v1

package
v1.33.0 Latest Latest
Warning

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

 Go to latest Published: May 31, 2025 License: Apache-2.0 Imports: 9 Imported by: 0

Documentation

Overview

Package v1 contains API Schema definitions for the tensor-fusion.ai v1 API group. +kubebuilder:object:generate=true +groupName=tensor-fusion.ai

Index

Constants

View Source 
const (
	TensorFusionPoolPhasePending    = TensorFusionPoolPhase(constants.PhasePending)
	TensorFusionPoolPhaseRunning    = TensorFusionPoolPhase(constants.PhaseRunning)
	TensorFusionPoolPhaseUpdating   = TensorFusionPoolPhase(constants.PhaseUpdating)
	TensorFusionPoolPhaseUnknown    = TensorFusionPoolPhase(constants.PhaseUnknown)
	TensorFusionPoolPhaseDestroying = TensorFusionPoolPhase(constants.PhaseDestroying)
)
View Source 
const (
	TensorFusionClusterPending    = TensorFusionClusterPhase(constants.PhasePending)
	TensorFusionClusterRunning    = TensorFusionClusterPhase(constants.PhaseRunning)
	TensorFusionClusterUpdating   = TensorFusionClusterPhase(constants.PhaseUpdating)
	TensorFusionClusterDestroying = TensorFusionClusterPhase(constants.PhaseDestroying)
	TensorFusionClusterUnknown    = TensorFusionClusterPhase(constants.PhaseUnknown)
)

Variables

View Source 
var (
	// GroupVersion is group version used to register these objects.
	GroupVersion = schema.GroupVersion{Group: "tensor-fusion.ai", Version: "v1"}

	// SchemeBuilder is used to add go types to the GroupVersionKind scheme.
	SchemeBuilder = &scheme.Builder{GroupVersion: GroupVersion}

	// AddToScheme adds the types in this group-version to the given scheme.
	AddToScheme = SchemeBuilder.AddToScheme
)

Functions

This section is empty.

Types

type AlertConfig 

type AlertConfig struct {
	// +optional
	Expression *runtime.RawExtension `json:"expression,omitempty"`
}

func (*AlertConfig) DeepCopy 

func (in *AlertConfig) DeepCopy() *AlertConfig

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

func (*AlertConfig) DeepCopyInto 

func (in *AlertConfig) DeepCopyInto(out *AlertConfig)

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

type AuthTypeEnum 

type AuthTypeEnum string

+kubebuilder:validation:Enum=accessKey;serviceAccountRole 

const (
	AuthTypeAccessKey          AuthTypeEnum = "accessKey"
	AuthTypeServiceAccountRole AuthTypeEnum = "serviceAccountRole"
)

type AutoFreeze 

type AutoFreeze struct {
	Qos             QoSLevel `json:"qos,omitempty"`
	FreezeToMemTTL  string   `json:"freezeToMemTTL,omitempty"`
	FreezeToDiskTTL string   `json:"freezeToDiskTTL,omitempty"`
	Enable          *bool    `json:"enable,omitempty"`
}

func (*AutoFreeze) DeepCopy 

func (in *AutoFreeze) DeepCopy() *AutoFreeze

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

func (*AutoFreeze) DeepCopyInto 

func (in *AutoFreeze) DeepCopyInto(out *AutoFreeze)

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

type AutoScalingConfig 

type AutoScalingConfig struct {
	// layer 1 vertical auto-scaling, turbo burst to existing GPU cards quickly
	// VPA-like, aggregate metrics data <1m
	AutoSetLimits AutoSetLimits `json:"autoSetLimits,omitempty"`

	// layer 2 horizontal auto-scaling, scale up to more GPU cards if max limits threshold hit
	// HPA-like, aggregate metrics data 1m-1h (when tf-worker scaled-up, should also trigger client pod's owner[Deployment etc.]'s replica increasing, check if KNative works)
	AutoSetReplicas AutoSetReplicas `json:"autoSetReplicas,omitempty"`

	// layer 3 adjusting, to match the actual usage in the long run
	// Adjust baseline requests to match the actual usage in longer period, such as 1day - 2weeks
	AutoSetRequests AutoSetRequests `json:"autoSetRequests,omitempty"`

	// additional layer to save VRAM, auto-freeze memory and cool down to RAM and Disk
	// Hypervisor will monitor and trigger freeze of inactive workers, Operator should mark them as scaled-to-zero and release the GPU pool resources, don't scale down CPU client part, so that they can continue to serve the traffic or scale down by other auto-scaling solutions like KEDA/KNative
	ScaleToZero ScaleToZero `json:"scaleToZero,omitempty"`
}

func (*AutoScalingConfig) DeepCopy 

func (in *AutoScalingConfig) DeepCopy() *AutoScalingConfig

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

func (*AutoScalingConfig) DeepCopyInto 

func (in *AutoScalingConfig) DeepCopyInto(out *AutoScalingConfig)

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

type AutoSetLimits 

type AutoSetLimits struct {
	Enable bool `json:"enable,omitempty"`

	// target resource to scale limits, such as "tflops", "vram", or "all" by default
	TargetResource string `json:"targetResource,omitempty"`

	EvaluationPeriod string `json:"evaluationPeriod,omitempty"`

	ExtraTFlopsBufferRatio string `json:"extraTFlopsBufferRatio,omitempty"`

	IgnoredDeltaRange string `json:"ignoredDeltaRange,omitempty"`

	ScaleUpStep string `json:"scaleUpStep,omitempty"`

	// the multiplier of requests, to avoid limit set too high, like 5.0
	MaxRatioToRequests string `json:"maxRatioToRequests,omitempty"`

	Prediction *SmartSchedulerModelInput `json:"prediction,omitempty"`
}

A typical autoLimits algorithm could be checking every 5m, look back 1 day data, select 99% of actual usage as preferredLimits, calculate finalPreferredLimits, which is preferredLimits*(1+extraBufferRatio) if they are equal with each other within a range (eg. 5%), do nothing if finalPreferredLimits is less than current limits and exceeded error range, set current limits to finalPreferredLimits if finalPreferredLimits > current limits and exceeded error range, set current limits to max(finalPreferredLimits, current limits * scaleUpStep) if AI prediction enabled, it helps to detect history pattern, and set more reasonable, explainable limit value the final set limits should be max(finalPreferredLimits, last(predict_value * (1 + extraTFlopsBufferRatio)))

func (*AutoSetLimits) DeepCopy 

func (in *AutoSetLimits) DeepCopy() *AutoSetLimits

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

func (*AutoSetLimits) DeepCopyInto 

func (in *AutoSetLimits) DeepCopyInto(out *AutoSetLimits)

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

type AutoSetReplicas 

type AutoSetReplicas struct {
	Enable                bool   `json:"enable,omitempty"`
	TargetTFlopsOfLimits  string `json:"targetTFlopsOfLimits,omitempty"`
	EvaluationPeriod      string `json:"evaluationPeriod,omitempty"`
	ScaleUpStep           string `json:"scaleUpStep,omitempty"`
	ScaleDownStep         string `json:"scaleDownStep,omitempty"`
	ScaleUpCoolDownTime   string `json:"scaleUpCoolDownTime,omitempty"`
	ScaleDownCoolDownTime string `json:"scaleDownCoolDownTime,omitempty"`
}

To handle burst traffic, scale up in short time (this feature requires GPU context migration & replication, not available yet)

func (*AutoSetReplicas) DeepCopy 

func (in *AutoSetReplicas) DeepCopy() *AutoSetReplicas

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

func (*AutoSetReplicas) DeepCopyInto 

func (in *AutoSetReplicas) DeepCopyInto(out *AutoSetReplicas)

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

type AutoSetRequests 

type AutoSetRequests struct {
	Enable bool `json:"enable,omitempty"`

	// target resource to scale requests, such as "tflops", "vram", or "all" by default
	TargetResource string `json:"targetResource,omitempty"`

	PercentileForAutoRequests string `json:"percentileForAutoRequests,omitempty"`

	// the request buffer ratio, for example actual usage is 1.0, 10% buffer will be 1.1 as final preferred requests
	ExtraBufferRatio string `json:"extraBufferRatio,omitempty"`

	EvaluationPeriod  string                   `json:"evaluationPeriod,omitempty"`
	AggregationPeriod string                   `json:"aggregationPeriod,omitempty"`
	Prediction        SmartSchedulerModelInput `json:"prediction,omitempty"`
}

func (*AutoSetRequests) DeepCopy 

func (in *AutoSetRequests) DeepCopy() *AutoSetRequests

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

func (*AutoSetRequests) DeepCopyInto 

func (in *AutoSetRequests) DeepCopyInto(out *AutoSetRequests)

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

type BudgetExceedStrategy 

type BudgetExceedStrategy string

+kubebuilder:validation:Enum=AlertOnly;AlertAndTerminateVM 

const (
	BudgetExceedStrategyAlertOnly           BudgetExceedStrategy = "AlertOnly"
	BudgetExceedStrategyAlertAndTerminateVM BudgetExceedStrategy = "AlertAndTerminateVM"
)

type CapacityConfig 

type CapacityConfig struct {
	// +optional
	MinResources *GPUOrCPUResourceUnit `json:"minResources,omitempty"`

	// +optional
	MaxResources *GPUOrCPUResourceUnit `json:"maxResources,omitempty"`

	// +optional
	WarmResources *GPUOrCPUResourceUnit `json:"warmResources,omitempty"`

	// +optional
	Oversubscription *Oversubscription `json:"oversubscription,omitempty"`
}

func (*CapacityConfig) DeepCopy 

func (in *CapacityConfig) DeepCopy() *CapacityConfig

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

func (*CapacityConfig) DeepCopyInto 

func (in *CapacityConfig) DeepCopyInto(out *CapacityConfig)

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

type ClientConfig 

type ClientConfig struct {
	OperatorEndpoint string `json:"operatorEndpoint,omitempty"`

	// +optional
	PatchToPod *runtime.RawExtension `json:"patchToPod,omitempty"`

	// +optional
	PatchToContainer *runtime.RawExtension `json:"patchToContainer,omitempty"`
}

func (*ClientConfig) DeepCopy 

func (in *ClientConfig) DeepCopy() *ClientConfig

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

func (*ClientConfig) DeepCopyInto 

func (in *ClientConfig) DeepCopyInto(out *ClientConfig)

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

type ComponentConfig 

type ComponentConfig struct {
	// +optional
	Worker *WorkerConfig `json:"worker,omitempty"`

	// +optional
	Hypervisor *HypervisorConfig `json:"hypervisor,omitempty"`

	// +optional
	NodeDiscovery *NodeDiscoveryConfig `json:"nodeDiscovery,omitempty"`

	// +optional
	Client *ClientConfig `json:"client,omitempty"`
}

Customize system components for seamless onboarding.

func (*ComponentConfig) DeepCopy 

func (in *ComponentConfig) DeepCopy() *ComponentConfig

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

func (*ComponentConfig) DeepCopyInto 

func (in *ComponentConfig) DeepCopyInto(out *ComponentConfig)

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

type ComputingVendorConfig 

type ComputingVendorConfig struct {
	Name string `json:"name,omitempty"`

	// support popular cloud providers
	Type ComputingVendorName `json:"type,omitempty"`

	AuthType AuthTypeEnum `json:"authType,omitempty"` // Authentication type (e.g., accessKey, serviceAccount).

