VNzHMVCBNG1j0fh3OrsFRkVUwStdDArbgBWoPAffktY=
 github.com/lithammer/dedent v1.1.0/go.mod h1:jrXYCQtgg0nJiN+StA2KgR7w6CiQNv9Fd/Z9BP0jIOc=
-github.com/longhorn/go-common-libs v0.0.0-20241109101504-614ae31d2212 h1:wmbGfP3yp4B35Nbd9mNwtOEOih4QrWvOXG3jXkisCgM=
-github.com/longhorn/go-common-libs v0.0.0-20241109101504-614ae31d2212/go.mod h1:MXrEgkGm+fI2dHRUuBCiKOG+UHySQXYC8ecjTFBvvho=
 github.com/longhorn/longhorn-manager v1.7.2 h1:hf0phTtk8Odq+zoVYO5JuHbKJOi4nK+iugUJ1FQw56M=
 github.com/longhorn/longhorn-manager v1.7.2/go.mod h1:QvOuwQP3TqICQHyJcPynjtAhF3TMCFeygVIpf9q7/cI=
 github.com/mailru/easyjson v0.7.7 h1:UGYAvKxe3sBsEDzO8ZeWOSlIQfWFlxbzLZe7hwFURr0=

pkg/consts/preflight.go

@@ -4,6 +4,8 @@ const (
 	AppNamePreflightChecker              = "longhorn-preflight-checker"
 	AppNamePreflightContainerOptimizedOS = "longhorn-gke-cos-node-agent"
 	AppNamePreflightInstaller            = "longhorn-preflight-installer"
+
+	AppNameCoreDNS = "coredns"
 )
 
 type DependencyModuleType int

pkg/local/preflight/checker.go

@@ -53,6 +53,16 @@ type Checker struct {
 func (local *Checker) Init() error {
 	local.collection.Log = &types.LogCollection{}
 
+	config, err := commonkube.GetInClusterConfig()
+	if err != nil {
+		return errors.Wrap(err, "failed to get client config")
+	}
+
+	local.kubeClient, err = kubeclient.NewForConfig(config)
+	if err != nil {
+		return errors.Wrap(err, "failed to get Kubernetes clientset")
+	}
+
 	osRelease, err := utils.GetOSRelease()
 	if err != nil {
 		return errors.Wrap(err, "failed to get OS release")
@@ -61,16 +71,6 @@ func (local *Checker) Init() error {
 	local.logger = logrus.WithField("os", local.osRelease)
 
 	if local.osRelease == fmt.Sprint(consts.OperatingSystemContainerOptimizedOS) {
-		config, err := commonkube.GetInClusterConfig()
-		if err != nil {
-			return errors.Wrap(err, "failed to get client config")
-		}
-
-		local.kubeClient, err = kubeclient.NewForConfig(config)
-		if err != nil {
-			return errors.Wrap(err, "failed to get Kubernetes clientset")
-		}
-
 		return nil
 	}
 
@@ -177,6 +177,8 @@ func (local *Checker) Init() error {
 
 // Run executes the preflight checks.
 func (local *Checker) Run() error {
+	local.checkCoreDNS()
+
 	switch local.osRelease {
 	case fmt.Sprint(consts.OperatingSystemContainerOptimizedOS):
 		logrus.Infof("Checking preflight for %v", consts.OperatingSystemContainerOptimizedOS)
@@ -466,3 +468,36 @@ func (local *Checker) checkNFSv4Support() error {
 	local.collection.Log.Error = append(local.collection.Log.Error, "NFS4 is not supported")
 	return nil
 }
+
+// checkCoreDNS checks if the CoreDNS deployment in the Kubernetes cluster
+// has multiple replicas and logs warnings if it does not.
+//
+// It retrieves the CoreDNS deployment from the kubeClient and checks the
+// number of replicas specified in the deployment spec. If the number of
+// replicas is less than 2, it logs a warning indicating that CoreDNS is
+// not set to run in multiple replicas. Additionally, it checks the number
+// of ready replicas in the deployment status and logs a warning if there
+// are fewer than 2 ready replicas.
+//
+// https://github.com/longhorn/longhorn/issues/9752
+func (local *Checker) checkCoreDNS() {
+	logrus.Info("Checking if CoreDNS has multiple replicas")
+
+	deployment, err := commonkube.GetDeployment(local.kubeClient, metav1.NamespaceSystem, consts.AppNameCoreDNS)
+	if err != nil {
+		local.collection.Log.Error = append(local.collection.Log.Error, fmt.Sprintf("failed to check CoreDNS: %v", err))
+		return
+	}
+
+	if deployment.Spec.Replicas == nil || *deployment.Spec.Replicas < 2 {
+		local.collection.Log.Warn = append(local.collection.Log.Warn, "CoreDNS is set with fewer than 2 replicas; consider increasing replica count for high availability")
+		return
+	}
+
+	if deployment.Status.ReadyReplicas < 2 {
+		local.collection.Log.Warn = append(local.collection.Log.Warn, "CoreDNS has fewer than 2 ready replicas; some replicas may not be running or ready")
+		return
+	}
+
+	local.collection.Log.Info = append(local.collection.Log.Info, fmt.Sprintf("CoreDNS is set with %d replicas and %d ready replicas", *deployment.Spec.Replicas, deployment.Status.ReadyReplicas))
+}

pkg/remote/preflight/checker.go

@@ -165,7 +165,7 @@ func (remote *Checker) newClusterRole() *rbacv1.ClusterRole {
 		Rules: []rbacv1.PolicyRule{
 			{
 				APIGroups: []string{"apps"},
-				Resources: []string{"daemonsets"},
+				Resources: []string{"daemonsets", "deployments"},
 				Verbs:     []string{"get"},
 			},
 		},

vendor/modules.txt

@@ -133,7 +133,7 @@ github.com/json-iterator/go
 # github.com/liggitt/tabwriter v0.0.0-20181228230101-89fcab3d43de
 ## explicit
 github.com/liggitt/tabwriter
-# github.com/longhorn/go-common-libs v0.0.0-20241109101504-614ae31d2212
+# github.com/longhorn/go-common-libs v0.0.0-20241109101504-614ae31d2212 => github.com/c3y1huang/go-common-libs v0.0.0-20241114031240-e1117130cd3b
 ## explicit; go 1.22.7
 github.com/longhorn/go-common-libs/exec
 github.com/longhorn/go-common-libs/io
@@ -928,3 +928,4 @@ sigs.k8s.io/structured-merge-diff/v4/value
 sigs.k8s.io/yaml
 sigs.k8s.io/yaml/goyaml.v2
 sigs.k8s.io/yaml/goyaml.v3
+# github.com/longhorn/go-common-libs => github.com/c3y1huang/go-common-libs v0.0.0-20241114031240-e1117130cd3b