nvidia

package

v1.1.5 Latest Latest Go to latest Published: May 10, 2021 License: Apache-2.0, BSD-2-Clause, BSD-3-Clause, + 8 more Imports: 32 Imported by: 0

Details

Valid go.mod file

The Go module system was introduced in Go 1.11 and is the official dependency management solution for Go.
Redistributable license

Redistributable licenses place minimal restrictions on how software can be used, modified, and redistributed.
Tagged version

Modules with tagged versions give importers more predictable builds.
Stable version

When a project reaches major version v1 it is considered stable.
Learn more about best practices

Repository

github.com/tkestack/gpu-manager

Links

Open Source Insights

Documentation ¶

Index ¶

Constants
func NewNvidiaTopoAllocator(config *config.Config, tree device.GPUTree, k8sClient kubernetes.Interface) allocator.GPUTopoService
func NewNvidiaTopoAllocatorForTest(config *config.Config, tree device.GPUTree, k8sClient kubernetes.Interface) allocator.GPUTopoService
func OrderPodsdByPredicateTime(pods []*v1.Pod) []*v1.Pod
type Evaluator
type NvidiaTopoAllocator
type PodsOrderedByPredicateTime

Constants ¶

View Source

const (
	ALLOCATE_SUCCESS = iota
	ALLOCATE_FAIL
	PREDICATE_MISSING
)

Variables ¶

This section is empty.

Functions ¶

func NewNvidiaTopoAllocator ¶

func NewNvidiaTopoAllocator(config *config.Config, tree device.GPUTree, k8sClient kubernetes.Interface) allocator.GPUTopoService

NewNvidiaTopoAllocator returns a new NvidiaTopoAllocator

func NewNvidiaTopoAllocatorForTest ¶

func NewNvidiaTopoAllocatorForTest(config *config.Config, tree device.GPUTree, k8sClient kubernetes.Interface) allocator.GPUTopoService

NewNvidiaTopoAllocatorForTest returns a new NvidiaTopoAllocator with fake docker client, just for testing.

func OrderPodsdByPredicateTime ¶

func OrderPodsdByPredicateTime(pods []*v1.Pod) []*v1.Pod

make the pod ordered by predicate time

Types ¶

type NvidiaTopoAllocator ¶

type NvidiaTopoAllocator struct {
	sync.Mutex
	// contains filtered or unexported fields
}

NvidiaTopoAllocator is an allocator for Nvidia GPU

func (*NvidiaTopoAllocator) Allocate ¶

func (ta *NvidiaTopoAllocator) Allocate(_ context.Context, reqs *pluginapi.AllocateRequest) (*pluginapi.AllocateResponse, error)

#lizard forgives Allocate tries to allocate GPU node for each request

func (*NvidiaTopoAllocator) GetDevicePluginOptions ¶

func (ta *NvidiaTopoAllocator) GetDevicePluginOptions(ctx context.Context, e *pluginapi.Empty) (*pluginapi.DevicePluginOptions, error)

GetDevicePluginOptions returns empty DevicePluginOptions

func (*NvidiaTopoAllocator) ListAndWatch ¶

func (ta *NvidiaTopoAllocator) ListAndWatch(e *pluginapi.Empty, s pluginapi.DevicePlugin_ListAndWatchServer) error

ListAndWatch is not implement

func (*NvidiaTopoAllocator) ListAndWatchWithResourceName ¶

func (ta *NvidiaTopoAllocator) ListAndWatchWithResourceName(resourceName string, e *pluginapi.Empty, s pluginapi.DevicePlugin_ListAndWatchServer) error

ListAndWatchWithResourceName send devices for request resource back to server

func (*NvidiaTopoAllocator) PreStartContainer ¶

func (ta *NvidiaTopoAllocator) PreStartContainer(ctx context.Context, req *pluginapi.PreStartContainerRequest) (*pluginapi.PreStartContainerResponse, error)

PreStartContainer find the podUID by comparing request deviceids with deviceplugin checkpoint data, then checks the validation of allocation of the pod. Update pod annotation if check success, otherwise evict the pod.

type PodsOrderedByPredicateTime ¶

type PodsOrderedByPredicateTime []*v1.Pod

func (PodsOrderedByPredicateTime) Len ¶

func (pods PodsOrderedByPredicateTime) Len() int

func (PodsOrderedByPredicateTime) Less ¶

func (pods PodsOrderedByPredicateTime) Less(i, j int) bool

func (PodsOrderedByPredicateTime) Swap ¶

func (pods PodsOrderedByPredicateTime) Swap(i, j int)

Source Files ¶

View all Source files

?	: This menu
/	: Search site
f or F	: Jump to
y or Y	: Canonical URL