pkg/scheduler/framework/plugins/noderesources/fit.go - master - githubmirror/kubernetes

pkg/scheduler/framework/plugins/noderesources/fit.go
875 строк37 KB
Natasha Sarkar
scheduler: modify NodeResourcesFit plugin to support deferred pod resizes with delta calculations
20 июл 2026, 15:42
20 июл 2026, 15:427322e27
/*
Copyright 2019 The Kubernetes Authors.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/

package noderesources

import (
	"context"
	"fmt"
	"strings"

	v1 "k8s.io/api/core/v1"
	resourceapi "k8s.io/api/resource/v1"
	"k8s.io/apimachinery/pkg/runtime"
	"k8s.io/apimachinery/pkg/util/diff"
	"k8s.io/apimachinery/pkg/util/sets"
	"k8s.io/component-helpers/resource"
	"k8s.io/dynamic-resource-allocation/cel"
	"k8s.io/klog/v2"
	fwk "k8s.io/kube-scheduler/framework"
	v1helper "k8s.io/kubernetes/pkg/apis/core/v1/helper"
	"k8s.io/kubernetes/pkg/scheduler/apis/config"
	"k8s.io/kubernetes/pkg/scheduler/apis/config/validation"
	"k8s.io/kubernetes/pkg/scheduler/framework"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/dynamicresources"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/feature"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/names"
	schedutil "k8s.io/kubernetes/pkg/scheduler/util"
	"k8s.io/utils/ptr"
)

var _ fwk.PreFilterPlugin = &Fit{}
var _ fwk.FilterPlugin = &Fit{}
var _ fwk.EnqueueExtensions = &Fit{}
var _ fwk.PreScorePlugin = &Fit{}
var _ fwk.ScorePlugin = &Fit{}
var _ fwk.SignPlugin = &Fit{}
var _ fwk.PlacementScorePlugin = &Fit{}

const (
	// Name is the name of the plugin used in the plugin registry and configurations.
	Name = names.NodeResourcesFit

	// preFilterStateKey is the key in CycleState to NodeResourcesFit pre-computed data.
	// Using the name of the plugin will likely help us avoid collisions with other plugins.
	preFilterStateKey = "PreFilter" + Name

	// preScoreStateKey is the key in CycleState to NodeResourcesFit pre-computed data for Scoring.
	preScoreStateKey = "PreScore" + Name
)

// nodeResourceStrategyTypeMap maps strategy to scorer implementation
var nodeResourceStrategyTypeMap = map[config.ScoringStrategyType]scorer{
	config.LeastAllocated: func(strategy *config.ScoringStrategy) *resourceAllocationScorer {
		resources := strategy.Resources
		return &resourceAllocationScorer{
			Name:      string(config.LeastAllocated),
			scorer:    leastResourceScorer(resources),
			resources: resources,
		}
	},
	config.MostAllocated: func(strategy *config.ScoringStrategy) *resourceAllocationScorer {
		resources := strategy.Resources
		return &resourceAllocationScorer{
			Name:      string(config.MostAllocated),
			scorer:    mostResourceScorer(resources),
			resources: resources,
		}
	},
	config.RequestedToCapacityRatio: func(strategy *config.ScoringStrategy) *resourceAllocationScorer {
		resources := strategy.Resources
		return &resourceAllocationScorer{
			Name:      string(config.RequestedToCapacityRatio),
			scorer:    requestedToCapacityRatioScorer(resources, strategy.RequestedToCapacityRatio.Shape),
			resources: resources,
		}
	},
}

// Fit is a plugin that checks if a node has sufficient resources.
type Fit struct {
	ignoredResources                                   sets.Set[string]
	ignoredResourceGroups                              sets.Set[string]
	enableInPlacePodVerticalScaling                    bool
	enablePodLevelResources                            bool
	enableDRAExtendedResource                          bool
	enableInPlacePodLevelResourcesVerticalScaling      bool
	enableInPlacePodVerticalScalingSchedulerPreemption bool
	handle                                             fwk.Handle
	*resourceAllocationScorer
	placementScorer *resourceAllocationScorer
}

// ScoreExtensions of the Score plugin.
func (f *Fit) ScoreExtensions() fwk.ScoreExtensions {
	return nil
}

// preFilterState computed at PreFilter and used at Filter.
type preFilterState struct {
	framework.Resource
}

// Clone the prefilter state.
func (s *preFilterState) Clone() fwk.StateData {
	return s
}

// preScoreState computed at PreScore and used at Score.
type preScoreState struct {
	// podRequests have the same order as the resources defined in NodeResourcesBalancedAllocationArgs.Resources,
	// same for other place we store a list like that.
	podRequests []int64
	// DRA extended resource scoring related info.
	*draPreScoreState
}

// Clone implements the mandatory Clone interface. We don't really copy the data since
// there is no need for that.
func (s *preScoreState) Clone() fwk.StateData {
	return s
}

