framework

package
v0.0.0-...-5ee5d88 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Apr 18, 2024 License: Apache-2.0 Imports: 15 Imported by: 0

Documentation

Index

Constants

View Source
const (
	// DefaultMilliCPURequest defines default milli cpu request number.
	DefaultMilliCPURequest int64 = 100 // 0.1 core
	// DefaultMemoryRequest defines default memory request size.
	DefaultMemoryRequest int64 = 200 * 1024 * 1024 // 200 MB
)

For each of these resources, a pod that doesn't request the resource explicitly will be treated as having requested the amount indicated below, for the purpose of computing priority only. This ensures that when scheduling zero-request pods, such pods will not all be scheduled to the machine with the smallest in-use request, and that when scheduling regular pods, such pods will not see zero-request pods as consuming no resources whatsoever. We chose these values to be similar to the resources that we give to cluster addon pods (#10653). But they are pretty arbitrary. As described in #11713, we use request instead of limit to deal with resource requirements.

View Source
const (
	// MaxClusterScore is the maximum score a Score plugin is expected to return.
	MaxClusterScore int64 = 100
	// MinClusterScore is the minimum score a Score plugin is expected to return.
	MinClusterScore int64 = 0
	// MaxTotalScore is the maximum total score.
	MaxTotalScore int64 = math.MaxInt64
)
View Source
const (
	// NoClusterAvailableMsg is used to format message when no clusters available.
	NoClusterAvailableMsg = "0/%v clusters are available"
)
View Source
const (
	ResourceGPU = corev1.ResourceName("nvidia.com/gpu")
)

Variables

View Source
var DefaultRatio int64 = 1

DefaultRatio as the ratio of any unknown requested resources

View Source
var DefaultRequestedRatioResources = ResourceToWeightMap{corev1.ResourceMemory: 1, corev1.ResourceCPU: 1, ResourceGPU: 4}

DefaultRequestedRatioResources is an empirical value derived from practice.

Functions

func ExtractClusterNames

func ExtractClusterNames(clusters []*fedcorev1a1.FederatedCluster) []string

func FindMatchingUntoleratedTaint

func FindMatchingUntoleratedTaint(
	taints []corev1.Taint,
	tolerations []corev1.Toleration,
	inclusionFilter taintsFilterFunc,
) (corev1.Taint, bool)

FindMatchingUntoleratedTaint checks if the given tolerations tolerates all the filtered taints, and returns the first taint without a toleration Returns true if there is an untolerated taint Returns false if all taints are tolerated

func GetKatalystResources

func GetKatalystResources(r *Resource) map[corev1.ResourceName]int64

func GetNonzeroRequestForResource

func GetNonzeroRequestForResource(resource corev1.ResourceName, requests *corev1.ResourceList) int64

GetNonzeroRequestForResource returns the default resource request if none is found or what is provided on the request.

func GetNonzeroRequests

func GetNonzeroRequests(requests *corev1.ResourceList) (int64, int64)

GetNonzeroRequests returns the default cpu and memory resource request if none is found or what is provided on the request.

func IsAttachableVolumeResourceName

func IsAttachableVolumeResourceName(name corev1.ResourceName) bool

func IsExtendedResourceName

func IsExtendedResourceName(name corev1.ResourceName) bool

IsExtendedResourceName returns true if: 1. the resource name is not in the default namespace; 2. resource name does not have "requests." prefix, to avoid confusion with the convention in quota 3. it satisfies the rules in IsQualifiedName() after converted into quota resource name

func IsHugePageResourceName

func IsHugePageResourceName(name corev1.ResourceName) bool

IsHugePageResourceName returns true if the resource name has the huge page resource prefix.

func IsNativeResource

func IsNativeResource(name corev1.ResourceName) bool

IsNativeResource returns true if the resource name is in the *kubernetes.io/ namespace. Partially-qualified (unprefixed) names are implicitly in the kubernetes.io/ namespace.

func IsPrefixedNativeResource

func IsPrefixedNativeResource(name corev1.ResourceName) bool

IsPrefixedNativeResource returns true if the resource name is in the *kubernetes.io/ namespace.

func IsScalarResourceName

func IsScalarResourceName(name corev1.ResourceName) bool

Extended and Hugepages resources

func MaxInt64

func MaxInt64(a int64, b int64) int64

func MinInt64

func MinInt64(a int64, b int64) int64

func PreCheck

func PreCheck(ctx context.Context, su *SchedulingUnit, cluster *fedcorev1a1.FederatedCluster) error

func TolerationsTolerateTaint

func TolerationsTolerateTaint(tolerations []corev1.Toleration, taint *corev1.Taint) bool

TolerationsTolerateTaint checks if taint is tolerated by any of the tolerations.

