OCPBUGS-100170: Rebase master to Kubernetes v1.36.3

.go-version

@@ -1 +1 @@
-1.26.4
+1.26.5

build/build-image/cross/VERSION

@@ -1 +1 @@
-v1.36.0-go1.26.4-bullseye.0
+v1.36.0-go1.26.5-bullseye.0

build/common.sh

@@ -77,8 +77,8 @@ readonly REMOTE_OUTPUT_BINPATH="${REMOTE_OUTPUT_SUBPATH}/bin"
 readonly REMOTE_OUTPUT_GOPATH="${REMOTE_OUTPUT_SUBPATH}/go"
 
 # These are the default versions (image tags) for their respective base images.
-readonly __default_distroless_iptables_version=v0.9.3
-readonly __default_go_runner_version=v2.4.0-go1.26.4-bookworm.0
+readonly __default_distroless_iptables_version=v0.9.6
+readonly __default_go_runner_version=v2.4.0-go1.26.5-bookworm.0
 readonly __default_setcap_version=bookworm-v1.0.6
 
 # The default image for all binaries which are dynamically linked.

build/dependencies.yaml

@@ -137,7 +137,7 @@ dependencies:
   # should also be updated, but go-runner is much harder to exploit and has
   # far less relevancy to go updates for Kubernetes more generally.
   - name: "registry.k8s.io/kube-cross: dependents"
-    version: v1.36.0-go1.26.4-bullseye.0
+    version: v1.36.0-go1.26.5-bullseye.0
     refPaths:
     - path: build/build-image/cross/VERSION
 
@@ -175,15 +175,15 @@ dependencies:
       match: registry\.k8s\.io\/build-image\/debian-base:[a-zA-Z]+\-v((([0-9]+)\.([0-9]+)\.([0-9]+)(?:-([0-9a-zA-Z-]+(?:\.[0-9a-zA-Z-]+)*))?)(?:\+([0-9a-zA-Z-]+(?:\.[0-9a-zA-Z-]+)*))?)
 
   - name: "registry.k8s.io/distroless-iptables: dependents"
-    version: v0.9.3
+    version: v0.9.6
     refPaths:
     - path: build/common.sh
       match: __default_distroless_iptables_version=
     - path: test/utils/image/manifest.go
       match: configs\[DistrolessIptables\] = Config{list\.BuildImageRegistry, "distroless-iptables", "v([0-9]+)\.([0-9]+)\.([0-9]+)"}
 
   - name: "registry.k8s.io/go-runner: dependents"
-    version: v2.4.0-go1.26.4-bookworm.0
+    version: v2.4.0-go1.26.5-bookworm.0
     refPaths:
     - path: build/common.sh
       match: __default_go_runner_version=

cmd/kubeadm/app/cmd/certs.go

@@ -350,7 +350,7 @@ func getInternalCfg(cfgPath string, client kubernetes.Interface, cfg kubeadmapiv
 		getNodeRegistration := true
 		getAPIEndpoint := staticpodutil.IsControlPlaneNode()
 		getComponentConfigs := true
-		internalcfg, err := configutil.FetchInitConfigurationFromCluster(client, printer, logPrefix, getNodeRegistration, getAPIEndpoint, getComponentConfigs)
+		internalcfg, err := configutil.FetchInitConfigurationFromCluster(client, printer, logPrefix, getNodeRegistration, getAPIEndpoint, getComponentConfigs, true)
 		if err == nil {
 			printer.Println() // add empty line to separate the FetchInitConfigurationFromCluster output from the command output
 			// certificate renewal or expiration checking doesn't depend on a running cluster, which means the CertificatesDir

cmd/kubeadm/app/cmd/join.go

@@ -715,7 +715,7 @@ func fetchInitConfiguration(client clientset.Interface) (*kubeadmapi.InitConfigu
 	getNodeRegistration := false
 	getAPIEndpoint := false
 	getComponentConfigs := true
-	initConfiguration, err := configutil.FetchInitConfigurationFromCluster(client, nil, "preflight", getNodeRegistration, getAPIEndpoint, getComponentConfigs)
+	initConfiguration, err := configutil.FetchInitConfigurationFromCluster(client, nil, "preflight", getNodeRegistration, getAPIEndpoint, getComponentConfigs, false)
 	if err != nil {
 		return nil, errors.Wrap(err, "unable to fetch the kubeadm-config ConfigMap")
 	}

cmd/kubeadm/app/cmd/reset.go

@@ -136,7 +136,7 @@ func newResetData(cmd *cobra.Command, opts *resetOptions, in io.Reader, out io.W
 		getNodeRegistration := true
 		getAPIEndpoint := staticpodutil.IsControlPlaneNode()
 		getComponentConfigs := true
-		initCfg, err = configutil.FetchInitConfigurationFromCluster(client, nil, "reset", getNodeRegistration, getAPIEndpoint, getComponentConfigs)
+		initCfg, err = configutil.FetchInitConfigurationFromCluster(client, nil, "reset", getNodeRegistration, getAPIEndpoint, getComponentConfigs, true)
 		if err != nil {
 			klog.Warningf("[reset] Unable to fetch the kubeadm-config ConfigMap from cluster: %v", err)
 		}

