kubernetes

package
v1.56.2 Latest Latest
Warning

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

 Go to latest Published: Jan 23, 2026 License: Apache-2.0 Imports: 40 Imported by: 0

Documentation

Index

Constants

View Source 
const (
	// DevicePluginPath is the path where device plugins should register
	DevicePluginPath = "/var/lib/kubelet/device-plugins"
	// KubeletSocket is the kubelet registration socket
	KubeletSocket = "kubelet.sock"
	// DevicePluginEndpoint is the endpoint name for this device plugin
	DevicePluginEndpoint = "tensor-fusion-index-%d.sock"
)

Variables

This section is empty.

Functions

func GetWorkerInfoFromHostPID 

func GetWorkerInfoFromHostPID(hostPID uint32) (*framework.ProcessMappingInfo, error)

GetWorkerInfoFromHostPID extracts worker information from a process's environment by reading /proc/{hostPID}/environ and /proc/{hostPID}/status Returns ProcessMappingInfo with Namespace, PodName, ContainerName for worker lookup

Types

type APIClient 

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

APIClient provides CRUD operations for GPU resources

func NewAPIClient 

func NewAPIClient(ctx context.Context, k8sClient client.Client) *APIClient

NewAPIClient creates a new API client instance with an existing client

func NewAPIClientFromConfig 

func NewAPIClientFromConfig(ctx context.Context, restConfig *rest.Config) (*APIClient, error)

NewAPIClientFromConfig creates a new API client instance from a rest.Config

func (*APIClient) CreateOrUpdateGPU 

func (a *APIClient) CreateOrUpdateGPU(
	gpuNodeName string, gpuID string,
	mutateFn func(gpuNode *tfv1.GPUNode, gpu *tfv1.GPU) error,
) error

CreateOrUpdateGPU creates or updates a GPU resource with metadata and status

func (*APIClient) DeleteGPU 

func (a *APIClient) DeleteGPU(uuid string) error

DeleteGPU deletes a GPU resource

func (*APIClient) GetGPU 

func (a *APIClient) GetGPU(uuid string) (*tfv1.GPU, error)

GetGPU retrieves a GPU resource by UUID

func (*APIClient) UpdateGPUNodeStatus 

func (a *APIClient) UpdateGPUNodeStatus(nodeName string, nodeInfo *api.NodeInfo) error

UpdateGPUNodeStatus updates the status of a GPUNode resource

func (*APIClient) UpdateGPUStatus 

func (a *APIClient) UpdateGPUStatus(gpu *tfv1.GPU) error

UpdateGPUStatus updates the status of a GPU resource using merge patch

type DevicePlugin 

type DevicePlugin struct {
	pluginapi.UnimplementedDevicePluginServer
	// contains filtered or unexported fields
}

DevicePlugin implements the Kubernetes device plugin interface

func NewDevicePlugins 

func NewDevicePlugins(ctx context.Context, deviceController framework.DeviceController, allocationController framework.WorkerAllocationController, kubeletClient *PodCacheManager) []*DevicePlugin

NewDevicePlugins creates a new device plugin instance

func (*DevicePlugin) Allocate 

func (dp *DevicePlugin) Allocate(ctx context.Context, req *pluginapi.AllocateRequest) (*pluginapi.AllocateResponse, error)

Allocate handles device allocation requests from kubelet

func (*DevicePlugin) GetDevicePluginOptions 

func (dp *DevicePlugin) GetDevicePluginOptions(ctx context.Context, req *pluginapi.Empty) (*pluginapi.DevicePluginOptions, error)

GetDevicePluginOptions returns options for the device plugin

func (*DevicePlugin) GetPreferredAllocation 

func (dp *DevicePlugin) GetPreferredAllocation(ctx context.Context, req *pluginapi.PreferredAllocationRequest) (*pluginapi.PreferredAllocationResponse, error)

GetPreferredAllocation returns preferred device allocation (optional)

func (*DevicePlugin) ListAndWatch 

func (dp *DevicePlugin) ListAndWatch(req *pluginapi.Empty, stream pluginapi.DevicePlugin_ListAndWatchServer) error

