controller

Documentation

Overview

Package controller contains logic for watching and synchronizing replicationControllers.

Index

• Constants
• type Expectations
• type PodControlInterface
• type PodExpectations
  • func (e *PodExpectations) Fulfilled() bool
  • func (e *PodExpectations) Seen(add, del int64)
• type RCExpectations
  • func NewRCExpectations() *RCExpectations
  • func (r *RCExpectations) CreationObserved(rc *api.ReplicationController)
  • func (r *RCExpectations) DeletionObserved(rc *api.ReplicationController)
  • func (r *RCExpectations) ExpectCreations(rc *api.ReplicationController, adds int) error
  • func (r *RCExpectations) ExpectDeletions(rc *api.ReplicationController, dels int) error
  • func (r *RCExpectations) GetExpectations(rc *api.ReplicationController) (*PodExpectations, bool, error)
  • func (r *RCExpectations) SatisfiedExpectations(rc *api.ReplicationController) bool
• type RCExpectationsManager
• type RealPodControl
• type ReplicationManager
  • func NewReplicationManager(kubeClient client.Interface, burstReplicas int) *ReplicationManager
  • func (rm *ReplicationManager) Run(workers int, stopCh <-chan struct{})
  • func (rm *ReplicationManager) SetEventRecorder(recorder record.EventRecorder)

Constants

const (
	// We'll attempt to recompute the required replicas of all replication controllers
	// the have fulfilled their expectations at least this often. This recomputation
	// happens based on contents in local pod storage.
	FullControllerResyncPeriod = 30 * time.Second

	// If a watch misdelivers info about a pod, it'll take at least this long
	// to rectify the number of replicas. Note that dropped deletes are only
	// rectified after the expectation times out because we don't know the
	// final resting state of the pod.
	PodRelistPeriod = 5 * time.Minute

	// If a watch drops a delete event for a pod, it'll take this long
	// before a dormant rc waiting for those packets is woken up anyway. It is
	// specifically targeted at the case where some problem prevents an update
	// of expectations, without it the RC could stay asleep forever. This should
	// be set based on the expected latency of watch events.
	//
	// Currently an rc can service (create *and* observe the watch events for said
	// creation) about 10-20 pods a second, so it takes about 1 min to service
	// 500 pods. Just creation is limited to 20qps, and watching happens with ~10-30s
	// latency/pod at the scale of 3000 pods over 100 nodes.
	ExpectationsTimeout = 3 * time.Minute

	// Realistic value of the burstReplica field for the replication manager based off
	// performance requirements for kubernetes 1.0.
	BurstReplicas = 500
)

const (
	CreatedByAnnotation = "kubernetes.io/created-by"
)

Types

type Expectations

type Expectations interface {
	Fulfilled() bool
}

Expectations are either fulfilled, or expire naturally.

type PodControlInterface

type PodControlInterface interface {
	// contains filtered or unexported methods
}

PodControlInterface is an interface that knows how to add or delete pods created as an interface to allow testing.

type PodExpectations

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

PodExpectations track pod creates/deletes.

func (*PodExpectations) Fulfilled

func (e *PodExpectations) Fulfilled() bool

Fulfilled returns true if this expectation has been fulfilled.

func (*PodExpectations) Seen

func (e *PodExpectations) Seen(add, del int64)

Seen decrements the add and del counters.

type RCExpectations

type RCExpectations struct {
	cache.Store
}

RCExpectations is a ttl cache mapping rcs to what they expect to see before being woken up for a sync.

func NewRCExpectations

func NewRCExpectations() *RCExpectations

NewRCExpectations returns a store for PodExpectations.

func (*RCExpectations) CreationObserved

func (r *RCExpectations) CreationObserved(rc *api.ReplicationController)

CreationObserved atomically decrements the `add` expecation count of the given replication controller.

func (*RCExpectations) DeletionObserved

func (r *RCExpectations) DeletionObserved(rc *api.ReplicationController)

DeletionObserved atomically decrements the `del` expectation count of the given replication controller.

func (*RCExpectations) ExpectCreations

func (r *RCExpectations) ExpectCreations(rc *api.ReplicationController, adds int) error

func (*RCExpectations) ExpectDeletions

func (r *RCExpectations) ExpectDeletions(rc *api.ReplicationController, dels int) error

func (*RCExpectations) GetExpectations

func (r *RCExpectations) GetExpectations(rc *api.ReplicationController) (*PodExpectations, bool, error)

GetExpectations returns the PodExpectations of the given rc.

func (*RCExpectations) SatisfiedExpectations

func (r *RCExpectations) SatisfiedExpectations(rc *api.ReplicationController) bool

SatisfiedExpectations returns true if the replication manager has observed the required adds/dels for the given rc. Add/del counts are established by the rc at sync time, and updated as pods are observed by the replication manager's podController.

type RCExpectationsManager

type RCExpectationsManager interface {
	GetExpectations(rc *api.ReplicationController) (*PodExpectations, bool, error)
	SatisfiedExpectations(rc *api.ReplicationController) bool
	ExpectCreations(rc *api.ReplicationController, adds int) error
	ExpectDeletions(rc *api.ReplicationController, dels int) error
	CreationObserved(rc *api.ReplicationController)
	DeletionObserved(rc *api.ReplicationController)
}

RCExpectationsManager is an interface that allows users to set and wait on expectations. Only abstracted out for testing.

type RealPodControl

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

RealPodControl is the default implementation of PodControllerInterface.

type ReplicationManager

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

ReplicationManager is responsible for synchronizing ReplicationController objects stored in the system with actual running pods.

func NewReplicationManager

func NewReplicationManager(kubeClient client.Interface, burstReplicas int) *ReplicationManager

NewReplicationManager creates a new ReplicationManager.

func (*ReplicationManager) Run

func (rm *ReplicationManager) Run(workers int, stopCh <-chan struct{})

Run begins watching and syncing.

func (*ReplicationManager) SetEventRecorder

func (rm *ReplicationManager) SetEventRecorder(recorder record.EventRecorder)

SetEventRecorder replaces the event recorder used by the replication manager with the given recorder. Only used for testing.