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

@@ -241,7 +241,7 @@ func newApplyData(cmd *cobra.Command, args []string, applyFlags *applyFlags) (*a
 	getNodeRegistration := true
 	isControlPlaneNode := true
 	getComponentConfigs := true
-	initCfg, err := configutil.FetchInitConfigurationFromCluster(client, nil, "upgrade", getNodeRegistration, isControlPlaneNode, getComponentConfigs)
+	initCfg, err := configutil.FetchInitConfigurationFromCluster(client, nil, "upgrade", getNodeRegistration, isControlPlaneNode, getComponentConfigs, false)
 	if err != nil {
 		if apierrors.IsNotFound(err) {
 			_, _ = printer.Printf("[upgrade] In order to upgrade, a ConfigMap called %q in the %q namespace must exist.\n", constants.KubeadmConfigConfigMap, metav1.NamespaceSystem)

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

@@ -96,7 +96,7 @@ func enforceRequirements(flagSet *pflag.FlagSet, flags *applyPlanFlags, args []s
 	getNodeRegistration := true
 	getAPIEndpoint := staticpodutil.IsControlPlaneNode()
 	getComponentConfigs := true
-	initCfg, err := configutil.FetchInitConfigurationFromCluster(client, printer, "upgrade/config", getNodeRegistration, getAPIEndpoint, getComponentConfigs)
+	initCfg, err := configutil.FetchInitConfigurationFromCluster(client, printer, "upgrade/config", getNodeRegistration, getAPIEndpoint, getComponentConfigs, false)
 	if err != nil {
 		return nil, nil, nil, nil, errors.Wrap(err, "[upgrade/init config] FATAL")
 	}

cmd/kubeadm/app/cmd/upgrade/diff.go

@@ -107,7 +107,7 @@ func validateManifestsPath(manifests ...string) (err error) {
 }
 
 // FetchInitConfigurationFunc defines the signature of the function which will fetch InitConfiguration from cluster.
-type FetchInitConfigurationFunc func(client clientset.Interface, printer output.Printer, logPrefix string, getNodeRegistration, getAPIEndpoint, getComponentConfigs bool) (*kubeadmapi.InitConfiguration, error)
+type FetchInitConfigurationFunc func(client clientset.Interface, printer output.Printer, logPrefix string, getNodeRegistration, getAPIEndpoint, getComponentConfigs, shortConfigMapGet bool) (*kubeadmapi.InitConfiguration, error)
 
 func runDiff(fs *pflag.FlagSet, flags *diffFlags, args []string, fetchInitConfigurationFromCluster FetchInitConfigurationFunc) error {
 	externalCfg := &v1beta4.UpgradeConfiguration{}
@@ -123,7 +123,7 @@ func runDiff(fs *pflag.FlagSet, flags *diffFlags, args []string, fetchInitConfig
 	getNodeRegistration := true
 	getAPIEndpoint := staticpodutil.IsControlPlaneNode()
 	getComponentConfigs := false
-	initCfg, err := fetchInitConfigurationFromCluster(client, &output.TextPrinter{}, "upgrade/diff", getNodeRegistration, getAPIEndpoint, getComponentConfigs)
+	initCfg, err := fetchInitConfigurationFromCluster(client, &output.TextPrinter{}, "upgrade/diff", getNodeRegistration, getAPIEndpoint, getComponentConfigs, false)
 	if err != nil {
 		return err
 	}

cmd/kubeadm/app/cmd/upgrade/diff_test.go

@@ -44,7 +44,7 @@ func createTestRunDiffFile(contents []byte) (string, error) {
 	return file.Name(), nil
 }
 
-func fakeFetchInitConfig(client clientset.Interface, printer output.Printer, logPrefix string, getNodeRegistration, getAPIEndpoint, getComponentConfigs bool) (*kubeadmapi.InitConfiguration, error) {
+func fakeFetchInitConfig(client clientset.Interface, printer output.Printer, logPrefix string, getNodeRegistration, getAPIEndpoint, getComponentConfigs, shortConfigMapGet bool) (*kubeadmapi.InitConfiguration, error) {
 	return &kubeadmapi.InitConfiguration{
 		ClusterConfiguration: kubeadmapi.ClusterConfiguration{
 			KubernetesVersion: "v1.0.1",

