kubernetes – k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpreemption Index | Files

package defaultpreemption

import "k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultpreemption"

Index

Constants 

const Name = names.DefaultPreemption

Name of the plugin used in the plugin registry and configurations.

Types

type DefaultPreemption 

type DefaultPreemption struct {
	Evaluator *preemption.Evaluator

	// IsEligiblePod returns whether a victim pod is allowed to be preempted by a preemptor pod.
	// This filtering is in addition to the internal requirement that the victim pod have lower
	// priority than the preemptor pod. Any customizations should always allow system services
	// to preempt normal pods, to avoid problems if system pods are unable to find space.
	IsEligiblePod IsEligiblePodFunc

	// MoreImportantPod is used to sort eligible victims in-place in descending order of highest to
	// lowest importance. Pods with higher importance are less likely to be preempted.
	// The default behavior is to order pods by descending priority, then descending runtime duration
	// for pods with equal priority.
	MoreImportantPod MoreImportantPodFunc
	// contains filtered or unexported fields
}

DefaultPreemption is a PostFilter plugin implements the preemption logic.

func New 

func New(_ context.Context, dpArgs runtime.Object, fh framework.Handle, fts feature.Features) (*DefaultPreemption, error)

New initializes a new plugin and returns it. The plugin type is retained to allow modification.

func (*DefaultPreemption) CandidatesToVictimsMap 

func (pl *DefaultPreemption) CandidatesToVictimsMap(candidates []preemption.Candidate) map[string]*extenderv1.Victims

This function is not applicable for out-of-tree preemption plugins that exercise different preemption candidates on the same nominated node.

func (*DefaultPreemption) EventsToRegister 

func (pl *DefaultPreemption) EventsToRegister(_ context.Context) ([]fwk.ClusterEventWithHint, error)

EventsToRegister returns the possible events that may make a Pod failed by this plugin schedulable.

func (*DefaultPreemption) GetOffsetAndNumCandidates 

func (pl *DefaultPreemption) GetOffsetAndNumCandidates(numNodes int32) (int32, int32)

GetOffsetAndNumCandidates chooses a random offset and calculates the number of candidates that should be shortlisted for dry running preemption.

func (*DefaultPreemption) Name 

func (pl *DefaultPreemption) Name() string

Name returns name of the plugin. It is used in logs, etc.

func (*DefaultPreemption) OrderedScoreFuncs 

func (pl *DefaultPreemption) OrderedScoreFuncs(ctx context.Context, nodesToVictims map[string]*extenderv1.Victims) []func(node string) int64

OrderedScoreFuncs returns a list of ordered score functions to select preferable node where victims will be preempted.

func (*DefaultPreemption) PodEligibleToPreemptOthers 

func (pl *DefaultPreemption) PodEligibleToPreemptOthers(_ context.Context, pod *v1.Pod, nominatedNodeStatus *fwk.Status) (bool, string)

PodEligibleToPreemptOthers returns one bool and one string. The bool indicates whether this pod should be considered for preempting other pods or not. The string includes the reason if this pod isn't eligible. There're several reasons:

  1. The pod has a preemptionPolicy of Never.
  2. The pod has already preempted other pods and the victims are in their graceful termination period. Currently we check the node that is nominated for this pod, and as long as there are terminating pods on this node, we don't attempt to preempt more pods.

func (*DefaultPreemption) PostFilter 

func (pl *DefaultPreemption) PostFilter(ctx context.Context, state fwk.CycleState, pod *v1.Pod, m framework.NodeToStatusReader) (*framework.PostFilterResult, *fwk.Status)

PostFilter invoked at the postFilter extension point.

func (*DefaultPreemption) PreEnqueue 

func (pl *DefaultPreemption) PreEnqueue(ctx context.Context, p *v1.Pod) *fwk.Status

func (*DefaultPreemption) SelectVictimsOnNode 

func (pl *DefaultPreemption) SelectVictimsOnNode(
	ctx context.Context,
	state fwk.CycleState,
	pod *v1.Pod,
	nodeInfo *framework.NodeInfo,
	pdbs []*policy.PodDisruptionBudget) ([]*v1.Pod, int, *fwk.Status)

SelectVictimsOnNode finds minimum set of pods on the given node that should be preempted in order to make enough room for "pod" to be scheduled.

type IsEligiblePodFunc 

type IsEligiblePodFunc func(nodeInfo *framework.NodeInfo, victim *framework.PodInfo, preemptor *v1.Pod) bool

IsEligiblePodFunc is a function which may be assigned to the DefaultPreemption plugin. This may implement rules/filtering around preemption eligibility, which is in addition to the internal requirement that the victim pod have lower priority than the preemptor pod. Any customizations should always allow system services to preempt normal pods, to avoid problems if system pods are unable to find space.

type MoreImportantPodFunc 

type MoreImportantPodFunc func(pod1, pod2 *v1.Pod) bool

MoreImportantPodFunc is a function which may be assigned to the DefaultPreemption plugin. Implementations should return true if the first pod is more important than the second pod and the second one should be considered for preemption before the first one. For performance reasons, the search for nodes eligible for preemption is done by omitting all eligible victims from a node then checking whether the preemptor fits on the node without them, before adding back victims (starting from the most important) that still fit with the preemptor. The default behavior is to not consider pod affinity between the preemptor and the victims, as affinity between pods that are eligible to preempt each other isn't recommended.

Source Files

default_preemption.go

Version
v1.34.0-beta.0
Published
Jul 16, 2025
Platform
windows/amd64
Imports
24 packages
Last checked
6 minutes ago –

Tools for package owners.