v1

package
v0.1.1 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Jan 16, 2020 License: Apache-2.0 Imports: 19 Imported by: 0

Documentation

Overview

Package v1 is the v1 version of the API. +groupName=kubeflow.org

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at 

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Index

Constants

View Source 
const (
	// ReplicaIndexLabel represents the label key for the replica-index, e.g. the value is 0, 1, 2.. etc
	ReplicaIndexLabel = "replica-index"

	// ReplicaTypeLabel represents the label key for the replica-type, e.g. the value is ps , worker etc.
	ReplicaTypeLabel = "replica-type"

	// GroupNameLabel represents the label key for group name, e.g. the value is kubeflow.org
	GroupNameLabel = "group-name"

	// JobNameLabel represents the label key for the job name, the value is job name
	JobNameLabel = "job-name"

	// JobRoleLabel represents the label key for the job role, e.g. the value is master
	JobRoleLabel = "job-role"
)

Variables

This section is empty.

Functions

func GetOpenAPIDefinitions 

func GetOpenAPIDefinitions(ref common.ReferenceCallback) map[string]common.OpenAPIDefinition

func RegisterDefaults 

func RegisterDefaults(scheme *runtime.Scheme) error

RegisterDefaults adds defaulters functions to the given scheme. Public to allow building arbitrary schemes. All generated defaulters are covering - they call all nested defaulters.

Types

type CleanPodPolicy 

type CleanPodPolicy string

+k8s:openapi-gen=true CleanPodPolicy describes how to deal with pods when the job is finished. 

const (
	CleanPodPolicyUndefined CleanPodPolicy = ""
	CleanPodPolicyAll       CleanPodPolicy = "All"
	CleanPodPolicyRunning   CleanPodPolicy = "Running"
	CleanPodPolicyNone      CleanPodPolicy = "None"
)

type ControllerInterface 

type ControllerInterface interface {
	// Returns the Controller name
	ControllerName() string

	// Returns the GroupVersionKind of the API
	GetAPIGroupVersionKind() schema.GroupVersionKind

	// Returns the GroupVersion of the API
	GetAPIGroupVersion() schema.GroupVersion

	// Returns the Group Name(value) in the labels of the job
	GetGroupNameLabelValue() string

	// Returns the Job from Informer Cache
	GetJobFromInformerCache(namespace, name string) (v1.Object, error)

	// Returns the Job from API server
	GetJobFromAPIClient(namespace, name string) (v1.Object, error)

	// GetPodsForJob returns the pods managed by the job. This can be achieved by selecting pods using label key "job-name"
	// i.e. all pods created by the job will come with label "job-name" = <this_job_name>
	GetPodsForJob(job interface{}) ([]*corev1.Pod, error)

	// GetServicesForJob returns the services managed by the job. This can be achieved by selecting services using label key "job-name"
	// i.e. all services created by the job will come with label "job-name" = <this_job_name>
	GetServicesForJob(job interface{}) ([]*corev1.Service, error)

	// DeleteJob deletes the job
	DeleteJob(job interface{}) error

	// UpdateJobStatus updates the job status and job conditions
	UpdateJobStatus(job interface{}, replicas map[ReplicaType]*ReplicaSpec, jobStatus *JobStatus) error

	// UpdateJobStatusInApiServer updates the job status in API server
	UpdateJobStatusInApiServer(job interface{}, jobStatus *JobStatus) error

	// CreateService creates the service
	CreateService(job interface{}, service *corev1.Service) error

	// DeleteService deletes the service
	DeleteService(job interface{}, name string, namespace string) error

	// CreatePod creates the pod
	CreatePod(job interface{}, pod *corev1.Pod) error

	// DeletePod deletes the pod
	DeletePod(job interface{}, pod *corev1.Pod) error

	// SetClusterSpec sets the cluster spec for the pod
	SetClusterSpec(job interface{}, podTemplate *corev1.PodTemplateSpec, rtype, index string) error

