scheduling

Documentation

Index

• Constants
• Variables
• func RegisterVolumeSchedulingMetrics()
• type AssumeCache
  • func NewAssumeCache(informer cache.SharedIndexInformer, description, indexName string, ...) AssumeCache
• type FakeVolumeBinder
  • func NewFakeVolumeBinder(config *FakeVolumeBinderConfig) *FakeVolumeBinder
  • func (b *FakeVolumeBinder) AssumePodVolumes(assumedPod *v1.Pod, nodeName string) (bool, error)
  • func (b *FakeVolumeBinder) BindPodVolumes(assumedPod *v1.Pod) error
  • func (b *FakeVolumeBinder) FindPodVolumes(pod *v1.Pod, node *v1.Node) (unboundVolumesSatisfied, boundVolumesSatsified bool, err error)
  • func (b *FakeVolumeBinder) GetBindingsCache() PodBindingCache
• type FakeVolumeBinderConfig
• type PVAssumeCache
  • func NewPVAssumeCache(informer cache.SharedIndexInformer) PVAssumeCache
• type PVCAssumeCache
  • func NewPVCAssumeCache(informer cache.SharedIndexInformer) PVCAssumeCache
• type PodBindingCache
  • func NewPodBindingCache() PodBindingCache
• type SchedulerVolumeBinder
  • func NewVolumeBinder(kubeClient clientset.Interface, nodeInformer coreinformers.NodeInformer, ...) SchedulerVolumeBinder

Constants

const VolumeSchedulerSubsystem = "scheduler_volume"

VolumeSchedulerSubsystem - subsystem name used by scheduler

Variables

var (
	// VolumeBindingRequestSchedulerBinderCache tracks the number of volume binder cache operations.
	VolumeBindingRequestSchedulerBinderCache = prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Subsystem: VolumeSchedulerSubsystem,
			Name:      "binder_cache_requests_total",
			Help:      "Total number for request volume binding cache",
		},
		[]string{"operation"},
	)
	// VolumeSchedulingStageLatency tracks the latency of volume scheduling operations.
	VolumeSchedulingStageLatency = prometheus.NewHistogramVec(
		prometheus.HistogramOpts{
			Subsystem: VolumeSchedulerSubsystem,
			Name:      "scheduling_duration_seconds",
			Help:      "Volume scheduling stage latency",
			Buckets:   prometheus.ExponentialBuckets(1000, 2, 15),
		},
		[]string{"operation"},
	)
	// VolumeSchedulingStageFailed tracks the number of failed volume scheduling operations.
	VolumeSchedulingStageFailed = prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Subsystem: VolumeSchedulerSubsystem,
			Name:      "scheduling_stage_error_total",
			Help:      "Volume scheduling stage error count",
		},
		[]string{"operation"},
	)
)

Functions

func RegisterVolumeSchedulingMetrics

func RegisterVolumeSchedulingMetrics()

RegisterVolumeSchedulingMetrics is used for scheduler, because the volume binding cache is a library used by scheduler process.

Types

type AssumeCache

type AssumeCache interface {
	// Assume updates the object in-memory only
	Assume(obj interface{}) error

	// Restore the informer cache's version of the object
	Restore(objName string)

	// Get the object by name
	Get(objName string) (interface{}, error)

	// Get the API object by name
	GetAPIObj(objName string) (interface{}, error)

	// List all the objects in the cache
	List(indexObj interface{}) []interface{}
}

AssumeCache is a cache on top of the informer that allows for updating objects outside of informer events and also restoring the informer cache's version of the object. Objects are assumed to be Kubernetes API objects that implement meta.Interface

func NewAssumeCache

func NewAssumeCache(informer cache.SharedIndexInformer, description, indexName string, indexFunc cache.IndexFunc) AssumeCache

NewAssumeCache creates an assume cache for genernal objects.

type FakeVolumeBinder

type FakeVolumeBinder struct {
	AssumeCalled bool
	BindCalled   bool
	// contains filtered or unexported fields
}

