scheduler

Documentation

Index

• Constants
• type AntidoteScheduler
  • func (s *AntidoteScheduler) PruneOrphanedNamespaces() error
  • func (s *AntidoteScheduler) PurgeOldLessons(sc ot.SpanContext) ([]string, error)
  • func (s *AntidoteScheduler) PurgeOldSessions(sc ot.SpanContext) error
  • func (s *AntidoteScheduler) Start() error
• type LessonHealthCheck
• type LessonHealthChecker
• type NetworkCrdClient

Constants

const JobBackoff = 3

JobBackoff controls the number of times a job is retried before we consider it failed. I **believe** this is actually equal to the number of "tries", not "retries". So think of this as the number of pods you expect to see if all of them failed. https://kubernetes.io/docs/concepts/workloads/controllers/job/

Types

type AntidoteScheduler

type AntidoteScheduler struct {
	KubeConfig    *rest.Config
	NC            *nats.Conn
	Config        config.AntidoteConfig
	Db            db.DataManager
	HealthChecker LessonHealthChecker

	// Client for interacting with normal Kubernetes resources
	Client kubernetes.Interface

	// Client for creating CRD defintions
	ClientExt kubernetesExt.Interface

	// Client for creating instances of our network CRD
	ClientCrd kubernetesCrd.Interface

	BuildInfo map[string]string
}

AntidoteScheduler is an Antidote service that receives commands for things like creating new lesson instances, moving existing livelessons to a different stage, deleting old lessons, etc.

func (*AntidoteScheduler) PruneOrphanedNamespaces

func (s *AntidoteScheduler) PruneOrphanedNamespaces() error

PruneOrphanedNamespaces seeks out all antidote-managed namespaces, and deletes them. This will effectively reset the cluster to a state with all of the remaining infrastructure in place, but no running lessons. Antidote doesn't manage itself, or any other Antidote services.

func (*AntidoteScheduler) PurgeOldLessons

func (s *AntidoteScheduler) PurgeOldLessons(sc ot.SpanContext) ([]string, error)

PurgeOldLessons identifies any kubernetes namespaces that are operating with our antidoteId, and among those, deletes the ones that have a lastAccessed timestamp that exceeds our configured TTL. This function is meant to be run in a loop within a goroutine, at a configured interval. Returns a slice of livelesson IDs to be deleted by the caller (not handled by this function)

func (*AntidoteScheduler) PurgeOldSessions

func (s *AntidoteScheduler) PurgeOldSessions(sc ot.SpanContext) error

PurgeOldSessions cleans up old LiveSessions according to the configured LiveSessionTTL

func (*AntidoteScheduler) Start

func (s *AntidoteScheduler) Start() error

Start is meant to be run as a goroutine. The "requests" channel will wait for new requests, attempt to schedule them, and put a results message on the "results" channel when finished (success or fail)

type LessonHealthCheck

type LessonHealthCheck struct{}

LessonHealthCheck performs network tests to determine health of endpoints within a running lesson

type LessonHealthChecker

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

LessonHealthChecker describes a struct which offers a variety of reachability tests for lesson endpoints.

type NetworkCrdClient

type NetworkCrdClient interface {
	UpdateNamespace(string)
	Create(obj *networkcrd.NetworkAttachmentDefinition) (*networkcrd.NetworkAttachmentDefinition, error)
	Update(obj *networkcrd.NetworkAttachmentDefinition) (*networkcrd.NetworkAttachmentDefinition, error)
	Delete(name string, options *meta_v1.DeleteOptions) error
	Get(name string) (*networkcrd.NetworkAttachmentDefinition, error)
	List(opts meta_v1.ListOptions) (*networkcrd.NetworkList, error)
}

NetworkCrdClient is an interface for the client for our custom network CRD. Allows for injection of mocks at test time.