nvidia

Documentation

Index

• Constants
• func NewSharedGPUManager(enableMPS, healthCheck, queryKubelet bool, bp MemoryUnit, ...) *sharedGPUManager
• func StackTrace(all bool) string
• type MemoryUnit
• type NvidiaDevicePlugin
  • func NewNvidiaDevicePlugin(mps, healthCheck, queryKubelet bool, client *client.KubeletClient) (*NvidiaDevicePlugin, error)
  • func (m *NvidiaDevicePlugin) Allocate(ctx context.Context, reqs *pluginapi.AllocateRequest) (*pluginapi.AllocateResponse, error)
  • func (m *NvidiaDevicePlugin) GetDeviceNameByIndex(index uint) (name string, found bool)
  • func (m *NvidiaDevicePlugin) GetDevicePluginOptions(context.Context, *pluginapi.Empty) (*pluginapi.DevicePluginOptions, error)
  • func (m *NvidiaDevicePlugin) ListAndWatch(e *pluginapi.Empty, s pluginapi.DevicePlugin_ListAndWatchServer) error
  • func (m *NvidiaDevicePlugin) PreStartContainer(context.Context, *pluginapi.PreStartContainerRequest) (*pluginapi.PreStartContainerResponse, error)
  • func (m *NvidiaDevicePlugin) Register(kubeletEndpoint, resourceName string) error
  • func (m *NvidiaDevicePlugin) Serve() error
  • func (m *NvidiaDevicePlugin) Start() error
  • func (m *NvidiaDevicePlugin) Stop() error

Constants

const (
	OptimisticLockErrorMsg = "the object has been modified; please apply your changes to the latest version and try again"

	EnvResourceIndex           = "NVIDIA_COM_GPU_MEM_IDX"
	EnvResourceByPod           = "NVIDIA_COM_GPU_MEM_POD"
	EnvResourceByContainer     = "NVIDIA_COM_GPU_MEM_CONTAINER"
	EnvResourceByDev           = "NVIDIA_COM_GPU_MEM_DEV"
	EnvAssignedFlag            = "NVIDIA_COM_GPU_MEM_ASSIGNED"
	EnvResourceAssumeTime      = "NVIDIA_COM_GPU_MEM_ASSUME_TIME"
	EnvResourceAssignTime      = "NVIDIA_COM_GPU_MEM_ASSIGN_TIME"
	EnvNodeLabelForDisableCGPU = "cgpu.disable.isolation"

	GiBPrefix = MemoryUnit("GiB")
	MiBPrefix = MemoryUnit("MiB")
)

Functions

func NewSharedGPUManager

func NewSharedGPUManager(enableMPS, healthCheck, queryKubelet bool, bp MemoryUnit, client *client.KubeletClient) *sharedGPUManager

func StackTrace

func StackTrace(all bool) string

Types

type MemoryUnit

type MemoryUnit string

MemoryUnit describes GPU Memory, now only supports Gi, Mi

type NvidiaDevicePlugin

type NvidiaDevicePlugin struct {
	sync.RWMutex
	// contains filtered or unexported fields
}

NvidiaDevicePlugin implements the Kubernetes device plugin API

func NewNvidiaDevicePlugin

func NewNvidiaDevicePlugin(mps, healthCheck, queryKubelet bool, client *client.KubeletClient) (*NvidiaDevicePlugin, error)

NewNvidiaDevicePlugin returns an initialized NvidiaDevicePlugin

func (*NvidiaDevicePlugin) Allocate

func (m *NvidiaDevicePlugin) Allocate(ctx context.Context,
	reqs *pluginapi.AllocateRequest) (*pluginapi.AllocateResponse, error)

Allocate which return list of devices.

func (*NvidiaDevicePlugin) GetDeviceNameByIndex

func (m *NvidiaDevicePlugin) GetDeviceNameByIndex(index uint) (name string, found bool)

func (*NvidiaDevicePlugin) GetDevicePluginOptions

func (m *NvidiaDevicePlugin) GetDevicePluginOptions(context.Context, *pluginapi.Empty) (*pluginapi.DevicePluginOptions, error)

func (*NvidiaDevicePlugin) ListAndWatch

func (m *NvidiaDevicePlugin) ListAndWatch(e *pluginapi.Empty, s pluginapi.DevicePlugin_ListAndWatchServer) error

ListAndWatch lists devices and update that list according to the health status

func (*NvidiaDevicePlugin) PreStartContainer

func (m *NvidiaDevicePlugin) PreStartContainer(context.Context, *pluginapi.PreStartContainerRequest) (*pluginapi.PreStartContainerResponse, error)

func (*NvidiaDevicePlugin) Register

func (m *NvidiaDevicePlugin) Register(kubeletEndpoint, resourceName string) error

Register registers the device plugin for the given resourceName with Kubelet.

func (*NvidiaDevicePlugin) Serve

func (m *NvidiaDevicePlugin) Serve() error

Serve starts the gRPC server and register the device plugin to Kubelet

func (*NvidiaDevicePlugin) Start

func (m *NvidiaDevicePlugin) Start() error

Start starts the gRPC server of the device plugin

func (*NvidiaDevicePlugin) Stop

func (m *NvidiaDevicePlugin) Stop() error

Stop stops the gRPC server