Types

type Affinity

type Affinity struct {
	// Describes cluster affinity scheduling rules for the scheduling unit.
	ClusterAffinity *ClusterAffinity `json:"clusterAffinity,omitempty"`
}

Affinity is a group of affinity scheduling rules.

type AutoMigrationInfo

type AutoMigrationInfo struct {
	// Describes the estimated max number of replicas a cluster can accommodate.
	EstimatedCapacity map[string]int64 `json:"estimatedCapacity,omitempty"`
	ScheduledReplicas map[string]int64 `json:"scheduledReplicas,omitempty"`
}

AutoMigrationInfo contains information related to auto migration.

type AutoMigrationSpec

type AutoMigrationSpec struct {
	Info                      *AutoMigrationInfo
	KeepUnschedulableReplicas bool
}

type ClusterAffinity

type ClusterAffinity struct {
	// If the affinity requirements specified by this field are not met at
	// scheduling time, the scheduling unit will not be scheduled onto the cluster.
	RequiredDuringSchedulingIgnoredDuringExecution *ClusterSelector `json:"requiredDuringSchedulingIgnoredDuringExecution,omitempty"`
	// The scheduler will prefer to schedule scheduling units to clusters that satisfy
	// the affinity expressions specified by this field, but it may choose
	// a cluster that violates one or more of the expressions. The cluster that is
	// most preferred is the one with the greatest sum of weights, i.e.
	// for each cluster that meets all of the scheduling requirements (resource
	// request, requiredDuringScheduling affinity expressions, etc.),
	// compute a sum by iterating through the elements of this field and adding
	// "weight" to the sum if the cluster matches the corresponding matchExpressions; the
	// cluster(s) with the highest sum are the most preferred.
	// +optional
	PreferredDuringSchedulingIgnoredDuringExecution []PreferredSchedulingTerm `json:"preferredDuringSchedulingIgnoredDuringExecution,omitempty"`
}

ClusterAffinity is a group of node affinity scheduling rules.

type ClusterReplicas

type ClusterReplicas struct {
	Cluster  *fedcorev1a1.FederatedCluster
	Replicas int64
}

type ClusterReplicasList

type ClusterReplicasList []ClusterReplicas

type ClusterScore

type ClusterScore struct {
	Cluster *fedcorev1a1.FederatedCluster
	Score   int64
}

type ClusterScoreList

type ClusterScoreList []ClusterScore

type ClusterSelector

type ClusterSelector struct {
	// A list of cluster selector terms. The terms are ORed.
	ClusterSelectorTerms []fedcorev1a1.ClusterSelectorTerm `json:"clusterSelectorTerms"`
}

ClusterSelector represents the union of the results of one or more label queries over a set of clusters; that is, it represents the OR of the selectors represented by the cluster selector terms.

type ClusterToResultMap

type ClusterToResultMap map[string]*Result

ClusterToResultMap declares map from cluster name to its result.

type Code

type Code int

Code is the Status code/type which is returned from plugins.

const (
	// Success means that plugin ran correctly and found resource schedulable.
	// NOTE: A nil status is also considered as "Success".
	Success Code = iota
	// Unschedulable is used when a plugin finds the resource unschedulable.
	// The accompanying status message should explain why the it is unschedulable.
	Unschedulable
	// Error is used for internal plugin errors, unexpected input, etc.
	Error
)

These are predefined codes used in a Status.

func (Code) String

func (c Code) String() string

type CustomMigrationInfo

type CustomMigrationInfo struct {
	// Describes the max number of replicas a cluster can accommodate.
	// Different from `maxReplicas` in policy, LimitedCapacity is not affected by `avoidDisruption`
	// and is a constraint that must be satisfied.
	LimitedCapacity map[string]int64 `json:"limitedCapacity,omitempty"`
	// Describes the list of unavailable clusters.
	UnavailableClusters []UnavailableCluster `json:"unavailableClusters,omitempty"`
}

type CustomMigrationSpec

type CustomMigrationSpec struct {
	Info *CustomMigrationInfo
}

type Diagnosis

type Diagnosis struct {
	ClusterToResultMap   ClusterToResultMap
	UnschedulablePlugins sets.Set[string]
}

Diagnosis records the details to diagnose a scheduling failure.

type FilterPlugin

type FilterPlugin interface {
	Plugin

	// Filter is called by the scheduling framework.
	Filter(context.Context, *SchedulingUnit, *fedcorev1a1.FederatedCluster) *Result
}