// PreScore calculates incoming pod's resource requests and writes them to the cycle state used.
func (f *Fit) PreScore(ctx context.Context, cycleState fwk.CycleState, pod *v1.Pod, nodes []fwk.NodeInfo) *fwk.Status {
	podRequests := f.calculatePodResourceRequestList(pod, f.resources)
	state := &preScoreState{
		podRequests: podRequests,
	}
	if f.enableDRAExtendedResource {
		draPreScoreState, status := getDRAPreScoredParams(f.draManager, f.resources)
		if status != nil {
			return status
		}
		if draPreScoreState != nil {
			state.draPreScoreState = draPreScoreState
		}
	}

	cycleState.Write(preScoreStateKey, state)
	return nil
}

func getPreScoreState(cycleState fwk.CycleState) (*preScoreState, error) {
	c, err := cycleState.Read(preScoreStateKey)
	if err != nil {
		return nil, fmt.Errorf("reading %q from cycleState: %w", preScoreStateKey, err)
	}

	s, ok := c.(*preScoreState)
	if !ok {
		return nil, fmt.Errorf("invalid PreScore state, got type %T", c)
	}
	return s, nil
}

// Name returns name of the plugin. It is used in logs, etc.
func (f *Fit) Name() string {
	return Name
}

// Filtering and scoring based on the container resources and overheads.
func (pl *Fit) SignPod(ctx context.Context, pod *v1.Pod) ([]fwk.SignFragment, *fwk.Status) {
	opts := ResourceRequestsOptions{
		EnablePodLevelResources:   pl.enablePodLevelResources,
		EnableDRAExtendedResource: pl.enableDRAExtendedResource,
	}
	if pl.enableDRAExtendedResource {
		podRequest := computePodResourceRequest(pod, opts)
		for rName, rQuant := range podRequest.ScalarResources {
			// Skip in case request quantity is zero
			if rQuant == 0 {
				continue
			}
			if shouldDelegateResourceToDRA(rName, nil, pl.draManager, opts) {
				return nil, fwk.NewStatus(fwk.Unschedulable, "signature disabled when dra extended resources are used in the cluster")
			}
		}
	}

	return []fwk.SignFragment{
		{Key: fwk.ResourcesSignerName, Value: computePodResourceRequest(pod, opts)},
	}, nil
}

func getScorer(strategy *config.ScoringStrategy) (*resourceAllocationScorer, error) {
	if strategy == nil {
		return nil, fmt.Errorf("scoring strategy not specified")
	}
	scorerFactory, exists := nodeResourceStrategyTypeMap[strategy.Type]
	if !exists {
		return nil, fmt.Errorf("scoring strategy %s is not supported", strategy.Type)
	}
	return scorerFactory(strategy), nil
}

// NewFit initializes a new plugin and returns it.
func NewFit(_ context.Context, plArgs runtime.Object, h fwk.Handle, fts feature.Features) (fwk.Plugin, error) {
	args, ok := plArgs.(*config.NodeResourcesFitArgs)
	if !ok {
		return nil, fmt.Errorf("want args to be of type NodeResourcesFitArgs, got %T", plArgs)
	}
	if err := validation.ValidateNodeResourcesFitArgs(nil, args); err != nil {
		return nil, err
	}

	scorer, err := getScorer(args.ScoringStrategy)
	if err != nil {
		return nil, err
	}
	if fts.EnableDRAExtendedResource {
		scorer.enableDRAExtendedResource = true
		scorer.draManager = h.SharedDRAManager()
		scorer.draFeatures = dynamicresources.AllocatorFeatures(fts)
		// Create a CEL cache for device class selector compilation
		// This cache improves performance by avoiding recompilation of the same CEL expressions
		scorer.DRACaches.celCache = cel.NewCache(10, cel.Features{
			EnableConsumableCapacity: fts.EnableDRAConsumableCapacity,
			EnableListTypeAttributes: fts.EnableDRAListTypeAttributes,
		})
	}

	pl := &Fit{
		ignoredResources:                                   sets.New(args.IgnoredResources...),
		ignoredResourceGroups:                              sets.New(args.IgnoredResourceGroups...),
		enableInPlacePodVerticalScaling:                    fts.EnableInPlacePodVerticalScaling,
		handle:                                             h,
		enablePodLevelResources:                            fts.EnablePodLevelResources,
		enableDRAExtendedResource:                          fts.EnableDRAExtendedResource,
		enableInPlacePodLevelResourcesVerticalScaling:      fts.EnableInPlacePodLevelResourcesVerticalScaling,
		enableInPlacePodVerticalScalingSchedulerPreemption: fts.EnableInPlacePodVerticalScalingSchedulerPreemption,
		resourceAllocationScorer:                           scorer,
	}

	if fts.EnableTopologyAwareWorkloadScheduling {
		placementScorer, err := getScorer(&config.ScoringStrategy{
			Type:      config.MostAllocated,
			Resources: args.ScoringStrategy.Resources,
		})
		if err != nil {
			return nil, err
		}
		pl.placementScorer = placementScorer
	}

	return pl, nil
}

// ResourceRequestsOptions contains feature gate flags for resource request computation.
type ResourceRequestsOptions struct {
	EnablePodLevelResources                            bool
	EnableDRAExtendedResource                          bool
	EnableDRANodeAllocatableResources                  bool
	EnableInPlacePodVerticalScalingSchedulerPreemption bool
}