	// Returns the default container name in pod
	GetDefaultContainerName() string

	// Get the default container port name
	GetDefaultContainerPortName() string

	// Get the default container port number
	GetDefaultContainerPortNumber() int32

	// Returns if this replica type with index specified is a master role.
	// MasterRole pod will have "job-role=master" set in its label
	IsMasterRole(replicas map[ReplicaType]*ReplicaSpec, rtype ReplicaType, index int) bool
}

ControllerInterface defines the Interface to be implemented by custom operators. e.g. tf-operator needs to implement this interface

type JobCondition 

type JobCondition struct {
	// Type of job condition.
	Type JobConditionType `json:"type"`
	// Status of the condition, one of True, False, Unknown.
	Status v1.ConditionStatus `json:"status"`
	// The reason for the condition's last transition.
	Reason string `json:"reason,omitempty"`
	// A human readable message indicating details about the transition.
	Message string `json:"message,omitempty"`
	// The last time this condition was updated.
	LastUpdateTime metav1.Time `json:"lastUpdateTime,omitempty"`
	// Last time the condition transitioned from one status to another.
	LastTransitionTime metav1.Time `json:"lastTransitionTime,omitempty"`
}

+k8s:openapi-gen=true +k8s:deepcopy-gen=true JobCondition describes the state of the job at a certain point.

func (*JobCondition) DeepCopy 

func (in *JobCondition) DeepCopy() *JobCondition

DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new JobCondition.

func (*JobCondition) DeepCopyInto 

func (in *JobCondition) DeepCopyInto(out *JobCondition)

DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.

type JobConditionType 

type JobConditionType string

JobConditionType defines all kinds of types of JobStatus. 

const (
	// JobCreated means the job has been accepted by the system,
	// but one or more of the pods/services has not been started.
	// This includes time before pods being scheduled and launched.
	JobCreated JobConditionType = "Created"

	// JobRunning means all sub-resources (e.g. services/pods) of this job
	// have been successfully scheduled and launched.
	// The training is running without error.
	JobRunning JobConditionType = "Running"

	// JobRestarting means one or more sub-resources (e.g. services/pods) of this job
	// reached phase failed but maybe restarted according to it's restart policy
	// which specified by user in v1.PodTemplateSpec.
	// The training is freezing/pending.
	JobRestarting JobConditionType = "Restarting"

	// JobSucceeded means all sub-resources (e.g. services/pods) of this job
	// reached phase have terminated in success.
	// The training is complete without error.
	JobSucceeded JobConditionType = "Succeeded"

	// JobFailed means one or more sub-resources (e.g. services/pods) of this job
	// reached phase failed with no restarting.
	// The training has failed its execution.
	JobFailed JobConditionType = "Failed"
)

type JobController 

type JobController struct {
	Controller ControllerInterface

	Config JobControllerConfiguration

	// KubeClientSet is a standard kubernetes clientset.
	KubeClientSet kubeclientset.Interface

	// KubeBatchClientSet is a standard kube-batch clientset.
	KubeBatchClientSet kubebatchclient.Interface

	// PodLister can list/get pods from the shared informer's store.
	PodLister corelisters.PodLister

	// ServiceLister can list/get services from the shared informer's store.
	ServiceLister corelisters.ServiceLister

	// PodInformerSynced returns true if the pod store has been synced at least once.
	PodInformerSynced cache.InformerSynced

	// ServiceInformerSynced returns true if the service store has been synced at least once.
	ServiceInformerSynced cache.InformerSynced

	// A TTLCache of pod/services creates/deletes each job expects to see
	// We use Job namespace/name + ReplicaType + pods/services as an expectation key,
	// For example, there is a TFJob with namespace "tf-operator" and name "tfjob-abc":
	// {
	//     "PS": {
	//         "Replicas": 2,
	//     },
	//     "Worker": {
	//         "Replicas": 4,
	//     }
	// }
	// We will create 4 expectations:
	// - "tf-operator/tfjob-abc/ps/services", expects 2 adds.
	// - "tf-operator/tfjob-abc/ps/pods", expects 2 adds.
	// - "tf-operator/tfjob-abc/worker/services", expects 4 adds.
	// - "tf-operator/tfjob-abc/worker/pods", expects 4 adds.
	Expectations controller.ControllerExpectationsInterface