	// +optional
	// +kubebuilder:default=true
	Enable *bool `json:"enable,omitempty"` // Enable or disable the computing vendor.

	Params ComputingVendorParams `json:"params,omitempty"`
}

ComputingVendorConfig defines the Cloud vendor connection such as AWS, GCP, Azure etc.

func (*ComputingVendorConfig) DeepCopy 

func (in *ComputingVendorConfig) DeepCopy() *ComputingVendorConfig

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

func (*ComputingVendorConfig) DeepCopyInto 

func (in *ComputingVendorConfig) DeepCopyInto(out *ComputingVendorConfig)

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

type ComputingVendorName 

type ComputingVendorName string

+kubebuilder:validation:Enum=aws;lambda-labs;gcp;azure;oracle-oci;ibm;openshift;vultr;together-ai;alibaba;nvidia;tencent;runpod;mock 

const (
	ComputingVendorAWS        ComputingVendorName = "aws"
	ComputingVendorGCP        ComputingVendorName = "gcp"
	ComputingVendorAzure      ComputingVendorName = "azure"
	ComputingVendorOracle     ComputingVendorName = "oracle-oci"
	ComputingVendorIBM        ComputingVendorName = "ibm"
	ComputingVendorOpenShift  ComputingVendorName = "openshift"
	ComputingVendorVultr      ComputingVendorName = "vultr"
	ComputingVendorTogetherAI ComputingVendorName = "together-ai"
	ComputingVendorLambdaLabs ComputingVendorName = "lambda-labs"
	ComputingVendorAlibaba    ComputingVendorName = "alibaba"
	ComputingVendorNvidia     ComputingVendorName = "nvidia"
	ComputingVendorTencent    ComputingVendorName = "tencent"
	ComputingVendorRunPod     ComputingVendorName = "runpod"

	// This is not unit/integration testing only, no cloud provider is involved
	ComputingVendorMock ComputingVendorName = "mock"
)

type ComputingVendorParams 

type ComputingVendorParams struct {
	// +optional
	DefaultRegion string `json:"defaultRegion,omitempty"` // Region for the computing vendor.

	// the secret of access key and secret key or config file, must be mounted as file path
	// +optional
	AccessKeyPath string `json:"accessKeyPath,omitempty"`
	// +optional
	SecretKeyPath string `json:"secretKeyPath,omitempty"`

	// preferred IAM role since it's more secure
	// +optional
	IAMRole string `json:"iamRole,omitempty"`

	// +optional
	ConfigFile string `json:"configFile,omitempty"`

	// +optional
	// User can set extra cloud vendor params, eg.
	// in ali cloud:" spotPriceLimit, spotDuration, spotInterruptionBehavior, systemDiskCategory, systemDiskSize, dataDiskPerformanceLevel
	// in aws cloud: TODO
	ExtraParams map[string]string `json:"extraParams,omitempty"`
}

func (*ComputingVendorParams) DeepCopy 

func (in *ComputingVendorParams) DeepCopy() *ComputingVendorParams

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

func (*ComputingVendorParams) DeepCopyInto 

func (in *ComputingVendorParams) DeepCopyInto(out *ComputingVendorParams)

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

type DataPipeline4ResourcesConfig 

type DataPipeline4ResourcesConfig struct {
	// +optional
	SyncToCloud *bool `json:"syncToCloud,omitempty"` // Whether to sync resources to the cloud.

	// +optional human readable time like 1h, 1d, default to 1h
	SyncPeriod string `json:"syncPeriod,omitempty"` // Period for syncing resources.
}

func (*DataPipeline4ResourcesConfig) DeepCopy 

func (in *DataPipeline4ResourcesConfig) DeepCopy() *DataPipeline4ResourcesConfig

DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DataPipeline4ResourcesConfig.

func (*DataPipeline4ResourcesConfig) DeepCopyInto 

func (in *DataPipeline4ResourcesConfig) DeepCopyInto(out *DataPipeline4ResourcesConfig)

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

type DataPipeline4TimeSeriesConfig 

type DataPipeline4TimeSeriesConfig struct {
	AggregationPeriods       []string          `json:"aggregationPeriods,omitempty"`       // List of aggregation periods.
	RawDataRetention         string            `json:"rawDataRetention,omitempty"`         // Retention period for raw data.
	AggregationDataRetention string            `json:"aggregationDataRetention,omitempty"` // Retention period for aggregated data.
	RemoteWrite              RemoteWriteConfig `json:"remoteWrite,omitempty"`              // Configuration for remote write.
}

func (*DataPipeline4TimeSeriesConfig) DeepCopy 

func (in *DataPipeline4TimeSeriesConfig) DeepCopy() *DataPipeline4TimeSeriesConfig

DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DataPipeline4TimeSeriesConfig.

func (*DataPipeline4TimeSeriesConfig) DeepCopyInto 

func (in *DataPipeline4TimeSeriesConfig) DeepCopyInto(out *DataPipeline4TimeSeriesConfig)

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

type DataPipelineResultRemoteWriteConfig 

type DataPipelineResultRemoteWriteConfig struct {
	Type string `json:"type,omitempty"` // Type of the connection (e.g., datadog).
	URL  string `json:"url,omitempty"`  // URL of the connection.
}

func (*DataPipelineResultRemoteWriteConfig) DeepCopy 

func (in *DataPipelineResultRemoteWriteConfig) DeepCopy() *DataPipelineResultRemoteWriteConfig

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

func (*DataPipelineResultRemoteWriteConfig) DeepCopyInto 

func (in *DataPipelineResultRemoteWriteConfig) DeepCopyInto(out *DataPipelineResultRemoteWriteConfig)

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

type DataPipelinesConfig 

type DataPipelinesConfig struct {
	Resources DataPipeline4ResourcesConfig `json:"resources,omitempty"`

	Timeseries DataPipeline4TimeSeriesConfig `json:"timeseries,omitempty"`
}

DataPipelinesConfig defines the aggregation jobs that can make statistics on the data and then report to cloud if configured.

func (*DataPipelinesConfig) DeepCopy 

func (in *DataPipelinesConfig) DeepCopy() *DataPipelinesConfig

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

func (*DataPipelinesConfig) DeepCopyInto 

func (in *DataPipelinesConfig) DeepCopyInto(out *DataPipelinesConfig)

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

type GPU 

type GPU struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Status GPUStatus `json:"status,omitempty"`
}

+kubebuilder:object:root=true +kubebuilder:subresource:status +kubebuilder:resource:scope=Cluster +kubebuilder:printcolumn:name="GPU Model",type="string",JSONPath=".spec.gpuModel" +kubebuilder:printcolumn:name="Phase",type="string",JSONPath=".status.phase" +kubebuilder:printcolumn:name="Total TFlops",type="string",JSONPath=".status.capacity.tflops" +kubebuilder:printcolumn:name="Total VRAM",type="string",JSONPath=".status.capacity.vram" +kubebuilder:printcolumn:name="Available TFlops",type="string",JSONPath=".status.available.tflops" +kubebuilder:printcolumn:name="Available VRAM",type="string",JSONPath=".status.available.vram" +kubebuilder:printcolumn:name="Device UUID",type="string",JSONPath=".status.uuid" GPU is the Schema for the gpus API.

func (*GPU) DeepCopy 

func (in *GPU) DeepCopy() *GPU

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

func (*GPU) DeepCopyInto 

func (in *GPU) DeepCopyInto(out *GPU)

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

func (*GPU) DeepCopyObject 

func (in *GPU) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type GPUFilter 

type GPUFilter struct {
	Type   string               `json:"type,omitempty"`
	Params runtime.RawExtension `json:"params,omitempty"`
}

GPUFilter is to select eligible GPUs for scheduling.

example: ```yaml - type: avoidTooMuchConnectionsOnSameGPU params: 

connectionNum: 150

- type: avoidDifferentZone params: 