FakeVolumeBinder represents a fake volume binder for testing.

func NewFakeVolumeBinder

func NewFakeVolumeBinder(config *FakeVolumeBinderConfig) *FakeVolumeBinder

NewFakeVolumeBinder sets up all the caches needed for the scheduler to make topology-aware volume binding decisions.

func (*FakeVolumeBinder) AssumePodVolumes

func (b *FakeVolumeBinder) AssumePodVolumes(assumedPod *v1.Pod, nodeName string) (bool, error)

AssumePodVolumes implements SchedulerVolumeBinder.AssumePodVolumes.

func (*FakeVolumeBinder) BindPodVolumes

func (b *FakeVolumeBinder) BindPodVolumes(assumedPod *v1.Pod) error

BindPodVolumes implements SchedulerVolumeBinder.BindPodVolumes.

func (*FakeVolumeBinder) FindPodVolumes

func (b *FakeVolumeBinder) FindPodVolumes(pod *v1.Pod, node *v1.Node) (unboundVolumesSatisfied, boundVolumesSatsified bool, err error)

FindPodVolumes implements SchedulerVolumeBinder.FindPodVolumes.

func (*FakeVolumeBinder) GetBindingsCache

func (b *FakeVolumeBinder) GetBindingsCache() PodBindingCache

GetBindingsCache implements SchedulerVolumeBinder.GetBindingsCache.

type FakeVolumeBinderConfig

type FakeVolumeBinderConfig struct {
	AllBound             bool
	FindUnboundSatsified bool
	FindBoundSatsified   bool
	FindErr              error
	AssumeErr            error
	BindErr              error
}

FakeVolumeBinderConfig holds configurations for fake volume binder.

type PVAssumeCache

type PVAssumeCache interface {
	AssumeCache

	GetPV(pvName string) (*v1.PersistentVolume, error)
	GetAPIPV(pvName string) (*v1.PersistentVolume, error)
	ListPVs(storageClassName string) []*v1.PersistentVolume
}

PVAssumeCache is a AssumeCache for PersistentVolume objects

func NewPVAssumeCache

func NewPVAssumeCache(informer cache.SharedIndexInformer) PVAssumeCache

NewPVAssumeCache creates a PV assume cache.

type PVCAssumeCache

type PVCAssumeCache interface {
	AssumeCache

	// GetPVC returns the PVC from the cache with given pvcKey.
	// pvcKey is the result of MetaNamespaceKeyFunc on PVC obj
	GetPVC(pvcKey string) (*v1.PersistentVolumeClaim, error)
	GetAPIPVC(pvcKey string) (*v1.PersistentVolumeClaim, error)
}

PVCAssumeCache is a AssumeCache for PersistentVolumeClaim objects

func NewPVCAssumeCache

func NewPVCAssumeCache(informer cache.SharedIndexInformer) PVCAssumeCache

NewPVCAssumeCache creates a PVC assume cache.

type PodBindingCache

type PodBindingCache interface {
	// UpdateBindings will update the cache with the given bindings for the
	// pod and node.
	UpdateBindings(pod *v1.Pod, node string, bindings []*bindingInfo, provisionings []*v1.PersistentVolumeClaim)

	// ClearBindings will clear the cached bindings for the given pod and node.
	ClearBindings(pod *v1.Pod, node string)

	// GetBindings will return the cached bindings for the given pod and node.
	// A nil return value means that the entry was not found. An empty slice
	// means that no binding operations are needed.
	GetBindings(pod *v1.Pod, node string) []*bindingInfo

	// A nil return value means that the entry was not found. An empty slice
	// means that no provisioning operations are needed.
	GetProvisionedPVCs(pod *v1.Pod, node string) []*v1.PersistentVolumeClaim

	// GetDecisions will return all cached decisions for the given pod.
	GetDecisions(pod *v1.Pod) nodeDecisions

	// DeleteBindings will remove all cached bindings and provisionings for the given pod.
	// TODO: separate the func if it is needed to delete bindings/provisionings individually
	DeleteBindings(pod *v1.Pod)
}