// shouldDelegateResourceToDRA checks if the given resource should be delegated to the DRA plugin.
// It returns true if:
//  1. The resource is not a scalar resource in the node's allocatable (not provided by device plugin)
//  2. A device class mapping exists for the resource in the cache (when draManager is available)
func shouldDelegateResourceToDRA(rName v1.ResourceName, nodeInfo fwk.NodeInfo, draManager fwk.SharedDRAManager, opts ResourceRequestsOptions) bool {
	if !opts.EnableDRAExtendedResource {
		return false
	}

	if nodeInfo != nil {
		if allocatable := nodeInfo.GetAllocatable().GetScalarResources()[rName]; allocatable > 0 {
			return false
		}
	}

	if draManager == nil {
		return false
	}

	// If draManager is available, check the cache for a mapping
	cache := draManager.DeviceClassResolver()
	return cache.GetDeviceClass(rName) != nil
}

// computePodResourceRequest returns a framework.Resource that covers the largest
// width in each resource dimension. Because init-containers run sequentially, we collect
// the max in each dimension iteratively. In contrast, we sum the resource vectors for
// regular containers since they run simultaneously.
//
// # The resources defined for Overhead should be added to the calculated Resource request sum
//
// Example:
//
// Pod:
//
//	InitContainers
//	  IC1:
//	    CPU: 2
//	    Memory: 1G
//	  IC2:
//	    CPU: 2
//	    Memory: 3G
//	Containers
//	  C1:
//	    CPU: 2
//	    Memory: 1G
//	  C2:
//	    CPU: 1
//	    Memory: 1G
//
// Result: CPU: 3, Memory: 3G
func computePodResourceRequest(pod *v1.Pod, opts ResourceRequestsOptions) *preFilterState {
	// pod hasn't scheduled yet so we don't need to worry about InPlacePodVerticalScalingEnabled
	reqs := resource.PodRequests(pod, resource.PodResourcesOptions{
		// SkipPodLevelResources is set to false when PodLevelResources feature is enabled.
		SkipPodLevelResources:                    !opts.EnablePodLevelResources,
		UseDRANodeAllocatableResourceClaimStatus: opts.EnableDRANodeAllocatableResources,
	})
	result := &preFilterState{}
	result.SetMaxResource(reqs)
	return result
}

// PreFilter invoked at the prefilter extension point.
func (f *Fit) PreFilter(ctx context.Context, cycleState fwk.CycleState, pod *v1.Pod, nodes []fwk.NodeInfo) (*fwk.PreFilterResult, *fwk.Status) {
	result := computePodResourceRequest(pod, ResourceRequestsOptions{EnablePodLevelResources: f.enablePodLevelResources})

	cycleState.Write(preFilterStateKey, result)
	return nil, nil
}

// PreFilterExtensions returns prefilter extensions, pod add and remove.
func (f *Fit) PreFilterExtensions() fwk.PreFilterExtensions {
	return nil
}

func getPreFilterState(cycleState fwk.CycleState) (*preFilterState, error) {
	c, err := cycleState.Read(preFilterStateKey)
	if err != nil {
		// preFilterState doesn't exist, likely PreFilter wasn't invoked.
		return nil, fmt.Errorf("error reading %q from cycleState: %w", preFilterStateKey, err)
	}

	s, ok := c.(*preFilterState)
	if !ok {
		return nil, fmt.Errorf("%+v  convert to NodeResourcesFit.preFilterState error", c)
	}
	return s, nil
}

// EventsToRegister returns the possible events that may make a Pod
// failed by this plugin schedulable.
func (f *Fit) EventsToRegister(_ context.Context) ([]fwk.ClusterEventWithHint, error) {
	events := []fwk.ClusterEventWithHint{
		{Event: fwk.ClusterEvent{Resource: fwk.AssignedPod, ActionType: fwk.Delete}, QueueingHintFn: f.isSchedulableAfterAssignedPodDelete},
		{Event: fwk.ClusterEvent{Resource: fwk.Node, ActionType: fwk.Add | fwk.UpdateNodeAllocatable}, QueueingHintFn: f.isSchedulableAfterNodeChange},
	}
	if f.enableDRAExtendedResource {
		events = append(events,
			// A pod might be waiting for an exteneded resurce from a class to get created or modified.
			fwk.ClusterEventWithHint{Event: fwk.ClusterEvent{Resource: fwk.DeviceClass, ActionType: fwk.Add | fwk.Update}, QueueingHintFn: f.isSchedulableAfterDeviceClassEvent})
	}
	if f.enableInPlacePodVerticalScaling {
		// If InPlacePodVerticalScaling (KEP 1287) is enabled, then UpdatePodScaleDown event should be registered
		// for this plugin since a Pod update may free up resources that make other Pods schedulable.
		events = append(events,
			fwk.ClusterEventWithHint{Event: fwk.ClusterEvent{Resource: fwk.AssignedPod, ActionType: fwk.UpdatePodScaleDown}, QueueingHintFn: f.isSchedulableAfterAssignedPodScaleDown},
			fwk.ClusterEventWithHint{Event: fwk.ClusterEvent{Resource: fwk.TargetPod, ActionType: fwk.UpdatePodScaleDown}, QueueingHintFn: f.isSchedulableAfterTargetPodScaleDown})
	}
	if f.enableInPlacePodVerticalScalingSchedulerPreemption {
		events = append(events,
			fwk.ClusterEventWithHint{Event: fwk.ClusterEvent{Resource: fwk.AssignedPod, ActionType: fwk.UpdatePodScaleUp}, QueueingHintFn: f.isSchedulableAfterAssignedPodScaleUp},
			fwk.ClusterEventWithHint{Event: fwk.ClusterEvent{Resource: fwk.TargetPod, ActionType: fwk.UpdatePodScaleUp}, QueueingHintFn: f.isSchedulableAfterTargetPodScaleUp})
	}
	return events, nil
}