# by default, GPU worker will be scheduled into the same zone as CPU Client Pod to align AZ and improve performance
topologyKey: topology.kubernetes.io/zone

```

func (*GPUFilter) DeepCopy 

func (in *GPUFilter) DeepCopy() *GPUFilter

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

func (*GPUFilter) DeepCopyInto 

func (in *GPUFilter) DeepCopyInto(out *GPUFilter)

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

type GPUList 

type GPUList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []GPU `json:"items"`
}

GPUList contains a list of GPU.

func (*GPUList) DeepCopy 

func (in *GPUList) DeepCopy() *GPUList

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

func (*GPUList) DeepCopyInto 

func (in *GPUList) DeepCopyInto(out *GPUList)

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

func (*GPUList) DeepCopyObject 

func (in *GPUList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type GPUNode 

type GPUNode struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   GPUNodeSpec   `json:"spec,omitempty"`
	Status GPUNodeStatus `json:"status,omitempty"`
}

+kubebuilder:object:root=true +kubebuilder:subresource:status +kubebuilder:resource:scope=Cluster +kubebuilder:printcolumn:name="Phase",type="string",JSONPath=".status.phase" +kubebuilder:printcolumn:name="Total TFlops",type="string",JSONPath=".status.totalTFlops" +kubebuilder:printcolumn:name="Total VRAM",type="string",JSONPath=".status.totalVRAM" +kubebuilder:printcolumn:name="Virtual TFlops",type="string",JSONPath=".status.virtualTFlops" +kubebuilder:printcolumn:name="Virtual VRAM",type="string",JSONPath=".status.virtualVRAM" +kubebuilder:printcolumn:name="Available TFlops",type="string",JSONPath=".status.availableTFlops" +kubebuilder:printcolumn:name="Available VRAM",type="string",JSONPath=".status.availableVRAM" +kubebuilder:printcolumn:name="GPU Count",type="integer",JSONPath=".status.totalGPUs" GPUNode is the Schema for the gpunodes API.

func (*GPUNode) DeepCopy 

func (in *GPUNode) DeepCopy() *GPUNode

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

func (*GPUNode) DeepCopyInto 

func (in *GPUNode) DeepCopyInto(out *GPUNode)

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

func (*GPUNode) DeepCopyObject 

func (in *GPUNode) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

func (*GPUNode) InitializeStatus 

func (node *GPUNode) InitializeStatus(initTFlops, initVRAM resource.Quantity, initGPUs int32)

func (*GPUNode) SetAnnotationToTriggerNodeSync 

func (node *GPUNode) SetAnnotationToTriggerNodeSync()

type GPUNodeAllocationDetails 

type GPUNodeAllocationDetails struct {
	PodID        string `json:"podID,omitempty"`
	PodName      string `json:"podName,omitempty"`
	Namespace    string `json:"namespace"`
	WorkloadName string `json:"workload,omitempty"`

	Requests GPUResourceUnit `json:"requests"`
	Limits   GPUResourceUnit `json:"limits"`
	QoS      QoSLevel        `json:"qos,omitempty"`
}

func (*GPUNodeAllocationDetails) DeepCopy 

func (in *GPUNodeAllocationDetails) DeepCopy() *GPUNodeAllocationDetails

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

func (*GPUNodeAllocationDetails) DeepCopyInto 

func (in *GPUNodeAllocationDetails) DeepCopyInto(out *GPUNodeAllocationDetails)

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

type GPUNodeClass 

type GPUNodeClass struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   GPUNodeClassSpec   `json:"spec,omitempty"`
	Status GPUNodeClassStatus `json:"status,omitempty"`
}

+kubebuilder:object:root=true +kubebuilder:subresource:status +kubebuilder:resource:scope=Cluster GPUNodeClass is the Schema for the gpunodeclasses API.

func (*GPUNodeClass) DeepCopy 

func (in *GPUNodeClass) DeepCopy() *GPUNodeClass

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

func (*GPUNodeClass) DeepCopyInto 

func (in *GPUNodeClass) DeepCopyInto(out *GPUNodeClass)

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

func (*GPUNodeClass) DeepCopyObject 

func (in *GPUNodeClass) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type GPUNodeClassList 

type GPUNodeClassList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []GPUNodeClass `json:"items"`
}

GPUNodeClassList contains a list of GPUNodeClass.

func (*GPUNodeClassList) DeepCopy 

func (in *GPUNodeClassList) DeepCopy() *GPUNodeClassList

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

func (*GPUNodeClassList) DeepCopyInto 

func (in *GPUNodeClassList) DeepCopyInto(out *GPUNodeClassList)

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

func (*GPUNodeClassList) DeepCopyObject 

func (in *GPUNodeClassList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type GPUNodeClassSpec 

type GPUNodeClassSpec struct {
	// +optional
	// The launch template to use for VM instances, if set, all other fields could be skipped
	LaunchTemplate NodeClassItemSelectorTerms `json:"launchTemplate"`

	// +optional
	// Could be private or public, varies in different cloud vendor, define where to query the OSImageID
	// +kubebuilder:default="Private"
	OSImageType OSImageTypeEnum `json:"osImageType,omitempty"`

	// the OS image identifier string, default to use first one, if not found, fallback to others
	OSImageSelectorTerms []NodeClassItemSelectorTerms `json:"osImageSelectorTerms,omitempty"`

	// +optional
	// The instance profile to use, assign IAM role and permissions for EC2 instances
	InstanceProfile string `json:"instanceProfile,omitempty"`

	// +optional
	// for AWS only, IMDSv2 metadata service options
	MetadataOptions *NodeClassMetadataOptions `json:"metadataOptions,omitempty"`

	// +optional
	SecurityGroupSelectorTerms []NodeClassItemSelectorTerms `json:"securityGroupSelectorTerms,omitempty"`

	// +optional
	SubnetSelectorTerms []NodeClassItemSelectorTerms `json:"subnetSelectorTerms,omitempty"` // Terms to select subnets

	// +optional
	BlockDeviceMappings []NodeClassBlockDeviceMappings `json:"blockDeviceMappings,omitempty"` // Block device mappings for the instance

	// +optional
	Tags map[string]string `json:"tags,omitempty"` // Tags associated with the resource

	// +optional
	UserData string `json:"userData,omitempty"` // User data script for the instance
}

GPUNodeClassSpec defines the desired state of GPUNodeClass.

func (*GPUNodeClassSpec) DeepCopy 

func (in *GPUNodeClassSpec) DeepCopy() *GPUNodeClassSpec

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

func (*GPUNodeClassSpec) DeepCopyInto 

func (in *GPUNodeClassSpec) DeepCopyInto(out *GPUNodeClassSpec)

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

type GPUNodeClassStatus 

type GPUNodeClassStatus struct {
}

GPUNodeClassStatus defines the observed state of GPUNodeClass.

func (*GPUNodeClassStatus) DeepCopy 

func (in *GPUNodeClassStatus) DeepCopy() *GPUNodeClassStatus

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

func (*GPUNodeClassStatus) DeepCopyInto 

func (in *GPUNodeClassStatus) DeepCopyInto(out *GPUNodeClassStatus)

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

type GPUNodeInfo 

type GPUNodeInfo struct {
	// +optional
	// only set when node is managed by TensorFusion
	InstanceID string `json:"instanceID,omitempty"`
	Region     string `json:"region,omitempty"`

	Hostname         string `json:"hostname,omitempty"`
	IP               string `json:"ip,omitempty"`
	KernelVersion    string `json:"kernelVersion,omitempty"`
	OSImage          string `json:"osImage,omitempty"`
	GPUDriverVersion string `json:"gpuDriverVersion,omitempty"`
	GPUModel         string `json:"gpuModel,omitempty"`
	GPUCount         int32  `json:"gpuCount,omitempty"`
	OperatingSystem  string `json:"operatingSystem,omitempty"`
	Architecture     string `json:"architecture,omitempty"`

	// Additional space for L1/L2 VRAM buffer
	RAMSize      resource.Quantity `json:"ramSize,omitempty"`
	DataDiskSize resource.Quantity `json:"dataDiskSize,omitempty"`
}

func (*GPUNodeInfo) DeepCopy 

func (in *GPUNodeInfo) DeepCopy() *GPUNodeInfo

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

func (*GPUNodeInfo) DeepCopyInto 

func (in *GPUNodeInfo) DeepCopyInto(out *GPUNodeInfo)

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

type GPUNodeList 

type GPUNodeList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []GPUNode `json:"items"`
}

GPUNodeList contains a list of GPUNode.

func (*GPUNodeList) DeepCopy 

func (in *GPUNodeList) DeepCopy() *GPUNodeList

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

func (*GPUNodeList) DeepCopyInto 

func (in *GPUNodeList) DeepCopyInto(out *GPUNodeList)

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

func (*GPUNodeList) DeepCopyObject 

func (in *GPUNodeList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type GPUNodeManageMode 

type GPUNodeManageMode string

+kubebuilder:validation:Enum=Manual;AutoSelect;Provisioned 

const (
	GPUNodeManageModeManual      GPUNodeManageMode = "Manual"
	GPUNodeManageModeAutoSelect  GPUNodeManageMode = "AutoSelect"
	GPUNodeManageModeProvisioned GPUNodeManageMode = "Provisioned"
)

type GPUNodeSpec 

type GPUNodeSpec struct {

	// +kubebuilder:default=AutoSelect
	ManageMode GPUNodeManageMode `json:"manageMode,omitempty"`

	// +optional
	CostPerHour string `json:"costPerHour,omitempty"`

	// if not all GPU cards should be used, specify the GPU card indices, default to empty,
	// onboard all GPU cards to the pool
	// +optional
	GPUCardIndices []int `json:"gpuCardIndices,omitempty"`

	// +optional
	CloudVendorParam string `json:"cloudVendorParam,omitempty"`
}

GPUNodeSpec defines the desired state of GPUNode.

func (*GPUNodeSpec) DeepCopy 

func (in *GPUNodeSpec) DeepCopy() *GPUNodeSpec

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

func (*GPUNodeSpec) DeepCopyInto 

func (in *GPUNodeSpec) DeepCopyInto(out *GPUNodeSpec)

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

type GPUNodeStatus 

type GPUNodeStatus struct {
	// the identifier of the kubernetes node, in nodeSelector mode, GPUNode name is the same as kubernetes node name because of it's owned by the Kubernetes node, while in node provisioning mode owned by the GPUNode, and K8S Node name is uncontrollable
	KubernetesNodeName string `json:"kubernetesNodeName"`

	// +kubebuilder:default=Pending
	Phase TensorFusionGPUNodePhase `json:"phase"`

	// +optional
	Conditions []metav1.Condition `json:"conditions,omitempty"`

	TotalTFlops resource.Quantity `json:"totalTFlops"`
	TotalVRAM   resource.Quantity `json:"totalVRAM"`

	VirtualTFlops resource.Quantity `json:"virtualTFlops"`
	VirtualVRAM   resource.Quantity `json:"virtualVRAM"`

	AvailableTFlops resource.Quantity `json:"availableTFlops"`
	AvailableVRAM   resource.Quantity `json:"availableVRAM"`

	// +optional
	VirtualAvailableTFlops *resource.Quantity `json:"virtualAvailableTFlops,omitempty"`
	// +optional
	VirtualAvailableVRAM *resource.Quantity `json:"virtualAvailableVRAM,omitempty"`

	// +optional
	HypervisorStatus NodeHypervisorStatus `json:"hypervisorStatus,omitempty"`

	// +optional
	NodeInfo GPUNodeInfo `json:"nodeInfo,omitempty"`

	// +optional
	LoadedModels *[]string `json:"loadedModels,omitempty"`

	TotalGPUs   int32 `json:"totalGPUs"`
	ManagedGPUs int32 `json:"managedGPUs"`

	// +optional
	ManagedGPUDeviceIDs []string `json:"managedGPUDeviceIDs,omitempty"`

	ObservedGeneration int64 `json:"observedGeneration,omitempty"`

	// Allocation details is for node compaction, and calculate used apps
	// +optional
	AllocationDetails *[]GPUNodeAllocationDetails `json:"allocationDetails,omitempty"`