cmd/kubeadm/app/cmd/upgrade/node.go

@@ -205,7 +205,7 @@ func newNodeData(cmd *cobra.Command, nodeOptions *nodeOptions, out io.Writer) (*
 	getNodeRegistration := true
 	getAPIEndpoint := isControlPlaneNode
 	getComponentConfigs := true
-	initCfg, err := configutil.FetchInitConfigurationFromCluster(client, nil, "upgrade", getNodeRegistration, getAPIEndpoint, getComponentConfigs)
+	initCfg, err := configutil.FetchInitConfigurationFromCluster(client, nil, "upgrade", getNodeRegistration, getAPIEndpoint, getComponentConfigs, false)
 	if err != nil {
 		return nil, errors.Wrap(err, "unable to fetch the kubeadm-config ConfigMap")
 	}

cmd/kubeadm/app/util/config/cluster.go

@@ -27,6 +27,7 @@ import (
 	"time"
 
 	authv1 "k8s.io/api/authentication/v1"
+	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/util/wait"
@@ -51,8 +52,10 @@ import (
 	kubeadmruntime "k8s.io/kubernetes/cmd/kubeadm/app/util/runtime"
 )
 
-// FetchInitConfigurationFromCluster fetches configuration from a ConfigMap in the cluster
-func FetchInitConfigurationFromCluster(client clientset.Interface, printer output.Printer, logPrefix string, getNodeRegistration, getAPIEndpoint, getComponentConfigs bool) (*kubeadmapi.InitConfiguration, error) {
+// FetchInitConfigurationFromCluster fetches configuration from a ConfigMap in the cluster.
+// If shortConfigMapGet is true, a short retry is used when fetching the kubeadm-config ConfigMap,
+// which is suitable for callers like "kubeadm reset" that don't need a long retry.
+func FetchInitConfigurationFromCluster(client clientset.Interface, printer output.Printer, logPrefix string, getNodeRegistration, getAPIEndpoint, getComponentConfigs, shortConfigMapGet bool) (*kubeadmapi.InitConfiguration, error) {
 	if printer == nil {
 		printer = &output.TextPrinter{}
 	}
@@ -61,7 +64,7 @@ func FetchInitConfigurationFromCluster(client clientset.Interface, printer outpu
 	_, _ = printer.Printf("[%s] Use 'kubeadm init phase upload-config kubeadm --config your-config-file' to re-upload it.\n", logPrefix)
 
 	// Fetch the actual config from cluster
-	cfg, err := getInitConfigurationFromCluster(constants.KubernetesDir, client, getNodeRegistration, getAPIEndpoint, getComponentConfigs)
+	cfg, err := getInitConfigurationFromCluster(constants.KubernetesDir, client, getNodeRegistration, getAPIEndpoint, getComponentConfigs, shortConfigMapGet)
 	if err != nil {
 		return nil, err
 	}
@@ -78,9 +81,31 @@ func FetchInitConfigurationFromCluster(client clientset.Interface, printer outpu
 }
 
 // getInitConfigurationFromCluster is separate only for testing purposes, don't call it directly, use FetchInitConfigurationFromCluster instead
-func getInitConfigurationFromCluster(kubeconfigDir string, client clientset.Interface, getNodeRegistration, getAPIEndpoint, getComponentConfigs bool) (*kubeadmapi.InitConfiguration, error) {
+func getInitConfigurationFromCluster(kubeconfigDir string, client clientset.Interface, getNodeRegistration, getAPIEndpoint, getComponentConfigs, shortConfigMapGet bool) (*kubeadmapi.InitConfiguration, error) {
 	// Also, the config map really should be KubeadmConfigConfigMap...
-	configMap, err := apiclient.GetConfigMapWithShortRetry(client, metav1.NamespaceSystem, constants.KubeadmConfigConfigMap)
+	var configMap *v1.ConfigMap
+	var err error
+	if shortConfigMapGet {
+		configMap, err = apiclient.GetConfigMapWithShortRetry(client, metav1.NamespaceSystem, constants.KubeadmConfigConfigMap)
+	} else {
+		var lastErr error
+		err = wait.PollUntilContextTimeout(context.Background(),
+			constants.KubernetesAPICallRetryInterval,
+			kubeadmapi.GetActiveTimeouts().KubernetesAPICall.Duration,
+			true, func(_ context.Context) (bool, error) {
+				var err error
+				configMap, err = client.CoreV1().ConfigMaps(metav1.NamespaceSystem).Get(
+					context.Background(), constants.KubeadmConfigConfigMap, metav1.GetOptions{})
+				if err == nil {
+					return true, nil
+				}
+				lastErr = err
+				return false, nil
+			})
+		if err != nil {
+			err = lastErr
+		}
+	}
 	if err != nil {
 		return nil, errors.Wrap(err, "failed to get config map")
 	}

cmd/kubeadm/app/util/config/cluster_test.go

@@ -540,7 +540,7 @@ func TestGetInitConfigurationFromCluster(t *testing.T) {
 			}
 