// isSchedulableAfterAssignedPodDelete is invoked whenever a scheduled pod
// is deleted and checks whether that change could make a previously unschedulable pod schedulable.
func (f *Fit) isSchedulableAfterAssignedPodDelete(logger klog.Logger, pod *v1.Pod, oldObj, newObj interface{}) (fwk.QueueingHint, error) {
	deletedPod, _, err := schedutil.As[*v1.Pod](oldObj, newObj)
	if err != nil {
		return fwk.Queue, err
	}

	if deletedPod.Spec.NodeName == "" && deletedPod.Status.NominatedNodeName == "" {
		logger.V(5).Info("the deleted pod was unscheduled and it wouldn't make the unscheduled pod schedulable", "pod", klog.KObj(pod), "deletedPod", klog.KObj(deletedPod))
		return fwk.QueueSkip, nil
	}

	if f.enableInPlacePodVerticalScalingSchedulerPreemption && resource.IsPodResizeDeferred(pod) {
		if deletedPod.Spec.NodeName != pod.Spec.NodeName {
			logger.V(5).Info("another scheduled pod was deleted, but it's on a different node than the deferred resize pod", "pod", klog.KObj(pod), "deletedPod", klog.KObj(deletedPod))
			return fwk.QueueSkip, nil
		}
	}

	// any deletion event to a scheduled pod could make the unscheduled pod schedulable.
	logger.V(5).Info("another scheduled pod was deleted, and it may make the unscheduled pod schedulable", "pod", klog.KObj(pod), "deletedPod", klog.KObj(deletedPod))
	return fwk.Queue, nil
}

// isSchedulableAfterTargetPodScaleDown is invoked when the target pod is scaled down, making itself schedulable.
func (f *Fit) isSchedulableAfterTargetPodScaleDown(logger klog.Logger, pod *v1.Pod, oldObj, newObj interface{}) (fwk.QueueingHint, error) {
	logger.V(5).Info("the target pod got scaled down and it may be schedulable now", "pod", klog.KObj(pod))
	return fwk.Queue, nil
}

// isSchedulableAfterTargetPodScaleUp is invoked when the target pod is scaled up.
// Requeuing is necessary because a pod parked in UnschedulableAndUnresolvable (fitting naturally)
// may exceed available headroom upon scaling up further, requiring preemption to clear space.
func (f *Fit) isSchedulableAfterTargetPodScaleUp(logger klog.Logger, pod *v1.Pod, oldObj, newObj interface{}) (fwk.QueueingHint, error) {
	if !resource.IsPodResizeDeferred(pod) {
		logger.V(5).Info("the target pod got scaled up, but it is not a deferred resize pod", "pod", klog.KObj(pod))
		return fwk.QueueSkip, nil
	}
	logger.V(5).Info("the target pod got scaled up, re-evaluating", "pod", klog.KObj(pod))
	return fwk.Queue, nil
}

// isSchedulableAfterAssignedPodScaleUp checks if another assigned pod scaled up on the same node.
// Requeuing is necessary because a competing scale-up can consume the free headroom a parked pod
// was waiting for, forcing the deferred pod back into active evaluation to initiate preemption.
func (f *Fit) isSchedulableAfterAssignedPodScaleUp(logger klog.Logger, pod *v1.Pod, oldObj, newObj interface{}) (fwk.QueueingHint, error) {
	_, modifiedPod, err := schedutil.As[*v1.Pod](oldObj, newObj)
	if err != nil {
		return fwk.Queue, err
	}
	if !resource.IsPodResizeDeferred(pod) {
		logger.V(5).Info("another assigned pod got scaled up, but target pod is not a deferred resize pod", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod))
		return fwk.QueueSkip, nil
	}
	if modifiedPod.Spec.NodeName != pod.Spec.NodeName {
		logger.V(5).Info("another assigned pod got scaled up, but on a different node than the deferred resize pod", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod))
		return fwk.QueueSkip, nil
	}
	logger.V(5).Info("another assigned pod got scaled up on the same node as the deferred resize pod, re-evaluating", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod))
	return fwk.Queue, nil
}

// isSchedulableAfterAssignedPodScaleDown is invoked when a scheduled pod got scaled down, and checks
// whether that change could make a previously unschedulable pod schedulable.
func (f *Fit) isSchedulableAfterAssignedPodScaleDown(logger klog.Logger, pod *v1.Pod, oldObj, newObj interface{}) (fwk.QueueingHint, error) {
	originalPod, modifiedPod, err := schedutil.As[*v1.Pod](oldObj, newObj)
	if err != nil {
		return fwk.Queue, err
	}

	if !f.haveEnoughRelevantResourcesDecreased(pod, originalPod, modifiedPod) {
		if loggerV := logger.V(10); loggerV.Enabled() {
			// Log more information.
			loggerV.Info("pod got scaled down, but the modification isn't related to the resource requests of the target pod", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod), "diff", diff.Diff(originalPod, modifiedPod))
		} else {
			logger.V(5).Info("pod got scaled down, but the modification isn't related to the resource requests of the target pod", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod))
		}
		return fwk.QueueSkip, nil
	}

	logger.V(5).Info("another scheduled pod got scaled down, and it may make the unscheduled pod schedulable", "pod", klog.KObj(pod), "modifiedPod", klog.KObj(modifiedPod))
	return fwk.Queue, nil
}