	// WorkQueue is a rate limited work queue. This is used to queue work to be
	// processed instead of performing it as soon as a change happens. This
	// means we can ensure we only process a fixed amount of resources at a
	// time, and makes it easy to ensure we are never processing the same item
	// simultaneously in two different workers.
	WorkQueue workqueue.RateLimitingInterface

	// Recorder is an event recorder for recording Event resources to the
	// Kubernetes API.
	Recorder record.EventRecorder
}

JobController abstracts other operators to manage the lifecycle of Jobs. User need to first implement the ControllerInterface(objectA) and then initialize a JobController(objectB) struct with objectA as the parameter. And then call objectB.ReconcileJobs as mentioned below, the ReconcileJobs method is the entrypoint to trigger the reconcile logic of the job controller

ReconcileJobs( 

job interface{},
replicas map[apiv1.ReplicaType]*apiv1.ReplicaSpec,
jobStatus apiv1.JobStatus,
runPolicy *apiv1.RunPolicy) error

func NewJobController 

func NewJobController(
	controllerImpl ControllerInterface,
	reconcilerSyncPeriod metav1.Duration,
	enableGangScheduling bool,
	kubeClientSet kubeclientset.Interface,
	kubeBatchClientSet kubebatchclient.Interface,
	kubeInformerFactory kubeinformers.SharedInformerFactory,
	workQueueName string) JobController

type JobControllerConfiguration 

type JobControllerConfiguration struct {
	// ReconcilerSyncLoopPeriod is the amount of time the reconciler sync states loop
	// wait between two reconciler sync.
	// It is set to 15 sec by default.
	// TODO(cph): maybe we can let it grows by multiple in the future
	// and up to 5 minutes to reduce idle loop.
	// e.g. 15s, 30s, 60s, 120s...
	ReconcilerSyncLoopPeriod metav1.Duration

	// Enable gang scheduling by kube-batch
	EnableGangScheduling bool
}

JobControllerConfiguration contains configuration of operator.

type JobStatus 

type JobStatus struct {
	// Conditions is an array of current observed job conditions.
	Conditions []JobCondition `json:"conditions"`

	// ReplicaStatuses is map of ReplicaType and ReplicaStatus,
	// specifies the status of each replica.
	ReplicaStatuses map[ReplicaType]*ReplicaStatus `json:"replicaStatuses"`

	// Represents time when the job was acknowledged by the job controller.
	// It is not guaranteed to be set in happens-before order across separate operations.
	// It is represented in RFC3339 form and is in UTC.
	StartTime *metav1.Time `json:"startTime,omitempty"`

	// Represents time when the job was completed. It is not guaranteed to
	// be set in happens-before order across separate operations.
	// It is represented in RFC3339 form and is in UTC.
	CompletionTime *metav1.Time `json:"completionTime,omitempty"`

	// Represents last time when the job was reconciled. It is not guaranteed to
	// be set in happens-before order across separate operations.
	// It is represented in RFC3339 form and is in UTC.
	LastReconcileTime *metav1.Time `json:"lastReconcileTime,omitempty"`
}

+k8s:openapi-gen=true +k8s:deepcopy-gen=true JobStatus represents the current observed state of the training Job.

func (*JobStatus) DeepCopy 

func (in *JobStatus) DeepCopy() *JobStatus

DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new JobStatus.

func (*JobStatus) DeepCopyInto 

func (in *JobStatus) DeepCopyInto(out *JobStatus)

DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.

type ReplicaSpec 

type ReplicaSpec struct {
	// Replicas is the desired number of replicas of the given template.
	// If unspecified, defaults to 1.
	Replicas *int32 `json:"replicas,omitempty"`

	// Template is the object that describes the pod that
	// will be created for this replica. RestartPolicy in PodTemplateSpec
	// will be overide by RestartPolicy in ReplicaSpec
	Template v1.PodTemplateSpec `json:"template,omitempty"`