PodBindingCache stores PV binding decisions per pod per node. Pod entries are removed when the Pod is deleted or updated to no longer be schedulable.

func NewPodBindingCache

func NewPodBindingCache() PodBindingCache

NewPodBindingCache creates a pod binding cache.

type SchedulerVolumeBinder

type SchedulerVolumeBinder interface {
	// FindPodVolumes checks if all of a Pod's PVCs can be satisfied by the node.
	//
	// If a PVC is bound, it checks if the PV's NodeAffinity matches the Node.
	// Otherwise, it tries to find an available PV to bind to the PVC.
	//
	// It returns true if all of the Pod's PVCs have matching PVs or can be dynamic provisioned,
	// and returns true if bound volumes satisfy the PV NodeAffinity.
	//
	// This function is called by the volume binding scheduler predicate and can be called in parallel
	FindPodVolumes(pod *v1.Pod, node *v1.Node) (unboundVolumesSatisified, boundVolumesSatisfied bool, err error)

	// AssumePodVolumes will:
	// 1. Take the PV matches for unbound PVCs and update the PV cache assuming
	// that the PV is prebound to the PVC.
	// 2. Take the PVCs that need provisioning and update the PVC cache with related
	// annotations set.
	//
	// It returns true if all volumes are fully bound
	//
	// This function will modify assumedPod with the node name.
	// This function is called serially.
	AssumePodVolumes(assumedPod *v1.Pod, nodeName string) (allFullyBound bool, err error)

	// BindPodVolumes will:
	// 1. Initiate the volume binding by making the API call to prebind the PV
	// to its matching PVC.
	// 2. Trigger the volume provisioning by making the API call to set related
	// annotations on the PVC
	// 3. Wait for PVCs to be completely bound by the PV controller
	//
	// This function can be called in parallel.
	BindPodVolumes(assumedPod *v1.Pod) error

	// GetBindingsCache returns the cache used (if any) to store volume binding decisions.
	GetBindingsCache() PodBindingCache
}

SchedulerVolumeBinder is used by the scheduler to handle PVC/PV binding and dynamic provisioning. The binding decisions are integrated into the pod scheduling workflow so that the PV NodeAffinity is also considered along with the pod's other scheduling requirements.

This integrates into the existing default scheduler workflow as follows:

1. The scheduler takes a Pod off the scheduler queue and processes it serially: a. Invokes all predicate functions, parallelized across nodes. FindPodVolumes() is invoked here. b. Invokes all priority functions. Future/TBD c. Selects the best node for the Pod. d. Cache the node selection for the Pod. AssumePodVolumes() is invoked here. i. If PVC binding is required, cache in-memory only: * For manual binding: update PV objects for prebinding to the corresponding PVCs. * For dynamic provisioning: update PVC object with a selected node from c) * For the pod, which PVCs and PVs need API updates. ii. Afterwards, the main scheduler caches the Pod->Node binding in the scheduler's pod cache, This is handled in the scheduler and not here. e. Asynchronously bind volumes and pod in a separate goroutine i. BindPodVolumes() is called first. It makes all the necessary API updates and waits for PV controller to fully bind and provision the PVCs. If binding fails, the Pod is sent back through the scheduler. ii. After BindPodVolumes() is complete, then the scheduler does the final Pod->Node binding.
2. Once all the assume operations are done in d), the scheduler processes the next Pod in the scheduler queue while the actual binding operation occurs in the background.

func NewVolumeBinder

func NewVolumeBinder(
	kubeClient clientset.Interface,
	nodeInformer coreinformers.NodeInformer,
	pvcInformer coreinformers.PersistentVolumeClaimInformer,
	pvInformer coreinformers.PersistentVolumeInformer,
	storageClassInformer storageinformers.StorageClassInformer,
	bindTimeout time.Duration) SchedulerVolumeBinder

NewVolumeBinder sets up all the caches needed for the scheduler to make volume binding decisions.