// haveAnyEnoughRelevantResourcesDecreased checks whether a scheduled pod's resource request that the target pod also requests got reduced.
func (f *Fit) haveEnoughRelevantResourcesDecreased(targetPod *v1.Pod, originalPod, modifiedPod *v1.Pod) bool {
	if modifiedPod.Spec.NodeName == "" {
		// If the update event is for a unscheduled Pod,
		// it wouldn't make targetPod schedulable.
		return false
	}

	if f.enableInPlacePodVerticalScalingSchedulerPreemption && resource.IsPodResizeDeferred(targetPod) {
		// A pod with deferred resize is already bound to a specific node, so changes to other nodes do not affect its schedulability.
		if modifiedPod.Spec.NodeName != targetPod.Spec.NodeName {
			return false
		}
	}

	// the other pod was scheduled, so modification or deletion may free up some resources.
	originalMaxResourceReq, modifiedMaxResourceReq := &framework.Resource{}, &framework.Resource{}
	opts := resource.PodResourcesOptions{UseStatusResources: f.enableInPlacePodVerticalScaling, InPlacePodLevelResourcesVerticalScalingEnabled: f.enableInPlacePodLevelResourcesVerticalScaling}
	originalMaxResourceReq.SetMaxResource(resource.PodRequests(originalPod, opts))
	modifiedMaxResourceReq.SetMaxResource(resource.PodRequests(modifiedPod, opts))

	// check whether the resource request of the modified pod is less than the original pod.
	podRequests := resource.PodRequests(targetPod, opts)
	for rName, rValue := range podRequests {
		if rValue.IsZero() {
			// We only care about the resources requested by the pod we are trying to schedule.
			continue
		}
		switch rName {
		case v1.ResourceCPU:
			if originalMaxResourceReq.MilliCPU > modifiedMaxResourceReq.MilliCPU {
				return true
			}
		case v1.ResourceMemory:
			if originalMaxResourceReq.Memory > modifiedMaxResourceReq.Memory {
				return true
			}
		case v1.ResourceEphemeralStorage:
			if originalMaxResourceReq.EphemeralStorage > modifiedMaxResourceReq.EphemeralStorage {
				return true
			}
		default:
			if schedutil.IsScalarResourceName(rName) && originalMaxResourceReq.ScalarResources[rName] > modifiedMaxResourceReq.ScalarResources[rName] {
				return true
			}
		}
	}
	return false
}

// isSchedulableAfterNodeChange is invoked whenever a node added or changed. It checks whether
// that change could make a previously unschedulable pod schedulable.
func (f *Fit) isSchedulableAfterNodeChange(logger klog.Logger, pod *v1.Pod, oldObj, newObj interface{}) (fwk.QueueingHint, error) {
	originalNode, modifiedNode, err := schedutil.As[*v1.Node](oldObj, newObj)
	if err != nil {
		return fwk.Queue, err
	}
	if f.enableInPlacePodVerticalScalingSchedulerPreemption && resource.IsPodResizeDeferred(pod) {
		// A pod with deferred resize is already bound to a specific node, so changes to other nodes do not affect its schedulability.
		if modifiedNode.Name != pod.Spec.NodeName {
			logger.V(5).Info("node was updated, but it is a different node than the deferred resize pod's node", "pod", klog.KObj(pod), "node", klog.KObj(modifiedNode))
			return fwk.QueueSkip, nil
		}
	}
	// Use the DRA manager's extended resource cache for event handlers
	var draManager fwk.SharedDRAManager
	if f.enableDRAExtendedResource {
		draManager = f.handle.SharedDRAManager()
	}

	opts := ResourceRequestsOptions{
		EnablePodLevelResources:   f.enablePodLevelResources,
		EnableDRAExtendedResource: f.enableDRAExtendedResource,
	}

	// Leaving in the queue, since the pod won't fit into the modified node anyway.
	if !isFit(pod, modifiedNode, draManager, opts) {
		logger.V(5).Info("node was created or updated, but it doesn't have enough resource(s) to accommodate this pod", "pod", klog.KObj(pod), "node", klog.KObj(modifiedNode))
		return fwk.QueueSkip, nil
	}
	// The pod will fit, so since it's add, unblock scheduling.
	if originalNode == nil {
		logger.V(5).Info("node was added and it might fit the pod's resource requests", "pod", klog.KObj(pod), "node", klog.KObj(modifiedNode))
		return fwk.Queue, nil
	}
	// The pod will fit, but since there was no increase in available resources, the change won't make the pod schedulable.
	if !haveAnyRequestedResourcesIncreased(pod, originalNode, modifiedNode, draManager, opts) {
		logger.V(5).Info("node was updated, but haven't changed the pod's resource requestments fit assessment", "pod", klog.KObj(pod), "node", klog.KObj(modifiedNode))
		return fwk.QueueSkip, nil
	}

	logger.V(5).Info("node was updated, and may now fit the pod's resource requests", "pod", klog.KObj(pod), "node", klog.KObj(modifiedNode))
	return fwk.Queue, nil
}