}

GPUNodeStatus defines the observed state of GPUNode.

func (*GPUNodeStatus) DeepCopy 

func (in *GPUNodeStatus) DeepCopy() *GPUNodeStatus

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

func (*GPUNodeStatus) DeepCopyInto 

func (in *GPUNodeStatus) DeepCopyInto(out *GPUNodeStatus)

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

type GPUOrCPUResourceUnit 

type GPUOrCPUResourceUnit struct {
	TFlops resource.Quantity `json:"tflops,omitempty"`

	VRAM resource.Quantity `json:"vram,omitempty"`

	// CPU/Memory is only available when CloudVendor connection is enabled
	// +optional
	CPU resource.Quantity `json:"cpu,omitempty"`

	// +optional
	Memory resource.Quantity `json:"memory,omitempty"`
}

func (*GPUOrCPUResourceUnit) DeepCopy 

func (in *GPUOrCPUResourceUnit) DeepCopy() *GPUOrCPUResourceUnit

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

func (*GPUOrCPUResourceUnit) DeepCopyInto 

func (in *GPUOrCPUResourceUnit) DeepCopyInto(out *GPUOrCPUResourceUnit)

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

type GPUPool 

type GPUPool struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   GPUPoolSpec   `json:"spec,omitempty"`
	Status GPUPoolStatus `json:"status,omitempty"`
}

+kubebuilder:printcolumn:name="Phase",type="string",JSONPath=".status.phase" +kubebuilder:printcolumn:name="TFlops Oversubscription",type="string",JSONPath=".spec.capacityConfig.oversubscription.tflopsOversellRatio" +kubebuilder:printcolumn:name="Mode",type="string",JSONPath=".status.mode" +kubebuilder:printcolumn:name="Default Scheduling Strategy",type="string",JSONPath=".spec.schedulingConfigTemplate" +kubebuilder:printcolumn:name="Total Nodes",type="string",JSONPath=".status.totalNodes" +kubebuilder:printcolumn:name="Total GPU",type="string",JSONPath=".status.totalGPUs" +kubebuilder:printcolumn:name="Total Tflops",type="string",JSONPath=".status.totalTFlops" +kubebuilder:printcolumn:name="Total VRAM",type="string",JSONPath=".status.totalVRAM" +kubebuilder:printcolumn:name="Available Tflops",type="string",JSONPath=".status.availableTFlops" +kubebuilder:printcolumn:name="Available VRAM",type="string",JSONPath=".status.availableVRAM"

func (*GPUPool) DeepCopy 

func (in *GPUPool) DeepCopy() *GPUPool

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

func (*GPUPool) DeepCopyInto 

func (in *GPUPool) DeepCopyInto(out *GPUPool)

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

func (*GPUPool) DeepCopyObject 

func (in *GPUPool) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type GPUPoolDefinition 

type GPUPoolDefinition struct {
	Name string `json:"name,omitempty"` // Name of the GPU pool.

	SpecTemplate GPUPoolSpec `json:"specTemplate"`
}

GPUPool defines how to create a GPU pool, could be URL or inline

func (*GPUPoolDefinition) DeepCopy 

func (in *GPUPoolDefinition) DeepCopy() *GPUPoolDefinition

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

func (*GPUPoolDefinition) DeepCopyInto 

func (in *GPUPoolDefinition) DeepCopyInto(out *GPUPoolDefinition)

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

type GPUPoolList 

type GPUPoolList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []GPUPool `json:"items"`
}

GPUPoolList contains a list of GPUPool.

func (*GPUPoolList) DeepCopy 

func (in *GPUPoolList) DeepCopy() *GPUPoolList

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

func (*GPUPoolList) DeepCopyInto 

func (in *GPUPoolList) DeepCopyInto(out *GPUPoolList)

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

func (*GPUPoolList) DeepCopyObject 

func (in *GPUPoolList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type GPUPoolSpec 

type GPUPoolSpec struct {
	CapacityConfig *CapacityConfig `json:"capacityConfig,omitempty"`

	NodeManagerConfig *NodeManagerConfig `json:"nodeManagerConfig,omitempty"`

	// +optional
	ObservabilityConfig *ObservabilityConfig `json:"observabilityConfig,omitempty"`

	// +optional
	QosConfig *QosConfig `json:"qosConfig,omitempty"`

	// +optional
	ComponentConfig *ComponentConfig `json:"componentConfig,omitempty"`

	// +optional
	SchedulingConfigTemplate *string `json:"schedulingConfigTemplate,omitempty"`
}

GPUPoolSpec defines the desired state of GPUPool.

func (*GPUPoolSpec) DeepCopy 

func (in *GPUPoolSpec) DeepCopy() *GPUPoolSpec

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

func (*GPUPoolSpec) DeepCopyInto 

func (in *GPUPoolSpec) DeepCopyInto(out *GPUPoolSpec)

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

type GPUPoolStatus 

type GPUPoolStatus struct {
	Cluster string `json:"cluster,omitempty"`

	// +kubebuilder:default=Pending
	Phase TensorFusionPoolPhase `json:"phase"`

	Conditions []metav1.Condition `json:"conditions,omitempty"`

	TotalNodes    int32 `json:"totalNodes,omitempty"`
	TotalGPUs     int32 `json:"totalGPUs,omitempty"`
	ReadyNodes    int32 `json:"readyNodes"`
	NotReadyNodes int32 `json:"notReadyNodes"`

	TotalTFlops resource.Quantity `json:"totalTFlops"`
	TotalVRAM   resource.Quantity `json:"totalVRAM"`

	VirtualTFlops resource.Quantity `json:"virtualTFlops"`
	VirtualVRAM   resource.Quantity `json:"virtualVRAM"`

	AvailableTFlops resource.Quantity `json:"availableTFlops"`
	AvailableVRAM   resource.Quantity `json:"availableVRAM"`

	// +optional
	VirtualAvailableTFlops *resource.Quantity `json:"virtualAvailableTFlops,omitempty"`
	// +optional
	VirtualAvailableVRAM *resource.Quantity `json:"virtualAvailableVRAM,omitempty"`

	// when updating any component version or config, pool controller will perform rolling update.
	// the status will be updated periodically, default to 5s, progress will be 0-100.
	// when the progress is 100, the component version or config is fully updated.
	ComponentStatus PoolComponentStatus `json:"componentStatus"`

	// TODO: calculated every 1h/1d/1w average
	UtilizedTFlopsPercent string `json:"utilizedTFlopsPercent,omitempty"`
	UtilizedVRAMPercent   string `json:"utilizedVRAMPercent,omitempty"`

	// TODO: updated with interval
	AllocatedTFlopsPercent string `json:"allocatedTFlopsPercent,omitempty"`
	AllocatedVRAMPercent   string `json:"allocatedVRAMPercent,omitempty"`

	// TODO: aggregated with interval
	SavedCostsPerMonth       string `json:"savedCostsPerMonth,omitempty"`
	PotentialSavingsPerMonth string `json:"potentialSavingsPerMonth,omitempty"`

	// +kubebuilder:default=""
	// If the budget is exceeded, the set value in comma separated string to indicate which period caused the exceeding.
	// If this field is not empty, scheduler will not schedule new AI workloads and stop scaling-up check.
	// TODO not implemented yet
	BudgetExceeded string `json:"budgetExceeded,omitempty"`

	// +optional
	LastCompactionTime *metav1.Time `json:"lastCompactionTime,omitempty"`
}

GPUPoolStatus defines the observed state of GPUPool.

func (*GPUPoolStatus) DeepCopy 

func (in *GPUPoolStatus) DeepCopy() *GPUPoolStatus

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

func (*GPUPoolStatus) DeepCopyInto 

func (in *GPUPoolStatus) DeepCopyInto(out *GPUPoolStatus)

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

type GPUResourcePricingUnit 

type GPUResourcePricingUnit struct {

	// +kubebuilder:default="$0.0069228"
	PerFP16TFlopsPerHour string `json:"perFP16TFlopsPerHour,omitempty"`

	// +kubebuilder:default="$0.01548"
	PerGBOfVRAMPerHour string `json:"perGBOfVRAMPerHour,omitempty"`
}

The default pricing based on second level pricing from https://modal.com/pricing with Tensor/CUDA Core : HBM = 2:1

func (*GPUResourcePricingUnit) DeepCopy 

func (in *GPUResourcePricingUnit) DeepCopy() *GPUResourcePricingUnit

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

func (*GPUResourcePricingUnit) DeepCopyInto 

func (in *GPUResourcePricingUnit) DeepCopyInto(out *GPUResourcePricingUnit)

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

type GPUResourceUnit 

type GPUResourceUnit struct {
	// Tera floating point operations per second
	TFlops resource.Quantity `json:"tflops,omitempty"`

	// VRAM is short for Video memory, namely GPU RAM
	VRAM resource.Quantity `json:"vram,omitempty"`
}

func (*GPUResourceUnit) DeepCopy 

func (in *GPUResourceUnit) DeepCopy() *GPUResourceUnit

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

func (*GPUResourceUnit) DeepCopyInto 

func (in *GPUResourceUnit) DeepCopyInto(out *GPUResourceUnit)

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

type GPUStatus 

type GPUStatus struct {
	// +kubebuilder:default=Pending
	Phase TensorFusionGPUPhase `json:"phase"`

	Capacity  *Resource `json:"capacity"`
	Available *Resource `json:"available"`

	UUID string `json:"uuid"`

	// The host match selector to schedule worker pods
	NodeSelector map[string]string `json:"nodeSelector"`
	GPUModel     string            `json:"gpuModel"`

	Message string `json:"message"`
}

GPUStatus defines the observed state of GPU.

func (*GPUStatus) DeepCopy 

func (in *GPUStatus) DeepCopy() *GPUStatus

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

func (*GPUStatus) DeepCopyInto 

func (in *GPUStatus) DeepCopyInto(out *GPUStatus)

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

type HypervisorConfig 

type HypervisorConfig struct {
	// +optional
	PodTemplate *runtime.RawExtension `json:"podTemplate,omitempty"`
}

func (*HypervisorConfig) DeepCopy 

func (in *HypervisorConfig) DeepCopy() *HypervisorConfig

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

func (*HypervisorConfig) DeepCopyInto 

func (in *HypervisorConfig) DeepCopyInto(out *HypervisorConfig)

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

type HypervisorScheduling 

type HypervisorScheduling struct {
	MultiProcessQueuing MultiProcessQueuing `json:"multiProcessQueuing,omitempty"`
}

func (*HypervisorScheduling) DeepCopy 

func (in *HypervisorScheduling) DeepCopy() *HypervisorScheduling

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

func (*HypervisorScheduling) DeepCopyInto 

func (in *HypervisorScheduling) DeepCopyInto(out *HypervisorScheduling)

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

type MaintenanceWindow 

type MaintenanceWindow struct {
	// crontab syntax.
	Includes []string `json:"includes,omitempty"`
}

func (*MaintenanceWindow) DeepCopy 

func (in *MaintenanceWindow) DeepCopy() *MaintenanceWindow

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

func (*MaintenanceWindow) DeepCopyInto 

func (in *MaintenanceWindow) DeepCopyInto(out *MaintenanceWindow)

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

type MonitorConfig 

type MonitorConfig struct {
	Interval string `json:"interval,omitempty"`
}

func (*MonitorConfig) DeepCopy 

func (in *MonitorConfig) DeepCopy() *MonitorConfig

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

func (*MonitorConfig) DeepCopyInto 

func (in *MonitorConfig) DeepCopyInto(out *MonitorConfig)

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

type MultiProcessQueuing 

type MultiProcessQueuing struct {
	// +optional
	Enable *bool `json:"enable,omitempty"`

	Interval string `json:"interval,omitempty"`

	QueueLevelTimeSlices []string `json:"queueLevelTimeSlices,omitempty"`
}

func (*MultiProcessQueuing) DeepCopy 

func (in *MultiProcessQueuing) DeepCopy() *MultiProcessQueuing

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

func (*MultiProcessQueuing) DeepCopyInto 

func (in *MultiProcessQueuing) DeepCopyInto(out *MultiProcessQueuing)

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

