Register and enable defaultpreemption plugin

- Enable defaultpreemption as a PostFilter plugin
- Remote legacy hard-coded preemption logic

cmd/kube-scheduler/app/server_test.go

@@ -124,6 +124,9 @@ profiles:
     filter:
       disabled:
       - name: "*"
+    postFilter:
+      disabled:
+      - name: "*"
     preScore:
       disabled:
       - name: "*"
@@ -175,6 +178,9 @@ profiles:
 			{Name: "PodTopologySpread"},
 			{Name: "InterPodAffinity"},
 		},
+		"PostFilterPlugin": {
+			{Name: "DefaultPreemption"},
+		},
 		"PreScorePlugin": {
 			{Name: "InterPodAffinity"},
 			{Name: "PodTopologySpread"},
@@ -221,16 +227,17 @@ profiles:
 			},
 			wantPlugins: map[string]map[string][]kubeschedulerconfig.Plugin{
 				"default-scheduler": {
-					"BindPlugin":      {{Name: "DefaultBinder"}},
-					"FilterPlugin":    {{Name: "NodeResourcesFit"}, {Name: "NodePorts"}},
-					"PreFilterPlugin": {{Name: "NodeResourcesFit"}, {Name: "NodePorts"}},
-					"PreScorePlugin":  {{Name: "InterPodAffinity"}, {Name: "TaintToleration"}},
-					"QueueSortPlugin": {{Name: "PrioritySort"}},
-					"ScorePlugin":     {{Name: "InterPodAffinity", Weight: 1}, {Name: "TaintToleration", Weight: 1}},
-					"ReservePlugin":   {{Name: "VolumeBinding"}},
-					"UnreservePlugin": {{Name: "VolumeBinding"}},
-					"PreBindPlugin":   {{Name: "VolumeBinding"}},
-					"PostBindPlugin":  {{Name: "VolumeBinding"}},
+					"BindPlugin":       {{Name: "DefaultBinder"}},
+					"FilterPlugin":     {{Name: "NodeResourcesFit"}, {Name: "NodePorts"}},
+					"PreFilterPlugin":  {{Name: "NodeResourcesFit"}, {Name: "NodePorts"}},
+					"PostFilterPlugin": {{Name: "DefaultPreemption"}},
+					"PreScorePlugin":   {{Name: "InterPodAffinity"}, {Name: "TaintToleration"}},
+					"QueueSortPlugin":  {{Name: "PrioritySort"}},
+					"ScorePlugin":      {{Name: "InterPodAffinity", Weight: 1}, {Name: "TaintToleration", Weight: 1}},
+					"ReservePlugin":    {{Name: "VolumeBinding"}},
+					"UnreservePlugin":  {{Name: "VolumeBinding"}},
+					"PreBindPlugin":    {{Name: "VolumeBinding"}},
+					"PostBindPlugin":   {{Name: "VolumeBinding"}},
 				},
 			},
 		},
@@ -307,6 +314,9 @@ profiles:
 						{Name: "PodTopologySpread"},
 						{Name: "InterPodAffinity"},
 					},
+					"PostFilterPlugin": {
+						{Name: "DefaultPreemption"},
+					},
 					"PreScorePlugin": {
 						{Name: "InterPodAffinity"},
 						{Name: "PodTopologySpread"},

pkg/scheduler/algorithmprovider/BUILD

@@ -15,6 +15,7 @@ go_library(
         "//pkg/scheduler/apis/config:go_default_library",
         "//pkg/scheduler/framework/plugins/defaultbinder:go_default_library",
         "//pkg/scheduler/framework/plugins/defaultpodtopologyspread:go_default_library",
+        "//pkg/scheduler/framework/plugins/defaultpreemption:go_default_library",
         "//pkg/scheduler/framework/plugins/imagelocality:go_default_library",
         "//pkg/scheduler/framework/plugins/interpodaffinity:go_default_library",
         "//pkg/scheduler/framework/plugins/nodeaffinity:go_default_library",
@@ -44,6 +45,7 @@ go_test(
         "//pkg/scheduler/apis/config:go_default_library",
         "//pkg/scheduler/framework/plugins/defaultbinder:go_default_library",
         "//pkg/scheduler/framework/plugins/defaultpodtopologyspread:go_default_library",
+        "//pkg/scheduler/framework/plugins/defaultpreemption:go_default_library",
         "//pkg/scheduler/framework/plugins/imagelocality:go_default_library",
         "//pkg/scheduler/framework/plugins/interpodaffinity:go_default_library",
         "//pkg/scheduler/framework/plugins/nodeaffinity:go_default_library",