// isSchedulableAfterDeviceClassChange is invoked whenever a device class added or changed. It checks whether
// that change could make a previously unschedulable pod schedulable.
func (f *Fit) isSchedulableAfterDeviceClassEvent(logger klog.Logger, pod *v1.Pod, oldObj, newObj interface{}) (fwk.QueueingHint, error) {
	originalClass, modifiedClass, err := schedutil.As[*resourceapi.DeviceClass](oldObj, newObj)
	if err != nil {
		return fwk.Queue, err
	}
	if originalClass != nil {
		if ptr.Deref(originalClass.Spec.ExtendedResourceName, "") == ptr.Deref(modifiedClass.Spec.ExtendedResourceName, "") {
			logger.V(5).Info("device class has identical extended resource name", "pod", klog.KObj(pod), "deviceclass", klog.KObj(modifiedClass))
			return fwk.QueueSkip, nil
		}
	} else {
		// only check implicit extended resource name for Add, as device class name does not change during Update.
		reqs := resource.PodRequests(pod, resource.PodResourcesOptions{})
		if _, ok := reqs[v1.ResourceName(resourceapi.ResourceDeviceClassPrefix+modifiedClass.Name)]; ok {
			logger.V(5).Info("device class was added, and may now fit the pod's resource requests", "pod", klog.KObj(pod), "deviceclass", klog.KObj(modifiedClass))
			return fwk.Queue, nil
		}
	}
	if modifiedClass.Spec.ExtendedResourceName != nil {
		reqs := resource.PodRequests(pod, resource.PodResourcesOptions{})
		if _, ok := reqs[v1.ResourceName(*modifiedClass.Spec.ExtendedResourceName)]; ok {
			logger.V(5).Info("device class was created or updated, and may fit the pod's resoruce requests", "pod", klog.KObj(pod), "deviceclass", klog.KObj(modifiedClass))
			return fwk.Queue, nil
		}
	}
	logger.V(5).Info("created or updated deivce class extended resource name is either nil, or does not match pod's resource request", "pod", klog.KObj(pod), "deviceclass", klog.KObj(modifiedClass))
	return fwk.QueueSkip, nil
}

// haveAnyRequestedResourcesIncreased returns true if any of the resources requested by the pod have increased or if allowed pod number increased.
func haveAnyRequestedResourcesIncreased(pod *v1.Pod, originalNode, modifiedNode *v1.Node, draManager fwk.SharedDRAManager, opts ResourceRequestsOptions) bool {
	podRequest := computePodResourceRequest(pod, opts)
	originalNodeInfo := framework.NewNodeInfo()
	originalNodeInfo.SetNode(originalNode)
	modifiedNodeInfo := framework.NewNodeInfo()
	modifiedNodeInfo.SetNode(modifiedNode)

	if modifiedNodeInfo.Allocatable.GetAllowedPodNumber() > originalNodeInfo.Allocatable.GetAllowedPodNumber() {
		return true
	}

	if podRequest.MilliCPU == 0 &&
		podRequest.Memory == 0 &&
		podRequest.EphemeralStorage == 0 &&
		len(podRequest.ScalarResources) == 0 {
		return false
	}

	if (podRequest.MilliCPU > 0 && modifiedNodeInfo.Allocatable.GetMilliCPU() > originalNodeInfo.Allocatable.GetMilliCPU()) ||
		(podRequest.Memory > 0 && modifiedNodeInfo.Allocatable.GetMemory() > originalNodeInfo.Allocatable.GetMemory()) ||
		(podRequest.EphemeralStorage > 0 && modifiedNodeInfo.Allocatable.GetEphemeralStorage() > originalNodeInfo.Allocatable.GetEphemeralStorage()) {
		return true
	}

	for rName, rQuant := range podRequest.ScalarResources {
		// Skip in case request quantity is zero
		if rQuant == 0 {
			continue
		}

		if modifiedNodeInfo.Allocatable.GetScalarResources()[rName] > originalNodeInfo.Allocatable.GetScalarResources()[rName] {
			return true
		}

		if shouldDelegateResourceToDRA(rName, modifiedNodeInfo, draManager, opts) {
			return true
		}
	}
	return false
}

// isFit checks if the pod fits the node. If the node is nil, it returns false.
// It constructs a fake NodeInfo object for the node and checks if the pod fits the node.
func isFit(pod *v1.Pod, node *v1.Node, draManager fwk.SharedDRAManager, opts ResourceRequestsOptions) bool {
	if node == nil {
		return false
	}
	nodeInfo := framework.NewNodeInfo()
	nodeInfo.SetNode(node)

	return len(Fits(pod, nodeInfo, draManager, opts)) == 0
}

// Filter invoked at the filter extension point.
// Checks if a node has sufficient resources, such as cpu, memory, gpu, opaque int resources etc to run a pod.
// It returns a list of insufficient resources, if empty, then the node has all the resources requested by the pod.
func (f *Fit) Filter(ctx context.Context, cycleState fwk.CycleState, pod *v1.Pod, nodeInfo fwk.NodeInfo) *fwk.Status {
	s, err := getPreFilterState(cycleState)
	if err != nil {
		return fwk.AsStatus(err)
	}

	var draManager fwk.SharedDRAManager
	if f.enableDRAExtendedResource {
		draManager = f.handle.SharedDRAManager()
	}

