Merge pull request kubernetes#122745 from kannon92/swap-no-swap-default

[KEP-2400] add no swap as the default option for swap

cmd/kubelet/app/server.go

@@ -105,6 +105,7 @@ import (
 	kubeletmetrics "k8s.io/kubernetes/pkg/kubelet/metrics"
 	"k8s.io/kubernetes/pkg/kubelet/server"
 	"k8s.io/kubernetes/pkg/kubelet/stats/pidlimit"
+	kubelettypes "k8s.io/kubernetes/pkg/kubelet/types"
 	kubeletutil "k8s.io/kubernetes/pkg/kubelet/util"
 	utilfs "k8s.io/kubernetes/pkg/util/filesystem"
 	"k8s.io/kubernetes/pkg/util/flock"
@@ -799,6 +800,14 @@ func run(ctx context.Context, s *options.KubeletServer, kubeDeps *kubelet.Depend
 			return fmt.Errorf("topology manager policy options %v require feature gates %q enabled",
 				s.TopologyManagerPolicyOptions, features.TopologyManagerPolicyOptions)
 		}
+		if utilfeature.DefaultFeatureGate.Enabled(features.NodeSwap) {
+			if !isCgroup2UnifiedMode() && s.MemorySwap.SwapBehavior == kubelettypes.LimitedSwap {
+				klog.InfoS("Swap feature is enabled and LimitedSwap but it is only supported with cgroupv2", "CGroupV2", isCgroup2UnifiedMode(), "SwapBehavior", s.MemorySwap.SwapBehavior)
+			}
+			if !s.FailSwapOn && s.MemorySwap.SwapBehavior == "" {
+				klog.InfoS("NoSwap is set due to memorySwapBehavior not specified", "memorySwapBehavior", s.MemorySwap.SwapBehavior, "FailSwapOn", s.FailSwapOn)
+			}
+		}
 
 		kubeDeps.ContainerManager, err = cm.NewContainerManager(
 			kubeDeps.Mounter,

pkg/features/kube_features.go

@@ -539,11 +539,13 @@ const (
 	// Allow pods to failover to a different node in case of non graceful node shutdown
 	NodeOutOfServiceVolumeDetach featuregate.Feature = "NodeOutOfServiceVolumeDetach"
 
-	// owner: @iholder101
+	// owner: @iholder101 @kannon92
+	// kep: https://kep.k8s.io/2400
 	// alpha: v1.22
-	// beta1: v1.28. For more info, please look at the KEP: https://kep.k8s.io/2400.
-	//
-	// Permits kubelet to run with swap enabled
+	// beta1: v1.28 (default=false)
+	// beta2: v.1.30 (default=true)
+
+	// Permits kubelet to run with swap enabled.
 	NodeSwap featuregate.Feature = "NodeSwap"
 
 	// owner: @mortent, @atiratree, @ravig
@@ -1105,7 +1107,7 @@ var defaultKubernetesFeatureGates = map[featuregate.Feature]featuregate.FeatureS
 
 	NodeOutOfServiceVolumeDetach: {Default: true, PreRelease: featuregate.GA, LockToDefault: true}, // remove in 1.31
 
-	NodeSwap: {Default: false, PreRelease: featuregate.Beta},
+	NodeSwap: {Default: true, PreRelease: featuregate.Beta},
 
 	PDBUnhealthyPodEvictionPolicy: {Default: true, PreRelease: featuregate.Beta},
 

pkg/kubelet/apis/config/types.go

@@ -666,8 +666,8 @@ type ShutdownGracePeriodByPodPriority struct {
 
 type MemorySwapConfiguration struct {
 	// swapBehavior configures swap memory available to container workloads. May be one of
-	// "", "LimitedSwap": workload combined memory and swap usage cannot exceed pod memory limit
-	// "UnlimitedSwap": workloads can use unlimited swap, up to the allocatable limit.
+	// "", "NoSwap": workloads can not use swap, default option.
+	// "LimitedSwap": workload swap usage is limited. The swap limit is proportionate to the container's memory request.
 	// +featureGate=NodeSwap
 	// +optional
 	SwapBehavior string

pkg/kubelet/apis/config/v1beta1/defaults_test.go

@@ -471,7 +471,7 @@ func TestSetDefaultsKubeletConfiguration(t *testing.T) {
 				IPTablesMasqueradeBit:                     utilpointer.Int32(1),
 				IPTablesDropBit:                           utilpointer.Int32(1),
 				FailSwapOn:                                utilpointer.Bool(true),
-				MemorySwap:                                v1beta1.MemorySwapConfiguration{SwapBehavior: "UnlimitedSwap"},
+				MemorySwap:                                v1beta1.MemorySwapConfiguration{SwapBehavior: "NoSwap"},
 				ContainerLogMaxSize:                       "1Mi",
 				ContainerLogMaxFiles:                      utilpointer.Int32(1),
 				ContainerLogMaxWorkers:                    utilpointer.Int32(1),
@@ -620,7 +620,7 @@ func TestSetDefaultsKubeletConfiguration(t *testing.T) {
 				IPTablesMasqueradeBit:                     utilpointer.Int32(1),
 				IPTablesDropBit:                           utilpointer.Int32(1),
 				FailSwapOn:                                utilpointer.Bool(true),
-				MemorySwap:                                v1beta1.MemorySwapConfiguration{SwapBehavior: "UnlimitedSwap"},
+				MemorySwap:                                v1beta1.MemorySwapConfiguration{SwapBehavior: "NoSwap"},
 				ContainerLogMaxSize:                       "1Mi",
 				ContainerLogMaxFiles:                      utilpointer.Int32(1),
 				ContainerLogMaxWorkers:                    utilpointer.Int32(1),

