Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type PodEvictionAdmission ¶
type PodEvictionAdmission interface {
// LoopInit initializes PodEvictionAdmission for next Updater loop with the live pods and
// pods currently controlled by VPA in this cluster.
LoopInit(allLivePods []*apiv1.Pod, vpaControlledPods map[*vpa_types.VerticalPodAutoscaler][]*apiv1.Pod)
// Admit returns true if PodEvictionAdmission decides that pod can be evicted with given recommendation.
Admit(pod *apiv1.Pod, recommendation *vpa_types.RecommendedPodResources) bool
// CleanUp cleans up any state that PodEvictionAdmission may keep. Called
// when no VPA objects are present in the cluster.
CleanUp()
}
PodEvictionAdmission controls evictions of pods.
func NewDefaultPodEvictionAdmission ¶
func NewDefaultPodEvictionAdmission() PodEvictionAdmission
NewDefaultPodEvictionAdmission constructs new PodEvictionAdmission that admits all pods.
func NewSequentialPodEvictionAdmission ¶
func NewSequentialPodEvictionAdmission(admissions []PodEvictionAdmission) PodEvictionAdmission
NewSequentialPodEvictionAdmission constructs PodEvictionAdmission that will chain provided PodEvictionAdmission objects
type PodPriority ¶
type PodPriority struct {
// Is any container outside of the recommended range.
OutsideRecommendedRange bool
// Does any container want to grow.
ScaleUp bool
// Relative difference between the total requested and total recommended resources.
ResourceDiff float64
}
PodPriority contains data for a pod update that can be used to prioritize between updates.
func (PodPriority) Less ¶
func (p PodPriority) Less(other PodPriority) bool
Less returns true if p is lower than other.
type PriorityProcessor ¶
type PriorityProcessor interface {
GetUpdatePriority(pod *apiv1.Pod, vpa *vpa_types.VerticalPodAutoscaler,
recommendation *vpa_types.RecommendedPodResources) PodPriority
}
PriorityProcessor calculates priority for pod updates.
func NewFakeProcessor ¶
func NewFakeProcessor(priorities map[string]PodPriority) PriorityProcessor
NewFakeProcessor returns a fake processor for testing that can be initialized with a map from pod name to priority expected to be returned.
func NewProcessor ¶
func NewProcessor() PriorityProcessor
NewProcessor creates a new default PriorityProcessor.
type UpdateConfig ¶
type UpdateConfig struct {
// MinChangePriority is the minimum change priority that will trigger a update.
// TODO: should have separate for Mem and CPU?
MinChangePriority float64
}
UpdateConfig holds configuration for UpdatePriorityCalculator
type UpdatePriorityCalculator ¶
type UpdatePriorityCalculator struct {
// contains filtered or unexported fields
}
UpdatePriorityCalculator is responsible for prioritizing updates on pods. It can returns a sorted list of pods in order of update priority. Update priority is proportional to fraction by which resources should be increased / decreased. i.e. pod with 10M current memory and recommendation 20M will have higher update priority than pod with 100M current memory and 150M recommendation (100% increase vs 50% increase)
func NewUpdatePriorityCalculator ¶
func NewUpdatePriorityCalculator(vpa *vpa_types.VerticalPodAutoscaler, config *UpdateConfig, recommendationProcessor vpa_api_util.RecommendationProcessor, priorityProcessor PriorityProcessor) UpdatePriorityCalculator
NewUpdatePriorityCalculator creates new UpdatePriorityCalculator for the given VPA object an update config. If the vpa resource policy is nil, there will be no policy restriction on update. If the given update config is nil, default values are used.
func (*UpdatePriorityCalculator) AddPod ¶
func (calc *UpdatePriorityCalculator) AddPod(pod *apiv1.Pod, now time.Time)
AddPod adds pod to the UpdatePriorityCalculator.
func (*UpdatePriorityCalculator) GetSortedPods ¶
func (calc *UpdatePriorityCalculator) GetSortedPods(admission PodEvictionAdmission) []*apiv1.Pod
GetSortedPods returns a list of pods ordered by update priority (highest update priority first)
Source Files