type NameNamespace 

type NameNamespace struct {
	Name      string `json:"name,omitempty"`
	Namespace string `json:"namespace,omitempty"`
}

func (*NameNamespace) DeepCopy 

func (in *NameNamespace) DeepCopy() *NameNamespace

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

func (*NameNamespace) DeepCopyInto 

func (in *NameNamespace) DeepCopyInto(out *NameNamespace)

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

type NodeClassBlockDeviceMappings 

type NodeClassBlockDeviceMappings struct {
	// +optional
	DeviceName string `json:"deviceName,omitempty"` // The device name for the block device

	EBS NodeClassBlockDeviceSettings `json:"ebs,omitempty"`
}

func (*NodeClassBlockDeviceMappings) DeepCopy 

func (in *NodeClassBlockDeviceMappings) DeepCopy() *NodeClassBlockDeviceMappings

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

func (*NodeClassBlockDeviceMappings) DeepCopyInto 

func (in *NodeClassBlockDeviceMappings) DeepCopyInto(out *NodeClassBlockDeviceMappings)

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

type NodeClassBlockDeviceSettings 

type NodeClassBlockDeviceSettings struct {
	VolumeSize string `json:"volumeSize,omitempty"`

	// +optional
	// Default value would varies based on the cloud vendor
	// For AWS it's gp3, for Alicloud it's cloud_essd
	VolumeType string `json:"volumeType,omitempty"`

	// +optional
	// +kubebuilder:default=true
	DeleteOnTermination bool `json:"deleteOnTermination,omitempty"` // Whether to delete the EBS volume on termination

	// +optional
	// +kubebuilder:default=true
	Encrypted bool `json:"encrypted,omitempty"` // Whether the EBS volume is encrypted

}

func (*NodeClassBlockDeviceSettings) DeepCopy 

func (in *NodeClassBlockDeviceSettings) DeepCopy() *NodeClassBlockDeviceSettings

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

func (*NodeClassBlockDeviceSettings) DeepCopyInto 

func (in *NodeClassBlockDeviceSettings) DeepCopyInto(out *NodeClassBlockDeviceSettings)

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

type NodeClassItemSelectorTerms 

type NodeClassItemSelectorTerms struct {

	// +optional
	// The item ID
	ID string `json:"id,omitempty"`

	// +optional
	// The item name
	Name string `json:"name,omitempty"`

	// +optional
	// Query by tags
	Tags map[string]string `json:"tags,omitempty"`
}

func (*NodeClassItemSelectorTerms) DeepCopy 

func (in *NodeClassItemSelectorTerms) DeepCopy() *NodeClassItemSelectorTerms

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

func (*NodeClassItemSelectorTerms) DeepCopyInto 

func (in *NodeClassItemSelectorTerms) DeepCopyInto(out *NodeClassItemSelectorTerms)

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

type NodeClassMetadataOptions 

type NodeClassMetadataOptions struct {
	// +optional
	// +kubebuilder:default=true
	HttpEndpoint bool `json:"httpEndpoint,omitempty"`

	// +optional
	// +kubebuilder:default=false
	HttpProtocolIPv6 bool `json:"httpProtocolIPv6,omitempty"`

	// +optional
	// +kubebuilder:default=1
	HttpPutResponseHopLimit int `json:"httpPutResponseHopLimit,omitempty"`

	// +optional
	// +kubebuilder:default="required"
	HttpTokens string `json:"httpTokens,omitempty"`
}

AWS IMDSv2 metadata service options

func (*NodeClassMetadataOptions) DeepCopy 

func (in *NodeClassMetadataOptions) DeepCopy() *NodeClassMetadataOptions

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

func (*NodeClassMetadataOptions) DeepCopyInto 

func (in *NodeClassMetadataOptions) DeepCopyInto(out *NodeClassMetadataOptions)

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

type NodeCompaction 

type NodeCompaction struct {
	// +kubebuilder:default="5m"
	Period string `json:"period,omitempty"`
}

func (*NodeCompaction) DeepCopy 

func (in *NodeCompaction) DeepCopy() *NodeCompaction

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

func (*NodeCompaction) DeepCopyInto 

func (in *NodeCompaction) DeepCopyInto(out *NodeCompaction)

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

type NodeDiscoveryConfig 

type NodeDiscoveryConfig struct {
	// +optional
	PodTemplate *runtime.RawExtension `json:"podTemplate,omitempty"`
}

func (*NodeDiscoveryConfig) DeepCopy 

func (in *NodeDiscoveryConfig) DeepCopy() *NodeDiscoveryConfig

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

func (*NodeDiscoveryConfig) DeepCopyInto 

func (in *NodeDiscoveryConfig) DeepCopyInto(out *NodeDiscoveryConfig)

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

type NodeHypervisorStatus 

type NodeHypervisorStatus struct {
	HypervisorState   string      `json:"hypervisorState,omitempty"`
	HypervisorVersion string      `json:"hypervisorVersion,omitempty"`
	LastHeartbeatTime metav1.Time `json:"lastHeartbeatTime,omitempty"`
}

func (*NodeHypervisorStatus) DeepCopy 

func (in *NodeHypervisorStatus) DeepCopy() *NodeHypervisorStatus

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

func (*NodeHypervisorStatus) DeepCopyInto 

func (in *NodeHypervisorStatus) DeepCopyInto(out *NodeHypervisorStatus)

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

type NodeManagerConfig 

type NodeManagerConfig struct {
	// +kubebuilder:default="AutoSelect"
	ProvisioningMode ProvisioningMode `json:"provisioningMode,omitempty"`

	// +optional
	NodeProvisioner *NodeProvisioner `json:"nodeProvisioner,omitempty"`

	// +optional
	NodeSelector *corev1.NodeSelector `json:"nodeSelector,omitempty"`

	// +optional
	NodeCompaction *NodeCompaction `json:"nodeCompaction,omitempty"`

	// +optional
	NodePoolRollingUpdatePolicy *NodeRollingUpdatePolicy `json:"nodePoolRollingUpdatePolicy,omitempty"`
}

func (*NodeManagerConfig) DeepCopy 

func (in *NodeManagerConfig) DeepCopy() *NodeManagerConfig

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

func (*NodeManagerConfig) DeepCopyInto 

func (in *NodeManagerConfig) DeepCopyInto(out *NodeManagerConfig)

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

type NodeProvisioner 

type NodeProvisioner struct {
	// Mode could be Karpenter or Native, for Karpenter mode, node provisioner will start dummy nodes to provision and warmup GPU nodes, do nothing for CPU nodes, for Native mode, provisioner will create or compact GPU & CPU nodes based on current pods
	// +kubebuilder:default=Native
	Mode NodeProvisionerMode `json:"mode,omitempty"`

	NodeClass string `json:"nodeClass,omitempty"`

	// +optional
	GPURequirements []Requirement `json:"gpuRequirements,omitempty"`
	// +optional
	GPUTaints []Taint `json:"gpuTaints,omitempty"`
	// +optional
	GPULabels map[string]string `json:"gpuNodeLabels,omitempty"`

	// +optional
	CPURequirements []Requirement `json:"cpuRequirements,omitempty"`
	// +optional
	CPUTaints []Taint `json:"cpuTaints,omitempty"`
	// +optional
	CPULabels map[string]string `json:"cpuNodeLabels,omitempty"`

	// +optional
	// NodeProvisioner will start an virtual billing based on public pricing or customized pricing, if the VM's costs exceeded any budget constraints, the new VM will not be created, and alerts will be generated
	Budget *PeriodicalBudget `json:"budget,omitempty"`
}

NodeProvisioner or NodeSelector, they are exclusive. NodeSelector is for existing GPUs, NodeProvisioner is for Karpenter-like auto management.

func (*NodeProvisioner) DeepCopy 

func (in *NodeProvisioner) DeepCopy() *NodeProvisioner

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

func (*NodeProvisioner) DeepCopyInto 

func (in *NodeProvisioner) DeepCopyInto(out *NodeProvisioner)

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

type NodeProvisionerMode 

type NodeProvisionerMode string

+kubebuilder:validation:Enum=Native;Karpenter 

const (
	NodeProvisionerModeNative    NodeProvisionerMode = "Native"
	NodeProvisionerModeKarpenter NodeProvisionerMode = "Karpenter"
)

type NodeRequirementKey 

type NodeRequirementKey string

+kubebuilder:validation:Enum=node.kubernetes.io/instance-type;kubernetes.io/arch;kubernetes.io/os;topology.kubernetes.io/region;topology.kubernetes.io/zone;karpenter.sh/capacity-type;tensor-fusion.ai/gpu-arch;tensor-fusion.ai/gpu-instance-family;tensor-fusion.ai/gpu-instance-size 

const (
	NodeRequirementKeyInstanceType    NodeRequirementKey = "node.kubernetes.io/instance-type"
	NodeRequirementKeyArchitecture    NodeRequirementKey = "kubernetes.io/arch"
	NodeRequirementKeyGPUArchitecture NodeRequirementKey = "tensor-fusion.ai/gpu-arch"

	NodeRequirementKeyOS     NodeRequirementKey = "kubernetes.io/os"
	NodeRequirementKeyRegion NodeRequirementKey = "topology.kubernetes.io/region"
	NodeRequirementKeyZone   NodeRequirementKey = "topology.kubernetes.io/zone"

	// capacity-type is charging method, can be spot/preemptive or on-demand
	NodeRequirementKeyCapacityType NodeRequirementKey = "karpenter.sh/capacity-type"

	NodeRequirementKeyInstanceFamily NodeRequirementKey = "tensor-fusion.ai/gpu-instance-family"
	NodeRequirementKeyInstanceSize   NodeRequirementKey = "tensor-fusion.ai/gpu-instance-size"
)

type NodeRollingUpdatePolicy 

type NodeRollingUpdatePolicy struct {

	// +kubebuilder:default=true
	// +optional
	AutoUpdate *bool `json:"autoUpdate,omitempty"`

	// +kubebuilder:default=100
	// +kubebuilder:validation:Minimum=0
	// +kubebuilder:validation:Maximum=100
	BatchPercentage int32 `json:"batchPercentage,omitempty"`

	// +kubebuilder:default="10m"
	BatchInterval string `json:"batchInterval,omitempty"`

	// +optional
	// +kubebuilder:default="10m"
	MaxDuration string `json:"maxDuration,omitempty"`

	// +optional
	MaintenanceWindow MaintenanceWindow `json:"maintenanceWindow,omitempty"`
}

func (*NodeRollingUpdatePolicy) DeepCopy 

func (in *NodeRollingUpdatePolicy) DeepCopy() *NodeRollingUpdatePolicy

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

func (*NodeRollingUpdatePolicy) DeepCopyInto 

func (in *NodeRollingUpdatePolicy) DeepCopyInto(out *NodeRollingUpdatePolicy)

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

type OSImageTypeEnum 

type OSImageTypeEnum string

+kubebuilder:validation:Enum=Private;Public;System 

const (
	OSImageTypePrivate OSImageTypeEnum = "Private"
	OSImageTypePublic  OSImageTypeEnum = "Public"
	OSImageTypeSystem  OSImageTypeEnum = "System"
)

type ObservabilityConfig 

type ObservabilityConfig struct {
	// +optional
	Monitor *MonitorConfig `json:"monitor,omitempty"`

	// +optional
	Alert *AlertConfig `json:"alert,omitempty"`
}

func (*ObservabilityConfig) DeepCopy 

func (in *ObservabilityConfig) DeepCopy() *ObservabilityConfig

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

func (*ObservabilityConfig) DeepCopyInto 

func (in *ObservabilityConfig) DeepCopyInto(out *ObservabilityConfig)

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

type Oversubscription 

type Oversubscription struct {
	// the percentage of Host RAM appending to GPU VRAM, default to 50%
	// +optional
	// +kubebuilder:default=50
	// +kubebuilder:validation:Minimum=0
	// +kubebuilder:validation:Maximum=100
	VRAMExpandToHostMem int32 `json:"vramExpandToHostMem,omitempty"`

	// the percentage of Host Disk appending to GPU VRAM, default to 70%
	// +optional
	// +kubebuilder:default=70
	// +kubebuilder:validation:Minimum=0
	// +kubebuilder:validation:Maximum=100
	VRAMExpandToHostDisk int32 `json:"vramExpandToHostDisk,omitempty"`

	// The multi of TFlops to oversell, default to 500%, indicates 5 times oversell
	// +optional
	// +kubebuilder:default=500
	// +kubebuilder:validation:Minimum=100
	// +kubebuilder:validation:Maximum=100000
