feat: add NodeResourcesFitPlus and ScarceResourceAvoidance plugin

apis/config/register.go

@@ -45,6 +45,8 @@ func addKnownTypes(scheme *runtime.Scheme) error {
 		&NetworkOverheadArgs{},
 		&SySchedArgs{},
 		&PeaksArgs{},
+		&NodeResourcesFitPlusArgs{},
+		&ScarceResourceAvoidanceArgs{},
 	)
 	return nil
 }

apis/config/types.go

@@ -298,3 +298,24 @@ type PowerModel struct {
 	// Power = K0 + K1 * e ^(K2 * x) : where x is utilisation
 	// Idle power of node will be K0 + K1
 }
+
+// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
+
+// ScarceResourceAvoidanceArgs defines the parameters for ScarceResourceAvoidance plugin.
+type ScarceResourceAvoidanceArgs struct {
+	metav1.TypeMeta
+	Resources []v1.ResourceName `json:"resources,omitempty"`
+}
+
+// +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
+
+// NodeResourcesFitPlusArgs defines the parameters for NodeResourcesFitPlus plugin.
+type NodeResourcesFitPlusArgs struct {
+	metav1.TypeMeta
+	Resources map[v1.ResourceName]ResourcesType `json:"resources"`
+}
+
+type ResourcesType struct {
+	Type   schedconfig.ScoringStrategyType `json:"type"`
+	Weight int64                           `json:"weight"`
+}

cmd/scheduler/main.go

@@ -18,6 +18,8 @@ package main
 
 import (
 	"os"
+	noderesourcesfitplus "sigs.k8s.io/scheduler-plugins/pkg/noderesourcefitplus"
+	"sigs.k8s.io/scheduler-plugins/pkg/scarceresourceavoidance"
 
 	"k8s.io/component-base/cli"
 	_ "k8s.io/component-base/metrics/prometheus/clientgo" // for rest client metric registration
@@ -64,6 +66,8 @@ func main() {
 		// app.WithPlugin(crossnodepreemption.Name, crossnodepreemption.New),
 		app.WithPlugin(podstate.Name, podstate.New),
 		app.WithPlugin(qos.Name, qos.New),
+		app.WithPlugin(noderesourcesfitplus.Name, noderesourcesfitplus.New),
+		app.WithPlugin(scarceresourceavoidance.Name, scarceresourceavoidance.New),
 	)
 
 	code := cli.Run(command)

