Add GetNodeGpuConfig to cloud provider

* Added GetNodeGpuConfig to cloud provider which returns a GpuConfig
  struct containing the gpu label, type and resource name if the node
  has a GPU.
* Added initial implementaion of the GetNodeGpuConfig to all cloud
  providers.

cluster-autoscaler/cloudprovider/alicloud/alicloud_cloud_provider.go

@@ -27,6 +27,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/config/dynamic"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 )
 
@@ -109,6 +110,12 @@ func (ali *aliCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (ali *aliCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(ali, node)
+}
+
 func (ali *aliCloudProvider) NodeGroups() []cloudprovider.NodeGroup {
 	result := make([]cloudprovider.NodeGroup, 0, len(ali.asgs))
 	for _, asg := range ali.asgs {

cluster-autoscaler/cloudprovider/aws/aws_cloud_provider.go

@@ -28,6 +28,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 	schedulerframework "k8s.io/kubernetes/pkg/scheduler/framework"
 )
@@ -84,6 +85,12 @@ func (aws *awsCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (aws *awsCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(aws, node)
+}
+
 // NodeGroups returns all node groups configured for this cloud provider.
 func (aws *awsCloudProvider) NodeGroups() []cloudprovider.NodeGroup {
 	asgs := aws.awsManager.getAsgs()

cluster-autoscaler/cloudprovider/azure/azure_cloud_provider.go

@@ -25,6 +25,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 )
 
@@ -79,6 +80,13 @@ func (azure *AzureCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (azure *AzureCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(azure, node)
+
+}
+
 // NodeGroups returns all node groups configured for this cloud provider.
 func (azure *AzureCloudProvider) NodeGroups() []cloudprovider.NodeGroup {
 	asgs := azure.azureManager.getNodeGroups()

cluster-autoscaler/cloudprovider/baiducloud/baiducloud_cloud_provider.go

@@ -28,6 +28,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/config/dynamic"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 	schedulerframework "k8s.io/kubernetes/pkg/scheduler/framework"
 )
@@ -168,6 +169,12 @@ func (baiducloud *baiducloudCloudProvider) GetAvailableGPUTypes() map[string]str
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (baiducloud *baiducloudCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(baiducloud, node)
+}
+
 // NodeGroupForNode returns the node group for the given node, nil if the node
 // should not be processed by cluster autoscaler, or non-nil error if such
 // occurred. Must be implemented.

cluster-autoscaler/cloudprovider/bizflycloud/bizflycloud_cloud_provider.go

@@ -27,6 +27,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 )
 
@@ -151,6 +152,12 @@ func (d *bizflycloudCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return nil
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (d *bizflycloudCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(d, node)
+}
+
 // Cleanup cleans up open resources before the cloud provider is destroyed,
 // i.e. go routines etc.
 func (d *bizflycloudCloudProvider) Cleanup() error {

cluster-autoscaler/cloudprovider/brightbox/brightbox_cloud_provider.go

@@ -30,6 +30,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider/brightbox/k8ssdk"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 )
 
@@ -204,6 +205,13 @@ func (b *brightboxCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (b *brightboxCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	klog.V(4).Info("GetNodeGpuConfig")
+	return gpu.GetNodeGPUFromCloudProvider(b, node)
+}
+
 // Cleanup cleans up open resources before the cloud provider is
 // destroyed, i.e. go routines etc.
 func (b *brightboxCloudProvider) Cleanup() error {

cluster-autoscaler/cloudprovider/cherryservers/cherry_cloud_provider.go

@@ -27,6 +27,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/config/dynamic"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 )
 
@@ -85,6 +86,12 @@ func (ccp *cherryCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (ccp *cherryCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(ccp, node)
+}
+
 // NodeGroups returns all node groups managed by this cloud provider.
 func (ccp *cherryCloudProvider) NodeGroups() []cloudprovider.NodeGroup {
 	groups := make([]cloudprovider.NodeGroup, len(ccp.nodeGroups))

cluster-autoscaler/cloudprovider/civo/civo_cloud_provider.go

@@ -27,6 +27,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	"k8s.io/klog/v2"
 )
 
@@ -146,6 +147,12 @@ func (d *civoCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return nil
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (d *civoCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(d, node)
+}
+
 // Cleanup cleans up open resources before the cloud provider is destroyed,
 // i.e. go routines etc.
 func (d *civoCloudProvider) Cleanup() error {

cluster-autoscaler/cloudprovider/cloud_provider.go

@@ -86,6 +86,13 @@ const (
 	RancherProviderName = "rancher"
 )
 
+// GpuConfig contains the label, type and the resource name for an accelerator (e.g. gpu)
+type GpuConfig struct {
+	Label        string
+	Type         string
+	ResourceName apiv1.ResourceName
+}
+
 // CloudProvider contains configuration info and functions for interacting with
 // cloud provider (GCE, AWS, etc).
 type CloudProvider interface {
@@ -127,6 +134,10 @@ type CloudProvider interface {
 	// GetAvailableGPUTypes return all available GPU types cloud provider supports.
 	GetAvailableGPUTypes() map[string]struct{}
 