pkg/scheduler/algorithmprovider/registry.go

@@ -26,6 +26,7 @@ import (
 	schedulerapi "k8s.io/kubernetes/pkg/scheduler/apis/config"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultbinder"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpodtopologyspread"
+	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpreemption"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/imagelocality"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/interpodaffinity"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/nodeaffinity"
@@ -109,6 +110,11 @@ func getDefaultConfig() *schedulerapi.Plugins {
 				{Name: interpodaffinity.Name},
 			},
 		},
+		PostFilter: &schedulerapi.PluginSet{
+			Enabled: []schedulerapi.Plugin{
+				{Name: defaultpreemption.Name},
+			},
+		},
 		PreScore: &schedulerapi.PluginSet{
 			Enabled: []schedulerapi.Plugin{
 				{Name: interpodaffinity.Name},

pkg/scheduler/algorithmprovider/registry_test.go

@@ -21,6 +21,7 @@ import (
 
 	"github.com/google/go-cmp/cmp"
 	"k8s.io/component-base/featuregate"
+	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpreemption"
 
 	utilfeature "k8s.io/apiserver/pkg/util/feature"
 	featuregatetesting "k8s.io/component-base/featuregate/testing"
@@ -80,6 +81,11 @@ func TestClusterAutoscalerProvider(t *testing.T) {
 				{Name: interpodaffinity.Name},
 			},
 		},
+		PostFilter: &schedulerapi.PluginSet{
+			Enabled: []schedulerapi.Plugin{
+				{Name: defaultpreemption.Name},
+			},
+		},
 		PreScore: &schedulerapi.PluginSet{
 			Enabled: []schedulerapi.Plugin{
 				{Name: interpodaffinity.Name},
@@ -177,6 +183,11 @@ func TestApplyFeatureGates(t *testing.T) {
 						{Name: interpodaffinity.Name},
 					},
 				},
+				PostFilter: &schedulerapi.PluginSet{
+					Enabled: []schedulerapi.Plugin{
+						{Name: defaultpreemption.Name},
+					},
+				},
 				PreScore: &schedulerapi.PluginSet{
 					Enabled: []schedulerapi.Plugin{
 						{Name: interpodaffinity.Name},
@@ -262,6 +273,11 @@ func TestApplyFeatureGates(t *testing.T) {
 						{Name: interpodaffinity.Name},
 					},
 				},
+				PostFilter: &schedulerapi.PluginSet{
+					Enabled: []schedulerapi.Plugin{
+						{Name: defaultpreemption.Name},
+					},
+				},
 				PreScore: &schedulerapi.PluginSet{
 					Enabled: []schedulerapi.Plugin{
 						{Name: interpodaffinity.Name},

