Merge pull request #56130 from anguslees/kubeadm-nodehealth

cmd/kubeadm/app/cmd/upgrade/apply.go

@@ -119,7 +119,7 @@ func NewCmdApply(parentFlags *cmdUpgradeFlags) *cobra.Command {
 func RunApply(flags *applyFlags) error {
 
 	// Start with the basics, verify that the cluster is healthy and get the configuration from the cluster (using the ConfigMap)
-	upgradeVars, err := enforceRequirements(flags.parent.featureGatesString, flags.parent.kubeConfigPath, flags.parent.cfgPath, flags.parent.printConfig, flags.dryRun)
+	upgradeVars, err := enforceRequirements(flags.parent.featureGatesString, flags.parent.kubeConfigPath, flags.parent.cfgPath, flags.parent.printConfig, flags.dryRun, flags.parent.ignorePreflightErrorsSet)
 	if err != nil {
 		return err
 	}

cmd/kubeadm/app/cmd/upgrade/common.go

@@ -48,14 +48,14 @@ type upgradeVariables struct {
 }
 
 // enforceRequirements verifies that it's okay to upgrade and then returns the variables needed for the rest of the procedure
-func enforceRequirements(featureGatesString, kubeConfigPath, cfgPath string, printConfig, dryRun bool) (*upgradeVariables, error) {
+func enforceRequirements(featureGatesString, kubeConfigPath, cfgPath string, printConfig, dryRun bool, ignoreChecksErrors sets.String) (*upgradeVariables, error) {
 	client, err := getClient(kubeConfigPath, dryRun)
 	if err != nil {
 		return nil, fmt.Errorf("couldn't create a Kubernetes client from file %q: %v", kubeConfigPath, err)
 	}
 
 	// Run healthchecks against the cluster
-	if err := upgrade.CheckClusterHealth(client); err != nil {
+	if err := upgrade.CheckClusterHealth(client, ignoreChecksErrors); err != nil {
 		return nil, fmt.Errorf("[upgrade/health] FATAL: %v", err)
 	}
 

cmd/kubeadm/app/cmd/upgrade/plan.go

@@ -55,7 +55,7 @@ func NewCmdPlan(parentFlags *cmdUpgradeFlags) *cobra.Command {
 // RunPlan takes care of outputting available versions to upgrade to for the user
 func RunPlan(parentFlags *cmdUpgradeFlags) error {
 	// Start with the basics, verify that the cluster is healthy, build a client and a versionGetter. Never set dry-run for plan.
-	upgradeVars, err := enforceRequirements(parentFlags.featureGatesString, parentFlags.kubeConfigPath, parentFlags.cfgPath, parentFlags.printConfig, false)
+	upgradeVars, err := enforceRequirements(parentFlags.featureGatesString, parentFlags.kubeConfigPath, parentFlags.cfgPath, parentFlags.printConfig, false, parentFlags.ignorePreflightErrorsSet)
 	if err != nil {
 		return err
 	}

cmd/kubeadm/app/phases/upgrade/BUILD

@@ -32,6 +32,7 @@ go_library(
         "//cmd/kubeadm/app/phases/etcd:go_default_library",
         "//cmd/kubeadm/app/phases/selfhosting:go_default_library",
         "//cmd/kubeadm/app/phases/uploadconfig:go_default_library",
+        "//cmd/kubeadm/app/preflight:go_default_library",
         "//cmd/kubeadm/app/util:go_default_library",
         "//cmd/kubeadm/app/util/apiclient:go_default_library",
         "//cmd/kubeadm/app/util/config:go_default_library",
@@ -42,8 +43,10 @@ go_library(
         "//vendor/k8s.io/api/core/v1:go_default_library",
         "//vendor/k8s.io/apimachinery/pkg/api/errors:go_default_library",
         "//vendor/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
+        "//vendor/k8s.io/apimachinery/pkg/labels:go_default_library",
         "//vendor/k8s.io/apimachinery/pkg/runtime:go_default_library",
         "//vendor/k8s.io/apimachinery/pkg/util/errors:go_default_library",
+        "//vendor/k8s.io/apimachinery/pkg/util/sets:go_default_library",
         "//vendor/k8s.io/client-go/kubernetes:go_default_library",
     ],
 )

cmd/kubeadm/app/phases/upgrade/health.go

@@ -24,73 +24,72 @@ import (
 	apps "k8s.io/api/apps/v1beta2"
 	"k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/util/sets"
 	clientset "k8s.io/client-go/kubernetes"
 	"k8s.io/kubernetes/cmd/kubeadm/app/constants"
+	"k8s.io/kubernetes/cmd/kubeadm/app/preflight"
 )
 
 // healthCheck is a helper struct for easily performing healthchecks against the cluster and printing the output
 type healthCheck struct {
-	description, okMessage, failMessage string
-	// f is invoked with a k8s client passed to it. Should return an optional warning and/or an error
+	name   string
+	client clientset.Interface
+	// f is invoked with a k8s client passed to it. Should return an optional error
 	f func(clientset.Interface) error
 }
 
