support UpdatePodScaleDown instead of UpdatePodRequest

Author: sanposhiho

pkg/scheduler/framework/events.go

@@ -57,8 +57,8 @@ var (
 	assignedPodOtherUpdate = ClusterEvent{Resource: Pod, ActionType: updatePodOther, Label: "AssignedPodUpdate"}
 	// AssignedPodDelete is the event when an assigned pod is deleted.
 	AssignedPodDelete = ClusterEvent{Resource: Pod, ActionType: Delete, Label: "AssignedPodDelete"}
-	// PodRequestChange is the event when a pod's resource request is changed.
-	PodRequestChange = ClusterEvent{Resource: Pod, ActionType: UpdatePodRequest, Label: "PodRequestChange"}
+	// PodRequestScaledDown is the event when a pod's resource request is scaled down.
+	PodRequestScaledDown = ClusterEvent{Resource: Pod, ActionType: UpdatePodScaleDown, Label: "PodRequestScaledDown"}
 	// PodLabelChange is the event when a pod's label is changed.
 	PodLabelChange = ClusterEvent{Resource: Pod, ActionType: UpdatePodLabel, Label: "PodLabelChange"}
 	// NodeSpecUnschedulableChange is the event when unschedulable node spec is changed.
@@ -108,7 +108,7 @@ var (
 func PodSchedulingPropertiesChange(newPod *v1.Pod, oldPod *v1.Pod) (events []ClusterEvent) {
 	podChangeExtracters := []podChangeExtractor{
 		extractPodLabelsChange,
-		extractPodResourceRequestChange,
+		extractPodScaleDown,
 	}
 
 	for _, fn := range podChangeExtracters {
@@ -128,13 +128,27 @@ func PodSchedulingPropertiesChange(newPod *v1.Pod, oldPod *v1.Pod) (events []Clu
 
 type podChangeExtractor func(newNode *v1.Pod, oldNode *v1.Pod) *ClusterEvent
 
-func extractPodResourceRequestChange(newPod, oldPod *v1.Pod) *ClusterEvent {
+// extractPodScaleDown interprets the update of a pod and returns PodRequestScaledDown event if any pod's resource request(s) is scaled down.
+func extractPodScaleDown(newPod, oldPod *v1.Pod) *ClusterEvent {
 	opt := resource.PodResourcesOptions{
 		InPlacePodVerticalScalingEnabled: utilfeature.DefaultFeatureGate.Enabled(features.InPlacePodVerticalScaling),
 	}
-	if !equality.Semantic.DeepEqual(resource.PodRequests(newPod, opt), resource.PodRequests(oldPod, opt)) {
-		return &PodRequestChange
+	newPodRequests := resource.PodRequests(newPod, opt)
+	oldPodRequests := resource.PodRequests(oldPod, opt)
+
+	for rName, oldReq := range oldPodRequests {
+		newReq, ok := newPodRequests[rName]
+		if !ok {
+			// The resource request of rName is removed.
+			return &PodRequestScaledDown
+		}
+
+		if oldReq.MilliValue() > newReq.MilliValue() {
+			// The resource request of rName is scaled down.
+			return &PodRequestScaledDown
+		}
 	}
+
 	return nil
 }
 

pkg/scheduler/framework/events_test.go

@@ -291,7 +291,7 @@ func Test_podSchedulingPropertiesChange(t *testing.T) {
 			},
 		},
 	}
-	podWithBigRequestAndLabel := &v1.Pod{
+	podWithSmallRequestAndLabel := &v1.Pod{
 		ObjectMeta: metav1.ObjectMeta{
 			Labels: map[string]string{"foo": "bar"},
 		},
@@ -300,7 +300,7 @@ func Test_podSchedulingPropertiesChange(t *testing.T) {
 				{
 					Name: "app",
 					Resources: v1.ResourceRequirements{
-						Requests: v1.ResourceList{v1.ResourceCPU: resource.MustParse("101m")},
+						Requests: v1.ResourceList{v1.ResourceCPU: resource.MustParse("100m")},
 					},
 				},
 			},
@@ -309,7 +309,7 @@ func Test_podSchedulingPropertiesChange(t *testing.T) {
 			ContainerStatuses: []v1.ContainerStatus{
 				{
 					Name:               "app",
-					AllocatedResources: v1.ResourceList{v1.ResourceCPU: resource.MustParse("101m")},
+					AllocatedResources: v1.ResourceList{v1.ResourceCPU: resource.MustParse("100m")},
 				},
 			},
 		},
@@ -347,16 +347,22 @@ func Test_podSchedulingPropertiesChange(t *testing.T) {
 			want:   []ClusterEvent{PodLabelChange},
 		},
 		{
-			name:   "only pod's resource request is updated",
+			name:   "pod's resource request is scaled down",
+			oldPod: podWithBigRequest,
+			newPod: podWithSmallRequest,
+			want:   []ClusterEvent{PodRequestScaledDown},
+		},
+		{
+			name:   "pod's resource request is scaled up",
 			oldPod: podWithSmallRequest,
 			newPod: podWithBigRequest,
-			want:   []ClusterEvent{PodRequestChange},
+			want:   []ClusterEvent{assignedPodOtherUpdate},
 		},
 		{
 			name:   "both pod's resource request and label are updated",
-			oldPod: podWithSmallRequest,
-			newPod: podWithBigRequestAndLabel,
-			want:   []ClusterEvent{PodLabelChange, PodRequestChange},
+			oldPod: podWithBigRequest,
+			newPod: podWithSmallRequestAndLabel,
+			want:   []ClusterEvent{PodLabelChange, PodRequestScaledDown},
 		},
 		{
 			name:   "untracked properties of pod is updated",