pkg/scheduler/apis/config/testing/compatibility_test.go

@@ -1414,6 +1414,9 @@ func TestAlgorithmProviderCompatibility(t *testing.T) {
 			{Name: "PodTopologySpread"},
 			{Name: "InterPodAffinity"},
 		},
+		"PostFilterPlugin": {
+			{Name: "DefaultPreemption"},
+		},
 		"PreScorePlugin": {
 			{Name: "InterPodAffinity"},
 			{Name: "PodTopologySpread"},
@@ -1483,6 +1486,9 @@ func TestAlgorithmProviderCompatibility(t *testing.T) {
 					{Name: "PodTopologySpread"},
 					{Name: "InterPodAffinity"},
 				},
+				"PostFilterPlugin": {
+					{Name: "DefaultPreemption"},
+				},
 				"PreScorePlugin": {
 					{Name: "InterPodAffinity"},
 					{Name: "PodTopologySpread"},
@@ -1572,6 +1578,9 @@ func TestPluginsConfigurationCompatibility(t *testing.T) {
 			{Name: "PodTopologySpread"},
 			{Name: "InterPodAffinity"},
 		},
+		"PostFilterPlugin": {
+			{Name: "DefaultPreemption"},
+		},
 		"PreScorePlugin": {
 			{Name: "InterPodAffinity"},
 			{Name: "PodTopologySpread"},
@@ -1771,6 +1780,11 @@ func TestPluginsConfigurationCompatibility(t *testing.T) {
 						{Name: "PodTopologySpread"},
 					},
 				},
+				PostFilter: &config.PluginSet{
+					Disabled: []config.Plugin{
+						{Name: "DefaultPreemption"},
+					},
+				},
 				PreScore: &config.PluginSet{
 					Disabled: []config.Plugin{
 						{Name: "InterPodAffinity"},
@@ -1917,6 +1931,9 @@ func TestPluginsConfigurationCompatibility(t *testing.T) {
 					{Name: "NodeResourcesFit"},
 					{Name: "NodeUnschedulable"},
 				},
+				"PostFilterPlugin": {
+					{Name: "DefaultPreemption"},
+				},
 				"PreScorePlugin": {
 					{Name: "TaintToleration"},
 					{Name: "DefaultPodTopologySpread"},

pkg/scheduler/eventhandlers.go

@@ -30,7 +30,6 @@ import (
 	coreinformers "k8s.io/client-go/informers/core/v1"
 	"k8s.io/client-go/tools/cache"
 	"k8s.io/kubernetes/pkg/features"
-	kubefeatures "k8s.io/kubernetes/pkg/features"
 	"k8s.io/kubernetes/pkg/scheduler/internal/queue"
 	"k8s.io/kubernetes/pkg/scheduler/profile"
 )
@@ -467,11 +466,6 @@ func addAllEventHandlers(
 			AddFunc: sched.onStorageClassAdd,
 		},
 	)
-
-	// TODO(Huang-Wei): remove this hack when defaultpreemption plugin is enabled.
-	if utilfeature.DefaultFeatureGate.Enabled(kubefeatures.PodDisruptionBudget) {
-		informerFactory.Policy().V1beta1().PodDisruptionBudgets().Lister()
-	}
 }
 
 func nodeSchedulingPropertiesChange(newNode *v1.Node, oldNode *v1.Node) string {

pkg/scheduler/framework/plugins/BUILD

@@ -12,6 +12,7 @@ go_library(
         "//pkg/scheduler/apis/config:go_default_library",
         "//pkg/scheduler/framework/plugins/defaultbinder:go_default_library",
         "//pkg/scheduler/framework/plugins/defaultpodtopologyspread:go_default_library",
+        "//pkg/scheduler/framework/plugins/defaultpreemption:go_default_library",
         "//pkg/scheduler/framework/plugins/imagelocality:go_default_library",
         "//pkg/scheduler/framework/plugins/interpodaffinity:go_default_library",
         "//pkg/scheduler/framework/plugins/nodeaffinity:go_default_library",

pkg/scheduler/framework/plugins/defaultpreemption/BUILD

@@ -6,10 +6,13 @@ go_library(
     importpath = "k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpreemption",
     visibility = ["//visibility:public"],
     deps = [
+        "//pkg/features:go_default_library",
         "//pkg/scheduler/core:go_default_library",
         "//pkg/scheduler/framework/v1alpha1:go_default_library",
+        "//pkg/scheduler/metrics:go_default_library",
         "//staging/src/k8s.io/api/core/v1:go_default_library",
         "//staging/src/k8s.io/apimachinery/pkg/runtime:go_default_library",
+        "//staging/src/k8s.io/apiserver/pkg/util/feature:go_default_library",
     ],
 )
 

pkg/scheduler/framework/plugins/defaultpreemption/default_preemption.go

@@ -18,11 +18,15 @@ package defaultpreemption
 
 import (
 	"context"
+	"time"
 