	TFlopsOversellRatio int32 `json:"tflopsOversellRatio,omitempty"`
}

func (*Oversubscription) DeepCopy 

func (in *Oversubscription) DeepCopy() *Oversubscription

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

func (*Oversubscription) DeepCopyInto 

func (in *Oversubscription) DeepCopyInto(out *Oversubscription)

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

type PeriodicalBudget 

type PeriodicalBudget struct {
	// +kubebuilder:default="100"
	BudgetPerDay string `json:"budgetPerDay,omitempty"`

	// +kubebuilder:default="1000"
	BudgetPerMonth string `json:"budgetPerMonth,omitempty"`

	// +kubebuilder:default="3000"
	BudgetPerQuarter string `json:"budgetPerQuarter,omitempty"`

	// +kubebuilder:default=AlertOnly
	BudgetExceedStrategy BudgetExceedStrategy `json:"budgetExceedStrategy,omitempty"`
}

The budget constraints in dollars

func (*PeriodicalBudget) DeepCopy 

func (in *PeriodicalBudget) DeepCopy() *PeriodicalBudget

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

func (*PeriodicalBudget) DeepCopyInto 

func (in *PeriodicalBudget) DeepCopyInto(out *PeriodicalBudget)

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

type PlacementConfig 

type PlacementConfig struct {
	// +kubebuilder:default=CompactFirst
	Mode PlacementMode `json:"mode"`

	// +kubebuilder:default=true
	// +optional
	AllowUsingLocalGPU *bool `json:"allowUsingLocalGPU,omitempty"` // If false, workloads will not be scheduled directly to GPU nodes with 'localGPU: true'.

	// +optional
	GPUFilters []GPUFilter `json:"gpuFilters,omitempty"`
}

func (*PlacementConfig) DeepCopy 

func (in *PlacementConfig) DeepCopy() *PlacementConfig

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

func (*PlacementConfig) DeepCopyInto 

func (in *PlacementConfig) DeepCopyInto(out *PlacementConfig)

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

type PlacementMode 

type PlacementMode string

+kubebuilder:validation:Enum=CompactFirst;LowLoadFirst 

const (
	// default to compactFirst for cost saving and energy saving
	PlacementModeCompactFirst PlacementMode = "CompactFirst"

	// in some cases, use lowLoadFirst for balance and fairness
	PlacementModeLowLoadFirst PlacementMode = "LowLoadFirst"
)

type PoolComponentStatus 

type PoolComponentStatus struct {
	WorkerVersion        string `json:"worker,omitempty"`
	WorkerConfigSynced   bool   `json:"workerConfigSynced,omitempty"`
	WorkerUpdateProgress int32  `json:"workerUpdateProgress,omitempty"`

	HypervisorVersion        string `json:"hypervisor,omitempty"`
	HypervisorConfigSynced   bool   `json:"hypervisorConfigSynced,omitempty"`
	HyperVisorUpdateProgress int32  `json:"hypervisorUpdateProgress,omitempty"`

	ClientVersion        string `json:"client,omitempty"`
	ClientConfigSynced   bool   `json:"clientConfigSynced,omitempty"`
	ClientUpdateProgress int32  `json:"clientUpdateProgress,omitempty"`
}

func (*PoolComponentStatus) DeepCopy 

func (in *PoolComponentStatus) DeepCopy() *PoolComponentStatus

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

func (*PoolComponentStatus) DeepCopyInto 

func (in *PoolComponentStatus) DeepCopyInto(out *PoolComponentStatus)

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

type PoolProvisioningStatus 

type PoolProvisioningStatus struct {
	InitializingNodes int32 `json:"initializingNodes,omitempty"`
	TerminatingNodes  int32 `json:"terminatingNodes,omitempty"`
	AvailableNodes    int32 `json:"availableNodes,omitempty"`
}

func (*PoolProvisioningStatus) DeepCopy 

func (in *PoolProvisioningStatus) DeepCopy() *PoolProvisioningStatus

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

func (*PoolProvisioningStatus) DeepCopyInto 

func (in *PoolProvisioningStatus) DeepCopyInto(out *PoolProvisioningStatus)

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

type ProvisioningMode 

type ProvisioningMode string

+kubebuilder:validation:Enum=Provisioned;AutoSelect 

const (
	ProvisioningModeProvisioned ProvisioningMode = "Provisioned"
	ProvisioningModeAutoSelect  ProvisioningMode = "AutoSelect"
)

type QoSLevel 

type QoSLevel string

+kubebuilder:validation:Enum=low;medium;high;critical 

const (
	QoSLow      QoSLevel = "low"
	QoSMedium   QoSLevel = "medium"
	QoSHigh     QoSLevel = "high"
	QoSCritical QoSLevel = "critical"
)

type QosConfig 

type QosConfig struct {
	Definitions []QosDefinition `json:"definitions,omitempty"`
	DefaultQoS  QoSLevel        `json:"defaultQoS,omitempty"`
	Pricing     []QosPricing    `json:"pricing,omitempty"`
}

Define different QoS and their price.

func (*QosConfig) DeepCopy 

func (in *QosConfig) DeepCopy() *QosConfig

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

func (*QosConfig) DeepCopyInto 

func (in *QosConfig) DeepCopyInto(out *QosConfig)

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

type QosDefinition 

type QosDefinition struct {
	Name        QoSLevel `json:"name,omitempty"`
	Description string   `json:"description,omitempty"`
	Priority    int      `json:"priority,omitempty"` // Range from 1-100, reflects the scheduling priority when GPU is full and tasks are in the queue.
}

func (*QosDefinition) DeepCopy 

func (in *QosDefinition) DeepCopy() *QosDefinition

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

func (*QosDefinition) DeepCopyInto 

func (in *QosDefinition) DeepCopyInto(out *QosDefinition)

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

type QosPricing 

type QosPricing struct {
	Qos QoSLevel `json:"qos,omitempty"`

	Requests GPUResourcePricingUnit `json:"requests,omitempty"`

	// Default requests and limitsOverRequests are same, indicates normal on-demand serverless GPU usage, in hands-on lab low QoS case, limitsOverRequests should be cheaper, for example Low QoS, ratio should be 0.5
	// +kubebuilder:default="1"
	LimitsOverRequestsChargingRatio string `json:"limitsOverRequests,omitempty"`
}

func (*QosPricing) DeepCopy 

func (in *QosPricing) DeepCopy() *QosPricing

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

func (*QosPricing) DeepCopyInto 

func (in *QosPricing) DeepCopyInto(out *QosPricing)

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

type ReBalanceThreshold 

type ReBalanceThreshold struct {
	MatchAny runtime.RawExtension `json:"matchAny,omitempty"`
}

func (*ReBalanceThreshold) DeepCopy 

func (in *ReBalanceThreshold) DeepCopy() *ReBalanceThreshold

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

func (*ReBalanceThreshold) DeepCopyInto 

func (in *ReBalanceThreshold) DeepCopyInto(out *ReBalanceThreshold)

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

type ReBalancerConfig 

type ReBalancerConfig struct {
	Enable                *bool              `json:"enable,omitempty"`
	Interval              string             `json:"interval,omitempty"`
	ReBalanceCoolDownTime string             `json:"reBalanceCoolDownTime,omitempty"`
	Threshold             ReBalanceThreshold `json:"threshold,omitempty"`
}

Avoid hot GPU devices and continuously balance the workload\nimplemented by trigger a simulation scheduling and advise better GPU nodes for scheduler

func (*ReBalancerConfig) DeepCopy 

func (in *ReBalancerConfig) DeepCopy() *ReBalancerConfig

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

func (*ReBalancerConfig) DeepCopyInto 

func (in *ReBalancerConfig) DeepCopyInto(out *ReBalancerConfig)

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

type RemoteWriteConfig 

type RemoteWriteConfig struct {
	Connection DataPipelineResultRemoteWriteConfig `json:"connection,omitempty"`
	Metrics    []string                            `json:"metrics,omitempty"` // List of metrics to remote write.
}

RemoteWriteConfig represents the configuration for remote write.

func (*RemoteWriteConfig) DeepCopy 

func (in *RemoteWriteConfig) DeepCopy() *RemoteWriteConfig

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

func (*RemoteWriteConfig) DeepCopyInto 

func (in *RemoteWriteConfig) DeepCopyInto(out *RemoteWriteConfig)

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

type Requirement 

type Requirement struct {
	Key NodeRequirementKey `json:"key,omitempty"`

	// +kubebuilder:default="In"
	// +kubebuilder:validation:Enum=In;Exists;DoesNotExist;Gt;Lt
	Operator corev1.NodeSelectorOperator `json:"operator,omitempty"`

	Values []string `json:"values,omitempty"`
}

func (*Requirement) DeepCopy 

func (in *Requirement) DeepCopy() *Requirement

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

func (*Requirement) DeepCopyInto 

func (in *Requirement) DeepCopyInto(out *Requirement)

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

type Resource 

type Resource struct {
	Tflops resource.Quantity `json:"tflops"`
	Vram   resource.Quantity `json:"vram"`
}

func (*Resource) DeepCopy 

func (in *Resource) DeepCopy() *Resource

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

func (*Resource) DeepCopyInto 

func (in *Resource) DeepCopyInto(out *Resource)

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

type Resources 

type Resources struct {
	Requests Resource `json:"requests"`
	Limits   Resource `json:"limits"`
}

func (*Resources) DeepCopy 

func (in *Resources) DeepCopy() *Resources

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

func (*Resources) DeepCopyInto 

func (in *Resources) DeepCopyInto(out *Resources)

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

type ScaleToZero 

type ScaleToZero struct {
	AutoFreeze         []AutoFreeze             `json:"autoFreeze,omitempty"`
	IntelligenceWarmup SmartSchedulerModelInput `json:"intelligenceWarmup,omitempty"`
}

func (*ScaleToZero) DeepCopy 

func (in *ScaleToZero) DeepCopy() *ScaleToZero

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

func (*ScaleToZero) DeepCopyInto 

func (in *ScaleToZero) DeepCopyInto(out *ScaleToZero)

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

type SchedulingConfigTemplate 

type SchedulingConfigTemplate struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   SchedulingConfigTemplateSpec   `json:"spec,omitempty"`
	Status SchedulingConfigTemplateStatus `json:"status,omitempty"`
}

+kubebuilder:object:root=true +kubebuilder:subresource:status +kubebuilder:resource:scope=Cluster +kubebuilder:printcolumn:name="Mode",type="string",JSONPath=".spec.placement.mode" +kubebuilder:printcolumn:name="Allow Local GPU",type="string",JSONPath=".spec.placement.allowLocalGPU" SchedulingConfigTemplate is the Schema for the schedulingconfigtemplates API.

func (*SchedulingConfigTemplate) DeepCopy 

func (in *SchedulingConfigTemplate) DeepCopy() *SchedulingConfigTemplate

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

func (*SchedulingConfigTemplate) DeepCopyInto 

func (in *SchedulingConfigTemplate) DeepCopyInto(out *SchedulingConfigTemplate)

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

func (*SchedulingConfigTemplate) DeepCopyObject 

func (in *SchedulingConfigTemplate) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type SchedulingConfigTemplateList 

type SchedulingConfigTemplateList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []SchedulingConfigTemplate `json:"items"`
}

SchedulingConfigTemplateList contains a list of SchedulingConfigTemplate.

func (*SchedulingConfigTemplateList) DeepCopy 

func (in *SchedulingConfigTemplateList) DeepCopy() *SchedulingConfigTemplateList

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

func (*SchedulingConfigTemplateList) DeepCopyInto 

func (in *SchedulingConfigTemplateList) DeepCopyInto(out *SchedulingConfigTemplateList)

