scheduler

Documentation

Index

• Constants
• func Start(logger *zap.Logger, port int, psUrl string)
• type Scheduler
  • func (s *Scheduler) GetHandler() http.Handler
  • func (s *Scheduler) Serve(port int)
• type SchedulerPolicy
• type SchedulerQueue
  • func NewQueue() SchedulerQueue
• type TaskOperation
• type ThroughputBasedPolicy

Constants

const (
	ThroughPutScaleDownThreshold = 1.2
	ThroughputScaleUpThreshold   = 1.05
)

Functions

func Start

func Start(logger *zap.Logger, port int, psUrl string)

Start starts all of the goroutines that will take care of the proper functioning of the scheduler 1) Find next parallelism 2) Fetch metrics 3) **maybe look for the failed tasks queue? If the router fails keep the task there and retry 4) API so the scheduler is reachable from the other components

Types

type Scheduler

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

func (*Scheduler) GetHandler

func (s *Scheduler) GetHandler() http.Handler

Create the handler for the scheduler to receive requests from the API

func (*Scheduler) Serve

func (s *Scheduler) Serve(port int)

Expose the API

type SchedulerPolicy

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

SchedulerPolicy defines the methods needed to be implemented by the scheduler in order to support the tasks from KubeML, these involve how to calculate the parallelism of the task based on previous performance of the task

type SchedulerQueue

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

SchedulerQueue is the queue that will be used for the scheduler

func NewQueue

func NewQueue() SchedulerQueue

NewQueue creates a queue for the scheduler

type TaskOperation

type TaskOperation int

const (
	CreateTask TaskOperation = iota
	UpdateTask
)

Declare the task operation

type ThroughputBasedPolicy

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

SchedulerPolicy defines the methods needed to be implemented by the scheduler in order to support the tasks from KubeML, these involve how to calculate the parallelism of the task based on previous performance of the task