pkg/noderesourcefitplus/node_resources_fit_plus.go

@@ -0,0 +1,152 @@
+package noderesourcesfitplus
+
+import (
+	"context"
+	"fmt"
+
+	v1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/kubernetes/pkg/api/v1/resource"
+	k8sConfig "k8s.io/kubernetes/pkg/scheduler/apis/config"
+	"k8s.io/kubernetes/pkg/scheduler/framework"
+	plfeature "k8s.io/kubernetes/pkg/scheduler/framework/plugins/feature"
+	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/noderesources"
+	"sigs.k8s.io/scheduler-plugins/apis/config"
+)
+
+const (
+	// Name is plugin name
+	Name = "NodeResourcesFitPlus"
+)
+
+var (
+	_ framework.ScorePlugin = &Plugin{}
+)
+
+type Plugin struct {
+	handle framework.Handle
+	args   *config.NodeResourcesFitPlusArgs
+}
+
+func New(_ context.Context, args runtime.Object, handle framework.Handle) (framework.Plugin, error) {
+
+	nodeResourcesFitPlusArgs, ok := args.(*config.NodeResourcesFitPlusArgs)
+
+	if !ok {
+		return nil, fmt.Errorf("want args to be of type NodeResourcesFitPlusArgs, got %T", args)
+	}
+
+	return &Plugin{
+		handle: handle,
+		args:   nodeResourcesFitPlusArgs,
+	}, nil
+}
+
+func (s *Plugin) Name() string {
+	return Name
+}
+
+func (s *Plugin) Score(ctx context.Context, state *framework.CycleState, p *v1.Pod, nodeName string) (int64, *framework.Status) {
+	nodeInfo, err := s.handle.SnapshotSharedLister().NodeInfos().Get(nodeName)
+	if err != nil {
+		return 0, framework.NewStatus(framework.Error, fmt.Sprintf("getting node %q from Snapshot: %v", nodeName, err))
+	}
+
+	var nodeScore int64
+	var weightSum int64
+
+	podRequest, _ := fitsRequest(computePodResourceRequest(p).Resource, nodeInfo)
+
+	for _, requestSourceName := range podRequest {
+		v, ok := s.args.Resources[requestSourceName]
+		if !ok {
+			continue
+		}
+		fit, err := noderesources.NewFit(ctx,
+			&k8sConfig.NodeResourcesFitArgs{
+				ScoringStrategy: &k8sConfig.ScoringStrategy{
+					Type: v.Type, // MostAllocated or LeastAllocated
+					Resources: []k8sConfig.ResourceSpec{
+						{Name: string(requestSourceName), Weight: 1},
+					},
+				},
+			}, s.handle, plfeature.Features{})
+
+		if err != nil {
+			return 0, framework.NewStatus(framework.Error, err.Error())
+		}
+		resourceScore, status := fit.(framework.ScorePlugin).Score(ctx, state, p, nodeName)
+		if !status.IsSuccess() {
+			return 0, framework.NewStatus(framework.Error, err.Error())
+		}
+
+		nodeScore += resourceScore * v.Weight
+		weightSum += v.Weight
+	}
+
+	if weightSum == 0 {
+		return framework.MaxNodeScore, framework.NewStatus(framework.Success, "")
+	}
+	scores := nodeScore / weightSum
+
+	return scores, nil
+}
+
+func (p *Plugin) ScoreExtensions() framework.ScoreExtensions {
+	return nil
+}
+
+type preFilterState struct {
+	framework.Resource
+}
+
+func computePodResourceRequest(pod *v1.Pod) *preFilterState {
+	// pod hasn't scheduled yet so we don't need to worry about InPlacePodVerticalScalingEnabled
+	reqs := resource.PodRequests(pod, resource.PodResourcesOptions{})
+	result := &preFilterState{}
+	result.SetMaxResource(reqs)
+	return result
+}
+
+func fitsRequest(podRequest framework.Resource, nodeInfo *framework.NodeInfo) ([]v1.ResourceName, []v1.ResourceName) {
+	var podRequestResource []v1.ResourceName
+	var nodeRequestResource []v1.ResourceName
+
+	if podRequest.MilliCPU > 0 {
+		podRequestResource = append(podRequestResource, v1.ResourceCPU)
+	}
+
+	if nodeInfo.Allocatable.MilliCPU > 0 {
+		nodeRequestResource = append(nodeRequestResource, v1.ResourceCPU)
+	}
+
+	if podRequest.Memory > 0 {
+		podRequestResource = append(podRequestResource, v1.ResourceMemory)
+	}
+
+	if nodeInfo.Allocatable.Memory > 0 {
+		nodeRequestResource = append(nodeRequestResource, v1.ResourceMemory)
+	}
+
+	if podRequest.EphemeralStorage > 0 {
+		podRequestResource = append(podRequestResource, v1.ResourceEphemeralStorage)
+	}
+
+	if nodeInfo.Allocatable.EphemeralStorage > 0 {
+		nodeRequestResource = append(nodeRequestResource, v1.ResourceEphemeralStorage)
+	}
+
+	for rName, rQuant := range podRequest.ScalarResources {
+		if rQuant > 0 {
+			podRequestResource = append(podRequestResource, rName)
+		}
+	}
+
+	for rName, rQuant := range nodeInfo.Allocatable.ScalarResources {
+		if rQuant > 0 {
+			nodeRequestResource = append(nodeRequestResource, rName)
+		}
+	}
+
+	return podRequestResource, nodeRequestResource
+}