 			getComponentConfigs := true
-			cfg, err := getInitConfigurationFromCluster(tmpdir, client, rt.getNodeRegistration, rt.getAPIEndpoint, getComponentConfigs)
+			cfg, err := getInitConfigurationFromCluster(tmpdir, client, rt.getNodeRegistration, rt.getAPIEndpoint, getComponentConfigs, true)
 			if rt.expectedError != (err != nil) {
 				t.Errorf("unexpected return err from getInitConfigurationFromCluster: %v", err)
 				return

cmd/kubeadm/app/util/etcd/etcd.go

@@ -584,14 +584,17 @@ func (c *Client) getMemberStatus(memberID uint64) (isLearner bool, started bool,
 func (c *Client) MemberPromote(learnerID uint64) error {
 	var (
 		lastError     error
+		isLearner     bool
+		isStarted     bool
 		learnerIDUint = strconv.FormatUint(learnerID, 16)
 	)
 
 	klog.V(1).Infof("[etcd] Waiting for a learner to start: %s", learnerIDUint)
 
 	err := wait.PollUntilContextTimeout(context.Background(), constants.EtcdAPICallRetryInterval, kubeadmapi.GetActiveTimeouts().EtcdAPICall.Duration,
 		true, func(_ context.Context) (bool, error) {
-			isLearner, started, err := c.getMemberStatus(learnerID)
+			var err error
+			isLearner, isStarted, err = c.getMemberStatus(learnerID)
 			if err != nil {
 				lastError = errors.WithMessagef(err, "failed to get member %s status", learnerIDUint)
 				return false, nil
@@ -600,7 +603,7 @@ func (c *Client) MemberPromote(learnerID uint64) error {
 				klog.V(1).Infof("[etcd] Member %s was already promoted.", learnerIDUint)
 				return true, nil
 			}
-			if !started {
+			if !isStarted {
 				klog.V(1).Infof("[etcd] Member %s is not started yet. Waiting for it to be started.", learnerIDUint)
 				lastError = errors.Errorf("the etcd member %s is not started", learnerIDUint)
 				return false, nil
@@ -611,6 +614,10 @@ func (c *Client) MemberPromote(learnerID uint64) error {
 		return lastError
 	}
 
+	if !isLearner {
+		return nil
+	}
+
 	klog.V(1).Infof("[etcd] Promoting a learner as a voting member: %s", learnerIDUint)
 
 	cli, err := c.newEtcdClient(c.Endpoints)
@@ -626,14 +633,39 @@ func (c *Client) MemberPromote(learnerID uint64) error {
 	// 2. context deadline exceeded
 	// 3. peer URLs already exists
 	// Once the client provides a way to check if the etcd learner is ready to promote, the retry logic can be revisited.
-	var promoteResp *clientv3.MemberPromoteResponse
+	var memberList []*etcdserverpb.Member
 	err = wait.PollUntilContextTimeout(context.Background(), constants.EtcdAPICallRetryInterval, kubeadmapi.GetActiveTimeouts().EtcdAPICall.Duration,
 		true, func(_ context.Context) (bool, error) {
+			// MemberPromote can return a transient client-side error even if the
+			// promotion already succeeded on the etcd side. Check the current
+			// member state before attempting another promotion so that retries
+			// remain idempotent.
+			resp, statusErr := c.listMembersOnce()
+			if statusErr != nil {
+				klog.V(5).Infof("[etcd] Failed to list members before promoting learner %s: %v", learnerIDUint, statusErr)
+				lastError = statusErr
+				return false, nil
+			}
+
+			for _, m := range resp.Members {
+				if m.ID != learnerID {
+					continue
+				}
+
+				if !m.IsLearner {
+					klog.V(1).Infof("[etcd] Member %s is already a voting member, treating promotion as successful", learnerIDUint)
+					memberList = resp.Members
+					return true, nil
+				}
+				break
+			}
+
 			ctx, cancel := context.WithTimeout(context.Background(), etcdTimeout)
 			defer cancel()
-			promoteResp, err = cli.MemberPromote(ctx, learnerID)
+			promoteResp, err := cli.MemberPromote(ctx, learnerID)
 			if err == nil {
 				klog.V(1).Infof("[etcd] The learner was promoted as a voting member: %s", learnerIDUint)
+				memberList = promoteResp.Members
 				return true, nil
 			}
 			klog.V(5).Infof("[etcd] Promoting the learner %s failed: %v", learnerIDUint, err)
@@ -644,7 +676,7 @@ func (c *Client) MemberPromote(learnerID uint64) error {
 		return lastError
 	}
 
-	for _, m := range promoteResp.Members {
+	for _, m := range memberList {
 		if m.ID == learnerID {
 			parsedPeerAddrs, err := url.Parse(m.PeerURLs[0])
 			if err != nil {