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

func (*SchedulingConfigTemplateList) DeepCopyObject 

func (in *SchedulingConfigTemplateList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type SchedulingConfigTemplateSpec 

type SchedulingConfigTemplateSpec struct {

	// place the client or worker to best matched nodes
	Placement PlacementConfig `json:"placement"`

	// scale the workload based on the usage and traffic
	// +optional
	AutoScaling *AutoScalingConfig `json:"autoScaling,omitempty"`

	// avoid hot GPU devices and continuously balance the workload
	// implemented by trigger a simulation scheduling and advise better GPU nodes for scheduler
	// +optional
	ReBalancer *ReBalancerConfig `json:"reBalancer,omitempty"`

	// single GPU device multi-process queuing and fair scheduling with QoS constraint
	// +optional
	Hypervisor *HypervisorScheduling `json:"hypervisor,omitempty"`
}

Place the workload to right nodes and scale smart.

func (*SchedulingConfigTemplateSpec) DeepCopy 

func (in *SchedulingConfigTemplateSpec) DeepCopy() *SchedulingConfigTemplateSpec

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

func (*SchedulingConfigTemplateSpec) DeepCopyInto 

func (in *SchedulingConfigTemplateSpec) DeepCopyInto(out *SchedulingConfigTemplateSpec)

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

type SchedulingConfigTemplateStatus 

type SchedulingConfigTemplateStatus struct {
}

SchedulingConfigTemplateStatus defines the observed state of SchedulingConfigTemplate.

func (*SchedulingConfigTemplateStatus) DeepCopy 

func (in *SchedulingConfigTemplateStatus) DeepCopy() *SchedulingConfigTemplateStatus

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

func (*SchedulingConfigTemplateStatus) DeepCopyInto 

func (in *SchedulingConfigTemplateStatus) DeepCopyInto(out *SchedulingConfigTemplateStatus)

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

type SmartSchedulerModelInput 

type SmartSchedulerModelInput struct {
	Enable            *bool  `json:"enable,omitempty"`
	Model             string `json:"model,omitempty"`
	HistoryDataPeriod string `json:"historyDataPeriod,omitempty"`
	PredictionPeriod  string `json:"predictionPeriod,omitempty"`
}

func (*SmartSchedulerModelInput) DeepCopy 

func (in *SmartSchedulerModelInput) DeepCopy() *SmartSchedulerModelInput

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

func (*SmartSchedulerModelInput) DeepCopyInto 

func (in *SmartSchedulerModelInput) DeepCopyInto(out *SmartSchedulerModelInput)

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

type StorageVendorConfig 

type StorageVendorConfig struct {
	Mode  string `json:"mode,omitempty"`  // Mode of the storage vendor (e.g., cloudnative-pg, timescale-db, RDS for PG).
	Image string `json:"image,omitempty"` // Image for the storage vendor (default to timescale).

	// +optional
	InstallCloudNativePGOperator *bool `json:"installCloudNativePGOperator,omitempty"` // Whether to install CloudNative-PG operator.

	StorageClass      string               `json:"storageClass,omitempty"`      // Storage class for the storage vendor.
	PGExtensions      []string             `json:"pgExtensions,omitempty"`      // List of PostgreSQL extensions to install.
	PGClusterTemplate runtime.RawExtension `json:"pgClusterTemplate,omitempty"` // Extra spec for the PostgreSQL cluster template.
}

StorageVendorConfig defines Postgres database with extensions for timeseries storage and other resource aggregation results, system events and diagnostics reports etc.

func (*StorageVendorConfig) DeepCopy 

func (in *StorageVendorConfig) DeepCopy() *StorageVendorConfig

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

func (*StorageVendorConfig) DeepCopyInto 

func (in *StorageVendorConfig) DeepCopyInto(out *StorageVendorConfig)

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

type Taint 

type Taint struct {
	// +kubebuilder:default=NoSchedule
	// +kubebuilder:validation:Enum=NoSchedule;NoExecute;PreferNoSchedule
	Effect corev1.TaintEffect `json:"effect,omitempty"`
	Key    string             `json:"key,omitempty"`
	Value  string             `json:"value,omitempty"`
}

func (*Taint) DeepCopy 

func (in *Taint) DeepCopy() *Taint

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

func (*Taint) DeepCopyInto 

func (in *Taint) DeepCopyInto(out *Taint)

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

type TensorFusionCluster 

type TensorFusionCluster struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   TensorFusionClusterSpec   `json:"spec,omitempty"`
	Status TensorFusionClusterStatus `json:"status,omitempty"`
}

+kubebuilder:printcolumn:name="Total Tflops",type="string",JSONPath=".status.totalTFlops" +kubebuilder:printcolumn:name="Total VRAM",type="string",JSONPath=".status.totalVRAM" +kubebuilder:printcolumn:name="Available Tflops",type="string",JSONPath=".status.availableTFlops" +kubebuilder:printcolumn:name="Available VRAM",type="string",JSONPath=".status.availableVRAM" TensorFusionCluster is the Schema for the tensorfusionclusters API.

func (*TensorFusionCluster) DeepCopy 

func (in *TensorFusionCluster) DeepCopy() *TensorFusionCluster

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

func (*TensorFusionCluster) DeepCopyInto 

func (in *TensorFusionCluster) DeepCopyInto(out *TensorFusionCluster)

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

func (*TensorFusionCluster) DeepCopyObject 

func (in *TensorFusionCluster) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

func (*TensorFusionCluster) RefreshStatus 

func (tfc *TensorFusionCluster) RefreshStatus(ownedPools []GPUPool)

func (*TensorFusionCluster) SetAsPending 

func (tfc *TensorFusionCluster) SetAsPending()

func (*TensorFusionCluster) SetAsReady 

func (tfc *TensorFusionCluster) SetAsReady(conditions ...metav1.Condition) bool

func (*TensorFusionCluster) SetAsUnknown 

func (tfc *TensorFusionCluster) SetAsUnknown(err error) bool

func (*TensorFusionCluster) SetAsUpdating 

func (tfc *TensorFusionCluster) SetAsUpdating(conditions ...metav1.Condition) bool

type TensorFusionClusterList 

type TensorFusionClusterList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []TensorFusionCluster `json:"items"`
}

TensorFusionClusterList contains a list of TensorFusionCluster.

func (*TensorFusionClusterList) DeepCopy 

func (in *TensorFusionClusterList) DeepCopy() *TensorFusionClusterList

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

func (*TensorFusionClusterList) DeepCopyInto 

func (in *TensorFusionClusterList) DeepCopyInto(out *TensorFusionClusterList)

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

func (*TensorFusionClusterList) DeepCopyObject 

func (in *TensorFusionClusterList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type TensorFusionClusterPhase 

type TensorFusionClusterPhase string

+kubebuilder:validation:Enum=Pending;Running;Updating;Destroying;Unknown TensorFusionClusterPhase represents the phase of the TensorFusionCluster resource.

type TensorFusionClusterSpec 

type TensorFusionClusterSpec struct {
	GPUPools []GPUPoolDefinition `json:"gpuPools,omitempty"`

	// +optional
	ComputingVendor *ComputingVendorConfig `json:"computingVendor,omitempty"`

	// +optional
	StorageVendor *StorageVendorConfig `json:"storageVendor,omitempty"`

	// +optional
	DataPipelines *DataPipelinesConfig `json:"dataPipelines,omitempty"`
}

TensorFusionClusterSpec defines the desired state of TensorFusionCluster.

func (*TensorFusionClusterSpec) DeepCopy 

func (in *TensorFusionClusterSpec) DeepCopy() *TensorFusionClusterSpec

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

func (*TensorFusionClusterSpec) DeepCopyInto 

func (in *TensorFusionClusterSpec) DeepCopyInto(out *TensorFusionClusterSpec)

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

type TensorFusionClusterStatus 

type TensorFusionClusterStatus struct {

	// +kubebuilder:default=Pending
	Phase TensorFusionClusterPhase `json:"phase,omitempty"`

	Conditions []metav1.Condition `json:"conditions,omitempty"`

	TotalPools int32 `json:"totalPools"`
	TotalNodes int32 `json:"totalNodes"`
	TotalGPUs  int32 `json:"totalGPUs"`

	TotalTFlops resource.Quantity `json:"totalTFlops"`
	TotalVRAM   resource.Quantity `json:"totalVRAM"`

	VirtualTFlops resource.Quantity `json:"virtualTFlops"`
	VirtualVRAM   resource.Quantity `json:"virtualVRAM"`

	AvailableTFlops resource.Quantity `json:"availableTFlops"`
	AvailableVRAM   resource.Quantity `json:"availableVRAM"`

	// +optional
	VirtualAvailableTFlops *resource.Quantity `json:"virtualAvailableTFlops,omitempty"`
	// +optional
	VirtualAvailableVRAM *resource.Quantity `json:"virtualAvailableVRAM,omitempty"`

	// +optional
	ReadyGPUPools []string `json:"readyGPUPools"`

	// +optional
	NotReadyGPUPools []string `json:"notReadyGPUPools"`

	// +kubebuilder:default=0
	//
	RetryCount int64 `json:"retryCount"`

	// TODO: calculated every 1h/1d/1w average
	UtilizedTFlopsPercent string `json:"utilizedTFlopsPercent,omitempty"`
	UtilizedVRAMPercent   string `json:"utilizedVRAMPercent,omitempty"`

	// TODO: updated with interval
	AllocatedTFlopsPercent string `json:"allocatedTFlopsPercent,omitempty"`
	AllocatedVRAMPercent   string `json:"allocatedVRAMPercent,omitempty"`

	// TODO: aggregated with interval
	SavedCostsPerMonth       string `json:"savedCostsPerMonth,omitempty"`
	PotentialSavingsPerMonth string `json:"potentialSavingsPerMonth,omitempty"`

	CloudVendorConfigHash string `json:"cloudVendorConfigHash,omitempty"`
}

TensorFusionClusterStatus defines the observed state of TensorFusionCluster.

func (*TensorFusionClusterStatus) DeepCopy 

func (in *TensorFusionClusterStatus) DeepCopy() *TensorFusionClusterStatus

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

func (*TensorFusionClusterStatus) DeepCopyInto 

func (in *TensorFusionClusterStatus) DeepCopyInto(out *TensorFusionClusterStatus)

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

type TensorFusionConnection 

type TensorFusionConnection struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   TensorFusionConnectionSpec   `json:"spec,omitempty"`
	Status TensorFusionConnectionStatus `json:"status,omitempty"`
}

TensorFusionConnection is the Schema for the tensorfusionconnections API.

func (*TensorFusionConnection) DeepCopy 

func (in *TensorFusionConnection) DeepCopy() *TensorFusionConnection

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

func (*TensorFusionConnection) DeepCopyInto 

func (in *TensorFusionConnection) DeepCopyInto(out *TensorFusionConnection)

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

func (*TensorFusionConnection) DeepCopyObject 

func (in *TensorFusionConnection) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type TensorFusionConnectionList 

type TensorFusionConnectionList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []TensorFusionConnection `json:"items"`
}

TensorFusionConnectionList contains a list of TensorFusionConnection.

func (*TensorFusionConnectionList) DeepCopy 

func (in *TensorFusionConnectionList) DeepCopy() *TensorFusionConnectionList

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

func (*TensorFusionConnectionList) DeepCopyInto 

func (in *TensorFusionConnectionList) DeepCopyInto(out *TensorFusionConnectionList)

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

func (*TensorFusionConnectionList) DeepCopyObject 

