contorller

Documentation

Index

• Constants
• type ETExtensionController
  • func NewETExtensionController(k8sClient *kubernetes.Clientset, ...) *ETExtensionController
  • func (pc *ETExtensionController) AddETJob(obj interface{})
  • func (pc *ETExtensionController) AddQueueUnit(obj interface{})
  • func (pc *ETExtensionController) DeleteETJob(obj interface{})
  • func (pc *ETExtensionController) DeleteQueueAnnotationInetJob(qu *v1alpha1.QueueUnit) error
  • func (pc *ETExtensionController) DeleteQueueUnit(obj interface{})
  • func (pc *ETExtensionController) Run(threadiness int, stopCh <-chan struct{}) error
  • func (pc *ETExtensionController) UpdateETJob(_, newObj interface{})
  • func (pc *ETExtensionController) UpdateQueueUnit(oldObj, newObj interface{})

Constants

const (
	// MaxRetries is the number of times a queue item will be retried before it is dropped out of the queue.
	// With the current rate-limiter in use (5ms*2^(maxRetries-1)) the following numbers represent the times
	// a queue item is going to be requeued:
	//
	// 1-10 retry times: 5ms, 10ms, 20ms, 40ms, 80ms, 160ms, 320ms, 640ms, 1.3s, 2.6s,
	// 11-20 retry times: 5.1s, 10.2s, 20.4s, 41s, 82s, 164s, 328s, 656s(11min), 1312s(21min), 2624s(43min)
	MaxRetries = 15
	// Suspend is a flag annotation for etJob to use the queueunit crd
	Suspend = "scheduling.x-k8s.io/suspend"
)

const (
	ConsumerRefKind       = "TrainingJob"
	ConsumerRefAPIVersion = "kai.alibabacloud.com/v1alpha1"
	// QuNameSuffix is the suffix of the queue unit name when create a new one.
	// In this way, different types of jobs with the same name will create different queue unit name.
	QuNameSuffix = "-pytorch-qu"
	Queuing      = "Queuing"
)

Types

type ETExtensionController

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

func NewETExtensionController

func NewETExtensionController(
	k8sClient *kubernetes.Clientset,
	queueInformer queueInformers.QueueUnitInformer,
	queueClient *queueversioned.Clientset,
	etJobInformer etjobinformers.TrainingJobInformer,
	etJobClient *etjobversioned.Clientset) *ETExtensionController

func (*ETExtensionController) AddETJob

func (pc *ETExtensionController) AddETJob(obj interface{})

func (*ETExtensionController) AddQueueUnit

func (pc *ETExtensionController) AddQueueUnit(obj interface{})

func (*ETExtensionController) DeleteETJob

func (pc *ETExtensionController) DeleteETJob(obj interface{})

func (*ETExtensionController) DeleteQueueAnnotationInetJob

func (pc *ETExtensionController) DeleteQueueAnnotationInetJob(qu *v1alpha1.QueueUnit) error

func (*ETExtensionController) DeleteQueueUnit

func (pc *ETExtensionController) DeleteQueueUnit(obj interface{})

func (*ETExtensionController) Run

func (pc *ETExtensionController) Run(threadiness int, stopCh <-chan struct{}) error

func (*ETExtensionController) UpdateETJob

func (pc *ETExtensionController) UpdateETJob(_, newObj interface{})

func (*ETExtensionController) UpdateQueueUnit

func (pc *ETExtensionController) UpdateQueueUnit(oldObj, newObj interface{})