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	utilfeature "k8s.io/apiserver/pkg/util/feature"
+	kubefeatures "k8s.io/kubernetes/pkg/features"
 	"k8s.io/kubernetes/pkg/scheduler/core"
 	framework "k8s.io/kubernetes/pkg/scheduler/framework/v1alpha1"
+	"k8s.io/kubernetes/pkg/scheduler/metrics"
 )
 
 const (
@@ -45,11 +49,22 @@ func (pl *DefaultPreemption) Name() string {
 // New initializes a new plugin and returns it.
 func New(_ runtime.Object, fh framework.FrameworkHandle) (framework.Plugin, error) {
 	pl := DefaultPreemption{fh}
+	if utilfeature.DefaultFeatureGate.Enabled(kubefeatures.PodDisruptionBudget) {
+		// A hack to initialize pdbLister in sharedInformerFactory.
+		fh.SharedInformerFactory().Policy().V1beta1().PodDisruptionBudgets().Lister()
+	}
 	return &pl, nil
 }
 
 // PostFilter invoked at the postFilter extension point.
 func (pl *DefaultPreemption) PostFilter(ctx context.Context, state *framework.CycleState, pod *v1.Pod, m framework.NodeToStatusMap) (*framework.PostFilterResult, *framework.Status) {
+	preemptionStartTime := time.Now()
+	defer func() {
+		metrics.PreemptionAttempts.Inc()
+		metrics.SchedulingAlgorithmPreemptionEvaluationDuration.Observe(metrics.SinceInSeconds(preemptionStartTime))
+		metrics.DeprecatedSchedulingDuration.WithLabelValues(metrics.PreemptionEvaluation).Observe(metrics.SinceInSeconds(preemptionStartTime))
+	}()
+
 	nnn, err := core.Preempt(ctx, pl.fh, state, pod, m)
 	if err != nil {
 		return nil, framework.NewStatus(framework.Error, err.Error())

pkg/scheduler/framework/plugins/registry.go

@@ -19,6 +19,7 @@ package plugins
 import (
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultbinder"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpodtopologyspread"
+	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpreemption"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/imagelocality"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/interpodaffinity"
 	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/nodeaffinity"
@@ -71,5 +72,6 @@ func NewInTreeRegistry() runtime.Registry {
 		serviceaffinity.Name:                       serviceaffinity.New,
 		queuesort.Name:                             queuesort.New,
 		defaultbinder.Name:                         defaultbinder.New,
+		defaultpreemption.Name:                     defaultpreemption.New,
 	}
 }

pkg/scheduler/framework/runtime/framework.go

@@ -525,7 +525,12 @@ func (f *frameworkImpl) runFilterPlugin(ctx context.Context, pl framework.Filter
 
 // RunPostFilterPlugins runs the set of configured PostFilter plugins until the first
 // Success or Error is met, otherwise continues to execute all plugins.
-func (f *frameworkImpl) RunPostFilterPlugins(ctx context.Context, state *framework.CycleState, pod *v1.Pod, filteredNodeStatusMap framework.NodeToStatusMap) (*framework.PostFilterResult, *framework.Status) {
+func (f *frameworkImpl) RunPostFilterPlugins(ctx context.Context, state *framework.CycleState, pod *v1.Pod, filteredNodeStatusMap framework.NodeToStatusMap) (_ *framework.PostFilterResult, status *framework.Status) {
+	startTime := time.Now()
+	defer func() {
+		metrics.FrameworkExtensionPointDuration.WithLabelValues(postFilter, status.Code().String()).Observe(metrics.SinceInSeconds(startTime))
+	}()
+
 	statuses := make(framework.PluginToStatus)
 	for _, pl := range f.postFilterPlugins {
 		r, s := f.runPostFilterPlugin(ctx, pl, state, pod, filteredNodeStatusMap)
@@ -537,6 +542,7 @@ func (f *frameworkImpl) RunPostFilterPlugins(ctx context.Context, state *framewo
 		}
 		statuses[pl.Name()] = s
 	}
+
 	return nil, statuses.Merge()
 }
 
@@ -936,6 +942,11 @@ func (f *frameworkImpl) HasFilterPlugins() bool {
 	return len(f.filterPlugins) > 0
 }
 