	opts := ResourceRequestsOptions{
		EnablePodLevelResources:                            f.enablePodLevelResources,
		EnableDRAExtendedResource:                          f.enableDRAExtendedResource,
		EnableInPlacePodVerticalScalingSchedulerPreemption: f.enableInPlacePodVerticalScalingSchedulerPreemption,
	}

	insufficientResources := fitsRequest(s, nodeInfo, f.ignoredResources, f.ignoredResourceGroups, draManager, opts, pod)

	if len(insufficientResources) != 0 {
		// We will keep all failure reasons.
		failureReasons := make([]string, 0, len(insufficientResources))
		statusCode := fwk.Unschedulable
		for i := range insufficientResources {
			failureReasons = append(failureReasons, insufficientResources[i].Reason)

			if insufficientResources[i].Unresolvable {
				statusCode = fwk.UnschedulableAndUnresolvable
			}
		}

		return fwk.NewStatus(statusCode, failureReasons...)
	}

	return nil
}

// InsufficientResource describes what kind of resource limit is hit and caused the pod to not fit the node.
type InsufficientResource struct {
	ResourceName v1.ResourceName
	// We explicitly have a parameter for reason to avoid formatting a message on the fly
	// for common resources, which is expensive for cluster autoscaler simulations.
	Reason    string
	Requested int64
	Used      int64
	Capacity  int64
	// Unresolvable indicates whether this node could be schedulable for the pod by the preemption,
	// which is determined by comparing the node's size and the pod's request.
	Unresolvable bool
}

// Fits checks if node have enough resources to host the pod.
func Fits(pod *v1.Pod, nodeInfo fwk.NodeInfo, draManager fwk.SharedDRAManager, opts ResourceRequestsOptions) []InsufficientResource {
	return fitsRequest(computePodResourceRequest(pod, opts), nodeInfo, nil, nil, draManager, opts, pod)
}

func fitsRequest(podRequest *preFilterState, nodeInfo fwk.NodeInfo, ignoredExtendedResources, ignoredResourceGroups sets.Set[string], draManager fwk.SharedDRAManager, opts ResourceRequestsOptions, pod *v1.Pod) []InsufficientResource {
	insufficientResources := make([]InsufficientResource, 0, 4)

	allowedPodNumber := nodeInfo.GetAllocatable().GetAllowedPodNumber()
	if len(nodeInfo.GetPods())+1 > allowedPodNumber {
		insufficientResources = append(insufficientResources, InsufficientResource{
			ResourceName: v1.ResourcePods,
			Reason:       "Too many pods",
			Requested:    1,
			Used:         int64(len(nodeInfo.GetPods())),
			Capacity:     int64(allowedPodNumber),
		})
	}

	if podRequest.MilliCPU == 0 &&
		podRequest.Memory == 0 &&
		podRequest.EphemeralStorage == 0 &&
		len(podRequest.ScalarResources) == 0 {
		return insufficientResources
	}

	deltaMilliCPU, deltaMemory, deltaEphemeralStorage, deltaScalarResources := adjustDeltasToAccomodateCacheDiscrepancy(opts, podRequest, nodeInfo, pod)

	if podRequest.MilliCPU > 0 && deltaMilliCPU > (nodeInfo.GetAllocatable().GetMilliCPU()-nodeInfo.GetRequested().GetMilliCPU()) {
		insufficientResources = append(insufficientResources, InsufficientResource{
			ResourceName: v1.ResourceCPU,
			Reason:       "Insufficient cpu",
			Requested:    podRequest.MilliCPU,
			Used:         nodeInfo.GetRequested().GetMilliCPU(),
			Capacity:     nodeInfo.GetAllocatable().GetMilliCPU(),
			Unresolvable: podRequest.MilliCPU > nodeInfo.GetAllocatable().GetMilliCPU(),
		})
	}
	if podRequest.Memory > 0 && deltaMemory > (nodeInfo.GetAllocatable().GetMemory()-nodeInfo.GetRequested().GetMemory()) {
		insufficientResources = append(insufficientResources, InsufficientResource{
			ResourceName: v1.ResourceMemory,
			Reason:       "Insufficient memory",
			Requested:    podRequest.Memory,
			Used:         nodeInfo.GetRequested().GetMemory(),
			Capacity:     nodeInfo.GetAllocatable().GetMemory(),
			Unresolvable: podRequest.Memory > nodeInfo.GetAllocatable().GetMemory(),
		})
	}
	if podRequest.EphemeralStorage > 0 &&
		deltaEphemeralStorage > (nodeInfo.GetAllocatable().GetEphemeralStorage()-nodeInfo.GetRequested().GetEphemeralStorage()) {
		insufficientResources = append(insufficientResources, InsufficientResource{
			ResourceName: v1.ResourceEphemeralStorage,
			Reason:       "Insufficient ephemeral-storage",
			Requested:    podRequest.EphemeralStorage,
			Used:         nodeInfo.GetRequested().GetEphemeralStorage(),
			Capacity:     nodeInfo.GetAllocatable().GetEphemeralStorage(),
			Unresolvable: podRequest.GetEphemeralStorage() > nodeInfo.GetAllocatable().GetEphemeralStorage(),
		})
	}

	for rName, rQuant := range deltaScalarResources {
		// Skip in case request quantity is zero
		if rQuant == 0 {
			continue
		}