FilterPlugin is an interface for filter plugins. These filters are used to filter out clusters that are not fit for the resource.

type FitError

type FitError struct {
	NumAllClusters int
	Diagnosis      Diagnosis
}

FitError describes a fit error of a cluster.

func (*FitError) Error

func (f *FitError) Error() string

Error returns detailed information of why the pod failed to fit on each node. A message format is "0/X clusters are available: <FilterMsg>. "

type Handle

type Handle interface {
	DynamicClient() dynamic.Interface
}

type InsufficientResource

type InsufficientResource struct {
	ResourceName corev1.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
}

InsufficientResource describes what kind of resource limit is hit and caused the pod to not fit the node.

type MigrationConfig

type MigrationConfig struct {
	// Describes the migration configuration for replicas in member clusters.
	ReplicasMigrations []ReplicasMigration `json:"replicasMigrations,omitempty"`
	// Describes the migration configuration for workloads in member clusters.
	WorkloadMigrations []WorkloadMigration `json:"workloadMigrations,omitempty"`
}

type Plugin

type Plugin interface {
	Name() string
}

Plugin is the parent type for all the scheduling framework plugins.

type PluginToClusterScore

type PluginToClusterScore map[string]ClusterScoreList

type PreferredSchedulingTerm

type PreferredSchedulingTerm struct {
	// Weight associated with matching the corresponding nodeSelectorTerm, in the range 1-100.
	Weight int32 `json:"weight"`
	// A node selector term, associated with the corresponding weight.
	Preference fedcorev1a1.ClusterSelectorTerm `json:"preference"`
}

An empty preferred scheduling term matches all objects with implicit weight 0 (i.e. it's a no-op). A null preferred scheduling term matches no objects (i.e. is also a no-op).

type ReplicasMigration

type ReplicasMigration struct {
	Cluster string `json:"cluster"`
	// Describes the max number of replicas a cluster can accommodate.
	// Different from `maxReplicas` in policy, LimitedCapacity is not affected by `avoidDisruption`
	// and is a constraint that must be satisfied.
	LimitedCapacity *int64 `json:"limitedCapacity"`
}

type ReplicasPlugin

type ReplicasPlugin interface {
	Plugin

	ReplicaScheduling(context.Context, *SchedulingUnit, []*fedcorev1a1.FederatedCluster) (ClusterReplicasList, *Result)
}

type Resource

type Resource struct {
	MilliCPU         int64 `json:"millicpu"`
	Memory           int64 `json:"memory"`
	EphemeralStorage int64 `json:"ephemeralStorage"`
	// ScalarResources
	ScalarResources map[corev1.ResourceName]int64 `json:"scalarResources"`
}

Resource is a collection of compute resource.

func GetResourceRequest

func GetResourceRequest(pod *corev1.Pod) *Resource

func NewResource

func NewResource(rl corev1.ResourceList) *Resource

NewResource creates a Resource from ResourceList

func (*Resource) Add

func (r *Resource) Add(rl corev1.ResourceList)

Add adds ResourceList into Resource.

func (*Resource) AddScalar

func (r *Resource) AddScalar(name corev1.ResourceName, quantity int64)

AddScalar adds a resource by a scalar value of this resource.

func (*Resource) Clone

func (r *Resource) Clone() *Resource

Clone returns a copy of this resource.

func (*Resource) HasScalarResource

func (r *Resource) HasScalarResource(name corev1.ResourceName) bool

HasScalarResource checks if Resource has the given scalar resource.

func (*Resource) ResourceList

func (r *Resource) ResourceList() corev1.ResourceList

ResourceList returns a resource list of this resource.

func (*Resource) SetMaxResource

func (r *Resource) SetMaxResource(rl corev1.ResourceList)

SetMaxResource compares with ResourceList and takes max value for each Resource.

func (*Resource) SetScalar

func (r *Resource) SetScalar(name corev1.ResourceName, quantity int64)

SetScalar sets a resource by a scalar value of this resource.

func (*Resource) Sub

func (r *Resource) Sub(rl corev1.ResourceList) error

Sub is used to subtract two resources. Return error when the minuend is less than the subtrahend.

type ResourceName

type ResourceName string

type ResourceToValueMap

type ResourceToValueMap map[corev1.ResourceName]int64

type ResourceToWeightMap

type ResourceToWeightMap map[corev1.ResourceName]int64

type Result

type Result struct {
	// contains filtered or unexported fields
}

Result indicates the result of running a plugin. It consists of a code, a message and (optionally) an error. When the status code is not `Success`, the reasons should explain why.