+// Check is part of the preflight.Checker interface
+func (c *healthCheck) Check() (warnings, errors []error) {
+	if err := c.f(c.client); err != nil {
+		return nil, []error{err}
+	}
+	return nil, nil
+}
+
+// Name is part of the preflight.Checker interface
+func (c *healthCheck) Name() string {
+	return c.name
+}
+
 // CheckClusterHealth makes sure:
 // - the API /healthz endpoint is healthy
-// - all Nodes are Ready
+// - all master Nodes are Ready
 // - (if self-hosted) that there are DaemonSets with at least one Pod for all control plane components
 // - (if static pod-hosted) that all required Static Pod manifests exist on disk
-func CheckClusterHealth(client clientset.Interface) error {
+func CheckClusterHealth(client clientset.Interface, ignoreChecksErrors sets.String) error {
 	fmt.Println("[upgrade] Making sure the cluster is healthy:")
 
-	healthChecks := []healthCheck{
-		{
-			description: "API Server health",
-			okMessage:   "Healthy",
-			failMessage: "Unhealthy",
+	healthChecks := []preflight.Checker{
+		&healthCheck{
+			name:   "APIServerHealth",
+			client: client,
 			f:      apiServerHealthy,
 		},
-		{
-			description: "Node health",
-			okMessage:   "All Nodes are healthy",
-			failMessage: "More than one Node unhealthy",
-			f:           nodesHealthy,
+		&healthCheck{
+			name:   "MasterNodesReady",
+			client: client,
+			f:      masterNodesReady,
 		},
 		// TODO: Add a check for ComponentStatuses here?
 	}
 
 	// Run slightly different health checks depending on control plane hosting type
 	if IsControlPlaneSelfHosted(client) {
-		healthChecks = append(healthChecks, healthCheck{
-			description: "Control plane DaemonSet health",
-			okMessage:   "All control plane DaemonSets are healthy",
-			failMessage: "More than one control plane DaemonSet unhealthy",
+		healthChecks = append(healthChecks, &healthCheck{
+			name:   "ControlPlaneHealth",
+			client: client,
 			f:      controlPlaneHealth,
 		})
 	} else {
-		healthChecks = append(healthChecks, healthCheck{
-			description: "Static Pod manifests exists on disk",
-			okMessage:   "All manifests exist on disk",
-			failMessage: "Some manifests don't exist on disk",
+		healthChecks = append(healthChecks, &healthCheck{
+			name:   "StaticPodManifest",
+			client: client,
 			f:      staticPodManifestHealth,
 		})
 	}
 
-	return runHealthChecks(client, healthChecks)
-}
-
-// runHealthChecks runs a set of health checks against the cluster
-func runHealthChecks(client clientset.Interface, healthChecks []healthCheck) error {
-	for _, check := range healthChecks {
-
-		err := check.f(client)
-		if err != nil {
-			fmt.Printf("[upgrade/health] Checking %s: %s\n", check.description, check.failMessage)
-			return fmt.Errorf("The cluster is not in an upgradeable state due to: %v", err)
-		}
-		fmt.Printf("[upgrade/health] Checking %s: %s\n", check.description, check.okMessage)
-	}
-	return nil
+	return preflight.RunChecks(healthChecks, os.Stderr, ignoreChecksErrors)
 }
 
 // apiServerHealthy checks whether the API server's /healthz endpoint is healthy
@@ -108,16 +107,25 @@ func apiServerHealthy(client clientset.Interface) error {
 	return nil
 }
 
-// nodesHealthy checks whether all Nodes in the cluster are in the Running state
-func nodesHealthy(client clientset.Interface) error {
-	nodes, err := client.CoreV1().Nodes().List(metav1.ListOptions{})
+// masterNodesReady checks whether all master Nodes in the cluster are in the Running state
+func masterNodesReady(client clientset.Interface) error {
+	selector := labels.SelectorFromSet(labels.Set(map[string]string{
+		constants.LabelNodeRoleMaster: "",
+	}))
+	masters, err := client.CoreV1().Nodes().List(metav1.ListOptions{
+		LabelSelector: selector.String(),
+	})
 	if err != nil {
-		return fmt.Errorf("couldn't list all nodes in cluster: %v", err)
+		return fmt.Errorf("couldn't list masters in cluster: %v", err)
+	}
+
+	if len(masters.Items) == 0 {
+		return fmt.Errorf("failed to find any nodes with master role")
 	}
 
-	notReadyNodes := getNotReadyNodes(nodes.Items)
-	if len(notReadyNodes) != 0 {
-		return fmt.Errorf("there are NotReady Nodes in the cluster: %v", notReadyNodes)
+	notReadyMasters := getNotReadyNodes(masters.Items)
+	if len(notReadyMasters) != 0 {
+		return fmt.Errorf("there are NotReady masters in the cluster: %v", notReadyMasters)
 	}
 	return nil
 }