pkg/scheduler/framework/plugins/noderesources/fit.go

@@ -252,7 +252,7 @@ func (f *Fit) EventsToRegister(_ context.Context) ([]framework.ClusterEventWithH
 	if f.enableInPlacePodVerticalScaling {
 		// If InPlacePodVerticalScaling (KEP 1287) is enabled, then PodRequestUpdate event should be registered
 		// for this plugin since a Pod update may free up resources that make other Pods schedulable.
-		podActionType |= framework.UpdatePodRequest
+		podActionType |= framework.UpdatePodScaleDown
 	}
 	return []framework.ClusterEventWithHint{
 		{Event: framework.ClusterEvent{Resource: framework.Pod, ActionType: podActionType}, QueueingHintFn: f.isSchedulableAfterPodChange},
@@ -296,7 +296,7 @@ func (f *Fit) isSchedulableAfterPodChange(logger klog.Logger, pod *v1.Pod, oldOb
 		return framework.QueueSkip, nil
 	}
 
-	logger.V(5).Info("the max request resources of another scheduled pod got reduced and it may make the unscheduled pod schedulable", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod))
+	logger.V(5).Info("another scheduled pod or the target pod itself got scaled down, and it may make the unscheduled pod schedulable", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod))
 	return framework.Queue, nil
 }
 

pkg/scheduler/framework/plugins/noderesources/fit_test.go

@@ -1095,7 +1095,7 @@ func TestEventsToRegister(t *testing.T) {
 			"Register events with InPlacePodVerticalScaling feature enabled",
 			true,
 			[]framework.ClusterEventWithHint{
-				{Event: framework.ClusterEvent{Resource: "Pod", ActionType: framework.UpdatePodRequest | framework.Delete}},
+				{Event: framework.ClusterEvent{Resource: "Pod", ActionType: framework.UpdatePodScaleDown | framework.Delete}},
 				{Event: framework.ClusterEvent{Resource: "Node", ActionType: framework.Add | framework.Update}},
 			},
 		},

pkg/scheduler/framework/types.go

@@ -66,8 +66,8 @@ const (
 	// It's better to narrow down the scope of the event by using them instead of Update event
 	// for better performance in requeueing.
 	UpdatePodLabel
-	// UpdatePodRequest is a update for pod's resource request calculated by resource.PodRequests() function.
-	UpdatePodRequest
+	// UpdatePodScaleDown is an update for pod's scale down (i.e., any resource request is reduced).
+	UpdatePodScaleDown
 
 	// updatePodOther is a update for pod's other fields.
 	// It's used only for the internal event handling, and thus unexported.
@@ -76,7 +76,7 @@ const (
 	All ActionType = 1<<iota - 1
 
 	// Use the general Update type if you don't either know or care the specific sub-Update type to use.
-	Update = UpdateNodeAllocatable | UpdateNodeLabel | UpdateNodeTaint | UpdateNodeCondition | UpdateNodeAnnotation | UpdatePodLabel | UpdatePodRequest | updatePodOther
+	Update = UpdateNodeAllocatable | UpdateNodeLabel | UpdateNodeTaint | UpdateNodeCondition | UpdateNodeAnnotation | UpdatePodLabel | UpdatePodScaleDown | updatePodOther
 )
 
 // GVK is short for group/version/kind, which can uniquely represent a particular API resource.

pkg/scheduler/internal/queue/scheduling_queue.go

@@ -1169,7 +1169,7 @@ func (p *PriorityQueue) AssignedPodAdded(logger klog.Logger, pod *v1.Pod) {
 // may make pending pods with matching affinity terms schedulable.
 func (p *PriorityQueue) AssignedPodUpdated(logger klog.Logger, oldPod, newPod *v1.Pod, event framework.ClusterEvent) {
 	p.lock.Lock()
-	if event.Resource == framework.Pod && event.ActionType&framework.UpdatePodRequest != 0 {
+	if event.Resource == framework.Pod && event.ActionType&framework.UpdatePodScaleDown != 0 {
 		// In this case, we don't want to pre-filter Pods by getUnschedulablePodsWithCrossTopologyTerm
 		// because Pod related events may make Pods that were rejected by NodeResourceFit schedulable.
 		p.moveAllToActiveOrBackoffQueue(logger, framework.AssignedPodUpdate, oldPod, newPod, nil)

pkg/scheduler/scheduler_test.go

@@ -893,7 +893,7 @@ func Test_UnionedGVKs(t *testing.T) {
 			name:    "plugins with default profile (InPlacePodVerticalScaling: enabled)",
 			plugins: schedulerapi.PluginSet{Enabled: defaults.PluginsV1.MultiPoint.Enabled},
 			want: map[framework.GVK]framework.ActionType{
-				framework.Pod:                   framework.Add | framework.UpdatePodLabel | framework.UpdatePodRequest | framework.Delete,
+				framework.Pod:                   framework.Add | framework.UpdatePodLabel | framework.UpdatePodScaleDown | framework.Delete,
 				framework.Node:                  framework.All,
 				framework.CSINode:               framework.All - framework.Delete,
 				framework.CSIDriver:             framework.All - framework.Delete,