func DefaultNormalizeScore

func DefaultNormalizeScore(maxPriority int64, reverse bool, scores ClusterScoreList) *Result

DefaultNormalizeScore generates a Normalize Score function that can normalize the scores to [0, maxPriority]. If reverse is set to true, it reverses the scores by subtracting it from maxPriority.

func NewResult

func NewResult(code Code, reasons ...string) *Result

NewResult makes a result out of the given arguments and returns its pointer.

func (*Result) AsError

func (s *Result) AsError() error

AsError returns nil if the Result is a success; otherwise returns an "error" object with a concatenated message on reasons of the Result.

func (*Result) Code

func (s *Result) Code() Code

Code returns code of the Result.

func (*Result) FailedPlugin

func (s *Result) FailedPlugin() string

FailedPlugin returns the failed plugin name.

func (*Result) IsSuccess

func (s *Result) IsSuccess() bool

IsSuccess returns true if and only if "Result" is nil or Code is "Success".

func (*Result) Message

func (s *Result) Message() string

Message returns a concatenated message on reasons of the Status.

func (*Result) Reasons

func (s *Result) Reasons() []string

Reasons returns reasons of the Result.

func (*Result) SetFailedPlugin

func (s *Result) SetFailedPlugin(plugin string)

SetFailedPlugin sets the given plugin name to s.failedPlugin.

func (*Result) WithFailedPlugin

func (s *Result) WithFailedPlugin(plugin string) *Result

WithFailedPlugin sets the given plugin name to s.failedPlugin, and returns the given result object.

type SchedulingUnit

type SchedulingUnit struct {
	GroupVersion schema.GroupVersion
	Kind         string
	Resource     string

	Namespace   string
	Name        string
	Labels      map[string]string
	Annotations map[string]string

	// Only care about the requests resources
	// TODO(all), limit resources, Best Effort resources
	// The Resource and total replica could be deliver by resource type
	// or define by annotations
	// Describes the schedule request
	DesiredReplicas *int64
	ResourceRequest Resource

	// Describes the current scheduling state
	CurrentClusters map[string]*int64
	AutoMigration   *AutoMigrationSpec
	CustomMigration CustomMigrationSpec

	// Controls the scheduling behavior
	SchedulingMode   fedcorev1a1.SchedulingMode
	ReplicasStrategy fedcorev1a1.ReplicasStrategy
	StickyCluster    bool
	AvoidDisruption  bool

	// Used to filter/select clusters
	ClusterSelector map[string]string
	ClusterNames    map[string]struct{}
	Affinity        *Affinity
	Tolerations     []corev1.Toleration
	MaxClusters     *int64
	MinReplicas     map[string]int64
	MaxReplicas     map[string]int64
	Weights         map[string]int64
	Priorities      map[string]int64

	EnableKatalystSupport bool
}

func (*SchedulingUnit) Key

func (s *SchedulingUnit) Key() string

type ScoreExtensions

type ScoreExtensions interface {
	// NormalizeScore is called for all cluster scores produced by the same plugin's "Score"
	// method. A successful run of NormalizeScore will update the scores list and return
	// a success status.
	NormalizeScore(context.Context, ClusterScoreList) *Result
}

ScoreExtensions is an interface for Score extended functionality.

type ScorePlugin

type ScorePlugin interface {
	Plugin

	// Score is called on each filtered Cluster. It must return success and an integer
	// indicating the rank of the cluster. All scoring plugins must return success or
	// the pod will be rejected.
	Score(context.Context, *SchedulingUnit, *fedcorev1a1.FederatedCluster) (int64, *Result)

	// ScoreExtensions returns a ScoreExtensions interface if it implements one, or nil if it does not.
	ScoreExtensions() ScoreExtensions
}

ScorePlugin is an interface that must be implemented by "Score" plugins to rank clusters that passed the filtering phase.

type SelectPlugin

type SelectPlugin interface {
	Plugin

	SelectClusters(context.Context, *SchedulingUnit, ClusterScoreList) (ClusterScoreList, *Result)
}

type UnavailableCluster

type UnavailableCluster struct {
	Cluster    string      `json:"cluster"`
	ValidUntil metav1.Time `json:"validUntil"`
}

type WorkloadMigration

type WorkloadMigration struct {
	Cluster string `json:"cluster"`
	// Describes the timestamp after which workloads will no longer be evicted from the cluster.
	ValidUntil *metav1.Time `json:"validUntil"`
}

Directories

Path Synopsis
plugins
rsp
TODO(all) - implement score for macroservice.
TODO(all) - implement score for macroservice.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL