Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

scheduler:support multi gpu share #2127

Merged
merged 1 commit into from
Jul 30, 2024
Merged

scheduler:support multi gpu share #2127

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
4 changes: 4 additions & 0 deletions pkg/features/features.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,6 +70,9 @@ const (

// EnableQuotaAdmission enables quota admission.
EnableQuotaAdmission featuregate.Feature = "EnableQuotaAdmission"

// Enable sync GPU shared resource from Device CRD
EnableSyncGPUSharedResource featuregate.Feature = "EnableSyncGPUSharedResource"
)

var defaultFeatureGates = map[featuregate.Feature]featuregate.FeatureSpec{
Expand All @@ -88,6 +91,7 @@ var defaultFeatureGates = map[featuregate.Feature]featuregate.FeatureSpec{
DisableDefaultQuota: {Default: false, PreRelease: featuregate.Alpha},
SupportParentQuotaSubmitPod: {Default: false, PreRelease: featuregate.Alpha},
EnableQuotaAdmission: {Default: false, PreRelease: featuregate.Alpha},
EnableSyncGPUSharedResource: {Default: true, PreRelease: featuregate.Alpha},
}

const (
Expand Down
18 changes: 13 additions & 5 deletions pkg/scheduler/plugins/deviceshare/devicehandler_gpu.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,17 +51,25 @@ func (h *GPUHandler) CalcDesiredRequestsAndCount(node *corev1.Node, pod *corev1.
requests := podRequests
desiredCount := int64(1)

memoryRatio := podRequests[apiext.ResourceGPUMemoryRatio]
multiDevices := memoryRatio.Value() > 100 && memoryRatio.Value()%100 == 0
if multiDevices {
gpuCore, gpuMem, gpuMemoryRatio := podRequests[apiext.ResourceGPUCore], podRequests[apiext.ResourceGPUMemory], podRequests[apiext.ResourceGPUMemoryRatio]
desiredCount = gpuMemoryRatio.Value() / 100
gpuShare, ok := podRequests[apiext.ResourceGPUShared]
gpuCore, gpuMem, gpuMemoryRatio := podRequests[apiext.ResourceGPUCore], podRequests[apiext.ResourceGPUMemory], podRequests[apiext.ResourceGPUMemoryRatio]
// gpu share mode
if ok && gpuShare.Value() > 0 {
desiredCount = gpuShare.Value()
} else {
if gpuMemoryRatio.Value() > 100 && gpuMemoryRatio.Value()%100 == 0 {
desiredCount = gpuMemoryRatio.Value() / 100
}
}

if desiredCount > 1 {
requests = corev1.ResourceList{
apiext.ResourceGPUCore: *resource.NewQuantity(gpuCore.Value()/desiredCount, resource.DecimalSI),
apiext.ResourceGPUMemory: *resource.NewQuantity(gpuMem.Value()/desiredCount, resource.BinarySI),
apiext.ResourceGPUMemoryRatio: *resource.NewQuantity(gpuMemoryRatio.Value()/desiredCount, resource.DecimalSI),
}
}

return requests, int(desiredCount), nil
}

Expand Down
217 changes: 216 additions & 1 deletion pkg/scheduler/plugins/deviceshare/plugin_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -597,7 +597,7 @@ func Test_Plugin_PreFilter(t *testing.T) {
},
},
},
wantStatus: framework.NewStatus(framework.UnschedulableAndUnresolvable, fmt.Sprintf("invalid resource unit %v: 101", apiext.ResourceGPUMemoryRatio)),
wantStatus: framework.NewStatus(framework.UnschedulableAndUnresolvable, fmt.Sprintf("invalid resource device requests: [%s]", apiext.ResourceGPUMemoryRatio)),
},
{
name: "pod has valid gpu request 1",
Expand Down Expand Up @@ -1848,6 +1848,221 @@ func Test_Plugin_Filter(t *testing.T) {
nodeInfo: testNodeInfo,
want: nil,
},
{
name: "pod stuck when use multi gpu",
state: &preFilterState{
skip: false,
podRequests: map[schedulingv1alpha1.DeviceType]corev1.ResourceList{
schedulingv1alpha1.GPU: {
apiext.ResourceGPUShared: resource.MustParse("4"),
apiext.ResourceGPUMemory: resource.MustParse("160G"),
},
},
},
// reserved: apiext.DeviceAllocations{},
nodeDeviceCache: &nodeDeviceCache{
nodeDeviceInfos: map[string]*nodeDevice{
"test-node": {
allocateSet: map[schedulingv1alpha1.DeviceType]map[types.NamespacedName]deviceResources{},
deviceFree: map[schedulingv1alpha1.DeviceType]deviceResources{
schedulingv1alpha1.GPU: {
0: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
1: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
2: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
3: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
4: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
5: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
6: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
7: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
8: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
},
deviceTotal: map[schedulingv1alpha1.DeviceType]deviceResources{
schedulingv1alpha1.GPU: {
0: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
1: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
2: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
3: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
4: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
5: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
6: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
7: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
8: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
},
deviceUsed: map[schedulingv1alpha1.DeviceType]deviceResources{},
vfAllocations: map[schedulingv1alpha1.DeviceType]*VFAllocation{},
numaTopology: &NUMATopology{},
deviceInfos: map[schedulingv1alpha1.DeviceType][]*schedulingv1alpha1.DeviceInfo{
schedulingv1alpha1.GPU: {
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-0",
Minor: pointer.Int32(0),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-1",
Minor: pointer.Int32(1),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-2",
Minor: pointer.Int32(2),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-3",
Minor: pointer.Int32(3),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-4",
Minor: pointer.Int32(4),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-5",
Minor: pointer.Int32(5),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-6",
Minor: pointer.Int32(6),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
{
Type: schedulingv1alpha1.GPU,
Health: true,
UUID: "123456-7",
Minor: pointer.Int32(7),
Resources: corev1.ResourceList{
apiext.ResourceGPUCore: resource.MustParse("100"),
apiext.ResourceGPUMemoryRatio: resource.MustParse("100"),
apiext.ResourceGPUMemory: resource.MustParse("80Gi"),
},
},
},
},
},
},
},
nodeInfo: testNodeInfo,
want: nil,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
Expand Down