		if v1helper.IsExtendedResourceName(rName) {
			// If this resource is one of the extended resources that should be ignored, we will skip checking it.
			// rName is guaranteed to have a slash due to API validation.
			var rNamePrefix string
			if ignoredResourceGroups.Len() > 0 {
				rNamePrefix = strings.Split(string(rName), "/")[0]
			}
			if ignoredExtendedResources.Has(string(rName)) || ignoredResourceGroups.Has(rNamePrefix) {
				continue
			}
		}

		if shouldDelegateResourceToDRA(rName, nodeInfo, draManager, opts) {
			continue
		}
		if podRequest.ScalarResources[rName] > 0 && rQuant > (nodeInfo.GetAllocatable().GetScalarResources()[rName]-nodeInfo.GetRequested().GetScalarResources()[rName]) {
			insufficientResources = append(insufficientResources, InsufficientResource{
				ResourceName: rName,
				Reason:       fmt.Sprintf("Insufficient %v", rName),
				Requested:    podRequest.ScalarResources[rName],
				Used:         nodeInfo.GetRequested().GetScalarResources()[rName],
				Capacity:     nodeInfo.GetAllocatable().GetScalarResources()[rName],
				Unresolvable: rQuant > nodeInfo.GetAllocatable().GetScalarResources()[rName],
			})
		}
	}

	return insufficientResources
}

// adjustDeltasToAccomodateCacheDiscrepancy calculates the resource requests to evaluate
// for a pod. For an assigned pod, its desired resources are already accounted for in the
// node cache (max(desired, allocated, actual)), so ideally we only check if the node is
// overallocated. This function exists to amortize asynchronous discrepancies between the
// pod info in the scheduling queue and the node snapshot cache.
func adjustDeltasToAccomodateCacheDiscrepancy(opts ResourceRequestsOptions, podRequest *preFilterState, nodeInfo fwk.NodeInfo, pod *v1.Pod) (int64, int64, int64, map[v1.ResourceName]int64) {
	deltaMilliCPU := podRequest.MilliCPU
	deltaMemory := podRequest.Memory
	deltaEphemeralStorage := podRequest.EphemeralStorage
	deltaScalarResources := podRequest.ScalarResources

	if !opts.EnableInPlacePodVerticalScalingSchedulerPreemption || pod == nil || len(pod.Spec.NodeName) == 0 || pod.Spec.NodeName != nodeInfo.Node().Name {
		return deltaMilliCPU, deltaMemory, deltaEphemeralStorage, deltaScalarResources
	}

	var cachedPodInfo fwk.PodInfo
	for _, pInfo := range nodeInfo.GetPods() {
		if pInfo.GetPod().UID == pod.UID {
			cachedPodInfo = pInfo
			break
		}
	}
	if cachedPodInfo == nil {
		return deltaMilliCPU, deltaMemory, deltaEphemeralStorage, deltaScalarResources
	}

	cachedRes := cachedPodInfo.CalculateResource().Resource
	// We take max(0, ...) to prevent negative deltas when podRequest < cachedRes (e.g., during scale-down
	// or asynchronous cache lag). Allowing a negative delta would improperly reduce the node's requested
	// usage before the Kubelet has actually freed the resources. If a stale larger cachedRes causes
	// preemption to fail, eventual cache convergence will emit a scale-down event to wake up the pod.
	deltaMilliCPU = max(0, podRequest.MilliCPU-cachedRes.GetMilliCPU())
	deltaMemory = max(0, podRequest.Memory-cachedRes.GetMemory())
	deltaEphemeralStorage = max(0, podRequest.EphemeralStorage-cachedRes.GetEphemeralStorage())

	adjustedScalars := make(map[v1.ResourceName]int64)
	cachedScalars := cachedRes.GetScalarResources()
	for rName, rQuant := range podRequest.ScalarResources {
		adjustedScalars[rName] = max(0, rQuant-cachedScalars[rName])
	}
	deltaScalarResources = adjustedScalars

	return deltaMilliCPU, deltaMemory, deltaEphemeralStorage, deltaScalarResources
}

// Score invoked at the Score extension point.
func (f *Fit) Score(ctx context.Context, state fwk.CycleState, pod *v1.Pod, nodeInfo fwk.NodeInfo) (int64, *fwk.Status) {
	s, err := getPreScoreState(state)
	if err != nil {
		s = &preScoreState{
			podRequests: f.calculatePodResourceRequestList(pod, f.resources),
		}
		if f.enableDRAExtendedResource {
			draPreScoreState, status := getDRAPreScoredParams(f.draManager, f.resources)
			if status != nil {
				return 0, status
			}
			if draPreScoreState != nil {
				s.draPreScoreState = draPreScoreState
			}
		}
	}

	return f.score(ctx, pod, nodeInfo, s.podRequests, s.draPreScoreState)
}

// PlacementScoreExtensions is not used by this plugin.
func (f *Fit) PlacementScoreExtensions() fwk.PlacementScoreExtensions {
	return nil
}

// ScorePlacement scores capacity ratio on all nodes in the placement including all assigned pod group pods' resource requests.
func (f *Fit) ScorePlacement(ctx context.Context, state fwk.PlacementCycleState, podGroup fwk.PodGroupInfo, placement *fwk.PodGroupAssignments) (int64, *fwk.Status) {
	return f.placementScorer.scorePlacement(ctx, podGroup, placement)
}