func (in *TensorFusionConnectionList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type TensorFusionConnectionSpec 

type TensorFusionConnectionSpec struct {
	WorkloadName string `json:"workloadName"`
}

TensorFusionConnectionSpec defines the desired state of TensorFusionConnection.

func (*TensorFusionConnectionSpec) DeepCopy 

func (in *TensorFusionConnectionSpec) DeepCopy() *TensorFusionConnectionSpec

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

func (*TensorFusionConnectionSpec) DeepCopyInto 

func (in *TensorFusionConnectionSpec) DeepCopyInto(out *TensorFusionConnectionSpec)

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

type TensorFusionConnectionStatus 

type TensorFusionConnectionStatus struct {
	Phase         WorkerPhase `json:"phase"`
	ConnectionURL string      `json:"connectionURL"`
	WorkerName    string      `json:"workerName"`
}

TensorFusionConnectionStatus defines the observed state of TensorFusionConnection.

func (*TensorFusionConnectionStatus) DeepCopy 

func (in *TensorFusionConnectionStatus) DeepCopy() *TensorFusionConnectionStatus

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

func (*TensorFusionConnectionStatus) DeepCopyInto 

func (in *TensorFusionConnectionStatus) DeepCopyInto(out *TensorFusionConnectionStatus)

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

type TensorFusionGPUNodePhase 

type TensorFusionGPUNodePhase string

+kubebuilder:validation:Enum=Pending;Provisioning;Migrating;Running;Succeeded;Failed;Unknown;Destroying 

const (
	TensorFusionGPUNodePhasePending    TensorFusionGPUNodePhase = constants.PhasePending
	TensorFusionGPUNodePhaseMigrating  TensorFusionGPUNodePhase = constants.PhaseMigrating
	TensorFusionGPUNodePhaseRunning    TensorFusionGPUNodePhase = constants.PhaseRunning
	TensorFusionGPUNodePhaseSucceeded  TensorFusionGPUNodePhase = constants.PhaseSucceeded
	TensorFusionGPUNodePhaseFailed     TensorFusionGPUNodePhase = constants.PhaseFailed
	TensorFusionGPUNodePhaseUnknown    TensorFusionGPUNodePhase = constants.PhaseUnknown
	TensorFusionGPUNodePhaseDestroying TensorFusionGPUNodePhase = constants.PhaseDestroying
)

type TensorFusionGPUPhase 

type TensorFusionGPUPhase string

+kubebuilder:validation:Enum=Pending;Provisioning;Running;Unknown;Destroying;Migrating 

const (
	TensorFusionGPUPhasePending    TensorFusionGPUPhase = constants.PhasePending
	TensorFusionGPUPhaseUpdating   TensorFusionGPUPhase = constants.PhaseUpdating
	TensorFusionGPUPhaseRunning    TensorFusionGPUPhase = constants.PhaseRunning
	TensorFusionGPUPhaseUnknown    TensorFusionGPUPhase = constants.PhaseUnknown
	TensorFusionGPUPhaseDestroying TensorFusionGPUPhase = constants.PhaseDestroying
	TensorFusionGPUPhaseMigrating  TensorFusionGPUPhase = constants.PhaseMigrating
)

type TensorFusionPoolPhase 

type TensorFusionPoolPhase string

+kubebuilder:validation:Enum=Pending;Running;Updating;Destroying;Unknown

type TensorFusionWorkload 

type TensorFusionWorkload struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   WorkloadProfileSpec        `json:"spec,omitempty"`
	Status TensorFusionWorkloadStatus `json:"status,omitempty"`
}

TensorFusionWorkload is the Schema for the tensorfusionworkloads API.

func (*TensorFusionWorkload) DeepCopy 

func (in *TensorFusionWorkload) DeepCopy() *TensorFusionWorkload

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

func (*TensorFusionWorkload) DeepCopyInto 

func (in *TensorFusionWorkload) DeepCopyInto(out *TensorFusionWorkload)

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

func (*TensorFusionWorkload) DeepCopyObject 

func (in *TensorFusionWorkload) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type TensorFusionWorkloadList 

type TensorFusionWorkloadList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []TensorFusionWorkload `json:"items"`
}

TensorFusionWorkloadList contains a list of TensorFusionWorkload.

func (*TensorFusionWorkloadList) DeepCopy 

func (in *TensorFusionWorkloadList) DeepCopy() *TensorFusionWorkloadList

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

func (*TensorFusionWorkloadList) DeepCopyInto 

func (in *TensorFusionWorkloadList) DeepCopyInto(out *TensorFusionWorkloadList)

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

func (*TensorFusionWorkloadList) DeepCopyObject 

func (in *TensorFusionWorkloadList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type TensorFusionWorkloadPhase added in v1.32.0 

type TensorFusionWorkloadPhase string

+kubebuilder:validation:Enum=Pending;Running;Failed;Unknown 

const (
	TensorFusionWorkloadPhasePending TensorFusionWorkloadPhase = "Pending"
	TensorFusionWorkloadPhaseRunning TensorFusionWorkloadPhase = "Running"
	TensorFusionWorkloadPhaseFailed  TensorFusionWorkloadPhase = "Failed"
)

type TensorFusionWorkloadStatus 

type TensorFusionWorkloadStatus struct {
	// +kubebuilder:default=Pending
	Phase TensorFusionWorkloadPhase `json:"phase,omitempty"`

	// Represents the latest available observations of the workload's current state.
	// +optional
	Conditions []metav1.Condition `json:"conditions,omitempty"`

	// replicas is the number of Pods created by the Workload controller.
	Replicas int32 `json:"replicas"`

	// readyReplicas is the number of pods created for this Workload with a Ready Condition.
	ReadyReplicas int32 `json:"readyReplicas,omitempty"`

	// Represents the status of each worker in the workload.
	// +optional
	WorkerStatuses []WorkerStatus `json:"workerStatuses,omitempty"`

	// Hash of the pod template used to create worker pods
	PodTemplateHash string `json:"podTemplateHash,omitempty"`
}

TensorFusionWorkloadStatus defines the observed state of TensorFusionWorkload.

func (*TensorFusionWorkloadStatus) DeepCopy 

func (in *TensorFusionWorkloadStatus) DeepCopy() *TensorFusionWorkloadStatus

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

func (*TensorFusionWorkloadStatus) DeepCopyInto 

func (in *TensorFusionWorkloadStatus) DeepCopyInto(out *TensorFusionWorkloadStatus)

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

type WorkerConfig 

type WorkerConfig struct {
	// +optional
	PodTemplate *runtime.RawExtension `json:"podTemplate,omitempty"`
}

func (*WorkerConfig) DeepCopy 

func (in *WorkerConfig) DeepCopy() *WorkerConfig

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

func (*WorkerConfig) DeepCopyInto 

func (in *WorkerConfig) DeepCopyInto(out *WorkerConfig)

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

type WorkerPhase 

type WorkerPhase string
const (
	WorkerPending WorkerPhase = "Pending"
	WorkerRunning WorkerPhase = "Running"
	WorkerFailed  WorkerPhase = "Failed"
)

type WorkerStatus 

type WorkerStatus struct {
	WorkerPhase WorkerPhase `json:"workerPhase"`

	WorkerName   string            `json:"workerName"`
	NodeSelector map[string]string `json:"nodeSelector,omitempty"`
	// +optional
	WorkerIp string `json:"workerIp,omitempty"`
	// +optional
	WorkerPort int `json:"workerPort,omitempty"`
	// +optional
	ResourceVersion string `json:"resourceVersion,omitempty"`
}

func (*WorkerStatus) DeepCopy 

func (in *WorkerStatus) DeepCopy() *WorkerStatus

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

func (*WorkerStatus) DeepCopyInto 

func (in *WorkerStatus) DeepCopyInto(out *WorkerStatus)

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

type WorkloadProfile added in v1.23.7 

type WorkloadProfile struct {
	metav1.TypeMeta   `json:",inline"`
	metav1.ObjectMeta `json:"metadata,omitempty"`

	Spec   WorkloadProfileSpec   `json:"spec,omitempty"`
	Status WorkloadProfileStatus `json:"status,omitempty"`
}

WorkloadProfile is the Schema for the workloadprofiles API.

func (*WorkloadProfile) DeepCopy added in v1.23.7 

func (in *WorkloadProfile) DeepCopy() *WorkloadProfile

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

func (*WorkloadProfile) DeepCopyInto added in v1.23.7 

func (in *WorkloadProfile) DeepCopyInto(out *WorkloadProfile)

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

func (*WorkloadProfile) DeepCopyObject added in v1.23.7 

func (in *WorkloadProfile) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type WorkloadProfileList added in v1.23.7 

type WorkloadProfileList struct {
	metav1.TypeMeta `json:",inline"`
	metav1.ListMeta `json:"metadata,omitempty"`
	Items           []WorkloadProfile `json:"items"`
}

WorkloadProfileList contains a list of WorkloadProfile.

func (*WorkloadProfileList) DeepCopy added in v1.23.7 

func (in *WorkloadProfileList) DeepCopy() *WorkloadProfileList

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

func (*WorkloadProfileList) DeepCopyInto added in v1.23.7 

func (in *WorkloadProfileList) DeepCopyInto(out *WorkloadProfileList)

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

func (*WorkloadProfileList) DeepCopyObject added in v1.23.7 

func (in *WorkloadProfileList) DeepCopyObject() runtime.Object

DeepCopyObject is an autogenerated deepcopy function, copying the receiver, creating a new runtime.Object.

type WorkloadProfileSpec added in v1.23.7 

type WorkloadProfileSpec struct {
	// +optional
	Replicas *int32 `json:"replicas,omitempty"`

	// +optional
	PoolName string `json:"poolName,omitempty"`

	Resources Resources `json:"resources,omitempty"`
	// +optional
	// Qos defines the quality of service level for the client.
	Qos QoSLevel `json:"qos,omitempty"`

	// +optional
	// Schedule the workload to the same GPU server that runs vGPU worker for best performance, default to false
	IsLocalGPU bool `json:"isLocalGPU,omitempty"`

	// +optional
	// GPUModel specifies the required GPU model (e.g., "A100", "H100")
	GPUModel string `json:"gpuModel,omitempty"`

	// The number of GPUs to be used by the workload, default to 1
	GPUCount uint `json:"gpuCount,omitempty"`

	// +optional
	// TODO, not implemented
	// This mode is only available when `is-local-gpu` set to true, in this mode, TensorFusion will also inject vGPU worker into init container, so that to achieve best performance, trade-off is user might by-pass the vGPU worker and using physical GPU directly
	NoStandaloneWorkerMode bool `json:"noStandaloneWorkerMode,omitempty"`

	// +optional
	// AutoScalingConfig configured here will override Pool's schedulingConfig
	// This field can not be fully supported in annotation, if user want to enable auto-scaling in annotation,
	// user can set tensor-fusion.ai/auto-limits|requests|replicas: 'true'
	AutoScalingConfig AutoScalingConfig `json:"autoScalingConfig,omitempty"`
}

WorkloadProfileSpec defines the desired state of WorkloadProfile.

func (*WorkloadProfileSpec) DeepCopy added in v1.23.7 

func (in *WorkloadProfileSpec) DeepCopy() *WorkloadProfileSpec

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

func (*WorkloadProfileSpec) DeepCopyInto added in v1.23.7 

func (in *WorkloadProfileSpec) DeepCopyInto(out *WorkloadProfileSpec)

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

type WorkloadProfileStatus added in v1.23.7 

type WorkloadProfileStatus struct {
}

WorkloadProfileStatus defines the observed state of WorkloadProfile.

func (*WorkloadProfileStatus) DeepCopy added in v1.23.7 

func (in *WorkloadProfileStatus) DeepCopy() *WorkloadProfileStatus

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

func (*WorkloadProfileStatus) DeepCopyInto added in v1.23.7 

func (in *WorkloadProfileStatus) DeepCopyInto(out *WorkloadProfileStatus)

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

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.