+// HasPostFilterPlugins returns true if at least one postFilter plugin is defined.
+func (f *frameworkImpl) HasPostFilterPlugins() bool {
+	return len(f.postFilterPlugins) > 0
+}
+
 // HasScorePlugins returns true if at least one score plugin is defined.
 func (f *frameworkImpl) HasScorePlugins() bool {
 	return len(f.scorePlugins) > 0

pkg/scheduler/framework/v1alpha1/interface.go

@@ -481,6 +481,9 @@ type Framework interface {
 	// HasFilterPlugins returns true if at least one filter plugin is defined.
 	HasFilterPlugins() bool
 
+	// HasPostFilterPlugins returns true if at least one postFilter plugin is defined.
+	HasPostFilterPlugins() bool
+
 	// HasScorePlugins returns true if at least one score plugin is defined.
 	HasScorePlugins() bool
 

pkg/scheduler/scheduler.go

@@ -481,17 +481,10 @@ func (sched *Scheduler) scheduleOne(ctx context.Context) {
 		// into the resources that were preempted, but this is harmless.
 		nominatedNode := ""
 		if fitError, ok := err.(*core.FitError); ok {
-			if sched.DisablePreemption {
+			if sched.DisablePreemption || !prof.HasPostFilterPlugins() {
 				klog.V(3).Infof("Pod priority feature is not enabled or preemption is disabled by scheduler configuration." +
 					" No preemption is performed.")
 			} else {
-				preemptionStartTime := time.Now()
-				// TODO(Huang-Wei): implement the preemption logic as a PostFilter plugin.
-				nominatedNode, _ = core.Preempt(schedulingCycleCtx, prof, state, pod, fitError.FilteredNodesStatuses)
-				metrics.PreemptionAttempts.Inc()
-				metrics.SchedulingAlgorithmPreemptionEvaluationDuration.Observe(metrics.SinceInSeconds(preemptionStartTime))
-				metrics.DeprecatedSchedulingDuration.WithLabelValues(metrics.PreemptionEvaluation).Observe(metrics.SinceInSeconds(preemptionStartTime))
-
 				// Run PostFilter plugins to try to make the pod schedulable in a future scheduling cycle.
 				result, status := prof.RunPostFilterPlugins(ctx, state, pod, fitError.FilteredNodesStatuses)
 				if status.Code() == framework.Error {

test/integration/scheduler/framework_test.go

@@ -601,19 +601,17 @@ func TestPostFilterPlugin(t *testing.T) {
 			expectPostFilterNumCalled: 0,
 		},
 		{
-			name:             "Filter failed and PostFilter passed",
-			rejectFilter:     true,
-			rejectPostFilter: false,
-			// TODO: change to <numNodes * 2> when the hard-coded preemption logic is removed.
-			expectFilterNumCalled:     numNodes * 3,
+			name:                      "Filter failed and PostFilter passed",
+			rejectFilter:              true,
+			rejectPostFilter:          false,
+			expectFilterNumCalled:     numNodes * 2,
 			expectPostFilterNumCalled: 1,
 		},
 		{
-			name:             "Filter failed and PostFilter failed",
-			rejectFilter:     true,
-			rejectPostFilter: true,
-			// TODO: change to <numNodes * 2> when the hard-coded preemption logic is removed.
-			expectFilterNumCalled:     numNodes * 3,
+			name:                      "Filter failed and PostFilter failed",
+			rejectFilter:              true,
+			rejectPostFilter:          true,
+			expectFilterNumCalled:     numNodes * 2,
 			expectPostFilterNumCalled: 1,
 		},
 	}
@@ -645,6 +643,11 @@ func TestPostFilterPlugin(t *testing.T) {
 						Enabled: []schedulerconfig.Plugin{
 							{Name: postfilterPluginName},
 						},
+						// Need to disable default in-tree PostFilter plugins, as they will
+						// call RunFilterPlugins and hence impact the "numFilterCalled".
+						Disabled: []schedulerconfig.Plugin{
+							{Name: "*"},
+						},
 					},
 				},
 			}