pkg/kubelet/apis/config/validation/validation.go

@@ -195,10 +195,10 @@ func ValidateKubeletConfiguration(kc *kubeletconfig.KubeletConfiguration, featur
 	if localFeatureGate.Enabled(features.NodeSwap) {
 		switch kc.MemorySwap.SwapBehavior {
 		case "":
+		case kubetypes.NoSwap:
 		case kubetypes.LimitedSwap:
-		case kubetypes.UnlimitedSwap:
 		default:
-			allErrors = append(allErrors, fmt.Errorf("invalid configuration: memorySwap.swapBehavior %q must be one of: \"\", %q, or %q", kc.MemorySwap.SwapBehavior, kubetypes.LimitedSwap, kubetypes.UnlimitedSwap))
+			allErrors = append(allErrors, fmt.Errorf("invalid configuration: memorySwap.swapBehavior %q must be one of: \"\", %q or %q", kc.MemorySwap.SwapBehavior, kubetypes.LimitedSwap, kubetypes.NoSwap))
 		}
 	}
 	if !localFeatureGate.Enabled(features.NodeSwap) && kc.MemorySwap != (kubeletconfig.MemorySwapConfiguration{}) {

pkg/kubelet/apis/config/validation/validation_test.go

@@ -367,7 +367,7 @@ func TestValidateKubeletConfiguration(t *testing.T) {
 			conf.MemorySwap.SwapBehavior = "invalid-behavior"
 			return conf
 		},
-		errMsg: "invalid configuration: memorySwap.swapBehavior \"invalid-behavior\" must be one of: \"\", \"LimitedSwap\", or \"UnlimitedSwap\"",
+		errMsg: "invalid configuration: memorySwap.swapBehavior \"invalid-behavior\" must be one of: \"\", \"LimitedSwap\" or \"NoSwap\"",
 	}, {
 		name: "specify MemorySwap.SwapBehavior without enabling NodeSwap",
 		configure: func(conf *kubeletconfig.KubeletConfiguration) *kubeletconfig.KubeletConfiguration {

pkg/kubelet/kuberuntime/kuberuntime_container_linux.go

@@ -168,6 +168,12 @@ func (m *kubeGenericRuntimeManager) configureContainerSwapResources(lcr *runtime
 		return
 	}
 	swapConfigurationHelper := newSwapConfigurationHelper(*m.machineInfo)
+	if m.memorySwapBehavior == kubelettypes.LimitedSwap {
+		if !isCgroup2UnifiedMode() {
+			swapConfigurationHelper.ConfigureNoSwap(lcr)
+			return
+		}
+	}
 
 	if !utilfeature.DefaultFeatureGate.Enabled(kubefeatures.NodeSwap) {
 		swapConfigurationHelper.ConfigureNoSwap(lcr)
@@ -177,10 +183,12 @@ func (m *kubeGenericRuntimeManager) configureContainerSwapResources(lcr *runtime
 	// NOTE(ehashman): Behavior is defined in the opencontainers runtime spec:
 	// https://github.com/opencontainers/runtime-spec/blob/1c3f411f041711bbeecf35ff7e93461ea6789220/config-linux.md#memory
 	switch m.memorySwapBehavior {
+	case kubelettypes.NoSwap:
+		swapConfigurationHelper.ConfigureNoSwap(lcr)
 	case kubelettypes.LimitedSwap:
 		swapConfigurationHelper.ConfigureLimitedSwap(lcr, pod, container)
 	default:
-		swapConfigurationHelper.ConfigureUnlimitedSwap(lcr)
+		swapConfigurationHelper.ConfigureNoSwap(lcr)
 	}
 }
 
@@ -401,19 +409,6 @@ func (m swapConfigurationHelper) ConfigureNoSwap(lcr *runtimeapi.LinuxContainerR
 	m.configureSwap(lcr, 0)
 }
 
-func (m swapConfigurationHelper) ConfigureUnlimitedSwap(lcr *runtimeapi.LinuxContainerResources) {
-	if !isCgroup2UnifiedMode() {
-		m.ConfigureNoSwap(lcr)
-		return
-	}
-
-	if lcr.Unified == nil {
-		lcr.Unified = map[string]string{}
-	}
-
-	lcr.Unified[cm.Cgroup2MaxSwapFilename] = "max"
-}
-
 func (m swapConfigurationHelper) configureSwap(lcr *runtimeapi.LinuxContainerResources, swapMemory int64) {
 	if !isCgroup2UnifiedMode() {
 		klog.ErrorS(fmt.Errorf("swap configuration is not supported with cgroup v1"), "swap configuration under cgroup v1 is unexpected")