	// Restart policy for all replicas within the job.
	// One of Always, OnFailure, Never and ExitCode.
	// Default to Never.
	RestartPolicy RestartPolicy `json:"restartPolicy,omitempty"`
}

+k8s:openapi-gen=true +k8s:deepcopy-gen=true ReplicaSpec is a description of the replica

func (*ReplicaSpec) DeepCopy 

func (in *ReplicaSpec) DeepCopy() *ReplicaSpec

DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new ReplicaSpec.

func (*ReplicaSpec) DeepCopyInto 

func (in *ReplicaSpec) DeepCopyInto(out *ReplicaSpec)

DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.

type ReplicaStatus 

type ReplicaStatus struct {
	// The number of actively running pods.
	Active int32 `json:"active,omitempty"`

	// The number of pods which reached phase Succeeded.
	Succeeded int32 `json:"succeeded,omitempty"`

	// The number of pods which reached phase Failed.
	Failed int32 `json:"failed,omitempty"`
}

+k8s:openapi-gen=true ReplicaStatus represents the current observed state of the replica.

func (*ReplicaStatus) DeepCopy 

func (in *ReplicaStatus) DeepCopy() *ReplicaStatus

DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new ReplicaStatus.

func (*ReplicaStatus) DeepCopyInto 

func (in *ReplicaStatus) DeepCopyInto(out *ReplicaStatus)

DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.

type ReplicaType 

type ReplicaType string

+k8s:openapi-gen=true ReplicaType represents the type of the replica. Each operator needs to define its own set of ReplicaTypes.

type RestartPolicy 

type RestartPolicy string

RestartPolicy describes how the replicas should be restarted. Only one of the following restart policies may be specified. If none of the following policies is specified, the default one is RestartPolicyAlways. 

const (
	RestartPolicyAlways    RestartPolicy = "Always"
	RestartPolicyOnFailure RestartPolicy = "OnFailure"
	RestartPolicyNever     RestartPolicy = "Never"

	// RestartPolicyExitCode policy means that user should add exit code by themselves,
	// The job operator will check these exit codes to
	// determine the behavior when an error occurs:
	// - 1-127: permanent error, do not restart.
	// - 128-255: retryable error, will restart the pod.
	RestartPolicyExitCode RestartPolicy = "ExitCode"
)

type RunPolicy 

type RunPolicy struct {
	// CleanPodPolicy defines the policy to kill pods after the job completes.
	// Default to Running.
	CleanPodPolicy *CleanPodPolicy `json:"cleanPodPolicy,omitempty"`

	// TTLSecondsAfterFinished is the TTL to clean up jobs.
	// It may take extra ReconcilePeriod seconds for the cleanup, since
	// reconcile gets called periodically.
	// Default to infinite.
	TTLSecondsAfterFinished *int32 `json:"ttlSecondsAfterFinished,omitempty"`

	// Specifies the duration in seconds relative to the startTime that the job may be active
	// before the system tries to terminate it; value must be positive integer.
	// +optional
	ActiveDeadlineSeconds *int64 `json:"activeDeadlineSeconds,omitempty"`

	// Optional number of retries before marking this job failed.
	// +optional
	BackoffLimit *int32 `json:"backoffLimit,omitempty"`

	// SchedulingPolicy defines the policy related to scheduling, e.g. gang-scheduling
	// +optional
	SchedulingPolicy *SchedulingPolicy `json:"schedulingPolicy,omitempty"`
}

+k8s:deepcopy-gen=true RunPolicy encapsulates various runtime policies of the distributed training job, for example how to clean up resources and how long the job can stay active.

func (*RunPolicy) DeepCopy 

func (in *RunPolicy) DeepCopy() *RunPolicy

DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new RunPolicy.

func (*RunPolicy) DeepCopyInto 

func (in *RunPolicy) DeepCopyInto(out *RunPolicy)

DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.

type SchedulingPolicy 

type SchedulingPolicy struct {
	MinAvailable *int32 `json:"minAvailable,omitempty"`
}

SchedulingPolicy encapsulates various scheduling policies of the distributed training job, for example `minAvailable` for gang-scheduling.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.