ListAndWatch streams device list and health updates

func (*DevicePlugin) PreStartContainer 

func (dp *DevicePlugin) PreStartContainer(ctx context.Context, req *pluginapi.PreStartContainerRequest) (*pluginapi.PreStartContainerResponse, error)

PreStartContainer is called before container start (optional)

func (*DevicePlugin) Start 

func (dp *DevicePlugin) Start() error

Start starts the device plugin gRPC server and registers with kubelet

func (*DevicePlugin) Stop 

func (dp *DevicePlugin) Stop() error

Stop stops the device plugin

type GPUInfo 

type GPUInfo struct {
	UUID          string
	DeviceName    string
	VRAMBytes     uint64
	TFlops        resource.Quantity
	Index         int32
	NUMANodeID    int32
	NodeName      string
	Vendor        string
	IsolationMode tfv1.IsolationModeType
}

GPUInfo contains information needed to create or update a GPU

type KubeletBackend 

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

func NewKubeletBackend 

func NewKubeletBackend(ctx context.Context, deviceController framework.DeviceController, allocationController framework.WorkerAllocationController, restConfig *rest.Config) (*KubeletBackend, error)

func (*KubeletBackend) GetDeviceChangeHandler 

func (b *KubeletBackend) GetDeviceChangeHandler() framework.DeviceChangeHandler

func (*KubeletBackend) GetProcessMappingInfo 

func (b *KubeletBackend) GetProcessMappingInfo(hostPID uint32) (*framework.ProcessMappingInfo, error)

func (*KubeletBackend) ListWorkers 

func (b *KubeletBackend) ListWorkers() []*api.WorkerInfo

func (*KubeletBackend) RegisterWorkerUpdateHandler 

func (b *KubeletBackend) RegisterWorkerUpdateHandler(handler framework.WorkerChangeHandler) error

RegisterWorkerUpdateHandler registers a handler for worker updates

func (*KubeletBackend) Start 

func (b *KubeletBackend) Start() error

func (*KubeletBackend) StartWorker 

func (b *KubeletBackend) StartWorker(worker *api.WorkerInfo) error

func (*KubeletBackend) Stop 

func (b *KubeletBackend) Stop() error

func (*KubeletBackend) StopWorker 

func (b *KubeletBackend) StopWorker(workerUID string) error

type PodCacheManager 

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

PodCacheManager manages pod watching and worker information extraction

func NewPodCacheManager 

func NewPodCacheManager(ctx context.Context, restConfig *rest.Config, nodeName string) (*PodCacheManager, error)

NewPodCacheManager creates a new pod cache manager

func (*PodCacheManager) GetAllPods 

func (kc *PodCacheManager) GetAllPods() map[string]*corev1.Pod

GetAllPods returns all pods currently in the cache

func (*PodCacheManager) GetPodByUID 

func (kc *PodCacheManager) GetPodByUID(podUID string) *corev1.Pod

GetPodByUID retrieves a pod from the cache by its UID

func (*PodCacheManager) GetWorkerInfoForAllocationByIndex 

func (kc *PodCacheManager) GetWorkerInfoForAllocationByIndex(podIndex int) (*api.WorkerInfo, error)

GetWorkerInfoForAllocationByIndex finds a pod by its index annotation and extracts worker info It implements a Pub/Sub pattern where callers subscribe to worker info changes for a specific pod index. If worker info is already available, it returns immediately. Otherwise, it waits for up to 10 minutes for the worker info to become available.

func (*PodCacheManager) RegisterWorkerInfoSubscriber 

func (kc *PodCacheManager) RegisterWorkerInfoSubscriber(name string, subscriber chan<- *api.WorkerInfo)

func (*PodCacheManager) Start 

func (kc *PodCacheManager) Start() error

Start starts watching pods on this node

func (*PodCacheManager) Stop 

func (kc *PodCacheManager) Stop()

Stop stops the pod cache manager

func (*PodCacheManager) UnregisterWorkerInfoSubscriber 

func (kc *PodCacheManager) UnregisterWorkerInfoSubscriber(name string)

Source Files

View all Source files

Directories

Path Synopsis

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.