+	// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+	// any GPUs, it returns nil.
+	GetNodeGpuConfig(*apiv1.Node) *GpuConfig
+
 	// Cleanup cleans up open resources before the cloud provider is destroyed, i.e. go routines etc.
 	Cleanup() error
 

cluster-autoscaler/cloudprovider/cloudstack/cloudstack_cloud_provider.go

@@ -25,6 +25,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 
 	v1 "k8s.io/api/core/v1"
 	klog "k8s.io/klog/v2"
@@ -103,6 +104,12 @@ func (provider *cloudStackCloudProvider) GetAvailableGPUTypes() map[string]struc
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (provider *cloudStackCloudProvider) GetNodeGpuConfig(node *v1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(provider, node)
+}
+
 // Pricing returns pricing model for this cloud provider or error if not available.
 func (provider *cloudStackCloudProvider) Pricing() (cloudprovider.PricingModel, errors.AutoscalerError) {
 	return nil, cloudprovider.ErrNotImplemented

cluster-autoscaler/cloudprovider/clusterapi/clusterapi_provider.go

@@ -33,6 +33,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 )
 
 const (
@@ -128,6 +129,12 @@ func (p *provider) GPULabel() string {
 	return GPULabel
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (p *provider) GetNodeGpuConfig(node *corev1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(p, node)
+}
+
 func newProvider(
 	name string,
 	rl *cloudprovider.ResourceLimiter,

cluster-autoscaler/cloudprovider/digitalocean/digitalocean_cloud_provider.go

@@ -27,6 +27,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	"k8s.io/klog/v2"
 )
 
@@ -148,6 +149,12 @@ func (d *digitaloceanCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return nil
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (d *digitaloceanCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(d, node)
+}
+
 // Cleanup cleans up open resources before the cloud provider is destroyed,
 // i.e. go routines etc.
 func (d *digitaloceanCloudProvider) Cleanup() error {

cluster-autoscaler/cloudprovider/exoscale/exoscale_cloud_provider.go

@@ -25,6 +25,7 @@ import (
 	egoscale "k8s.io/autoscaler/cluster-autoscaler/cloudprovider/exoscale/internal/github.com/exoscale/egoscale/v2"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 )
 
 var _ cloudprovider.CloudProvider = (*exoscaleCloudProvider)(nil)
@@ -176,6 +177,12 @@ func (e *exoscaleCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return nil
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (e *exoscaleCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(e, node)
+}
+
 // Cleanup cleans up open resources before the cloud provider is destroyed, i.e. go routines etc.
 func (e *exoscaleCloudProvider) Cleanup() error {
 	return nil

cluster-autoscaler/cloudprovider/externalgrpc/externalgrpc_cloud_provider.go

@@ -36,6 +36,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider/externalgrpc/protos"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 )
 
@@ -262,6 +263,12 @@ func (e *externalGrpcCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return gpuTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (e *externalGrpcCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(e, node)
+}
+
 // Cleanup cleans up open resources before the cloud provider is destroyed, i.e. go routines etc.
 func (e *externalGrpcCloudProvider) Cleanup() error {
 	ctx, cancel := context.WithTimeout(context.Background(), grpcTimeout)

cluster-autoscaler/cloudprovider/gce/gce_cloud_provider.go

@@ -27,6 +27,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 	schedulerframework "k8s.io/kubernetes/pkg/scheduler/framework"
 )
@@ -80,6 +81,12 @@ func (gce *GceCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (gce *GceCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(gce, node)
+}
+
 // NodeGroups returns all node groups configured for this cloud provider.
 func (gce *GceCloudProvider) NodeGroups() []cloudprovider.NodeGroup {
 	migs := gce.gceManager.GetMigs()

cluster-autoscaler/cloudprovider/hetzner/hetzner_cloud_provider.go

@@ -29,6 +29,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	"k8s.io/klog/v2"
 )
 
@@ -156,6 +157,12 @@ func (d *HetznerCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return nil
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (d *HetznerCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(d, node)
+}
+
 // Cleanup cleans up open resources before the cloud provider is destroyed,
 // i.e. go routines etc.
 func (d *HetznerCloudProvider) Cleanup() error {

cluster-autoscaler/cloudprovider/huaweicloud/huaweicloud_cloud_provider.go

@@ -26,6 +26,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/config/dynamic"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	klog "k8s.io/klog/v2"
 )
 
@@ -160,6 +161,12 @@ func (hcp *huaweicloudCloudProvider) GetAvailableGPUTypes() map[string]struct{}
 	return availableGPUTypes
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (hcp *huaweicloudCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(hcp, node)
+}
+
 // Cleanup currently does nothing.
 func (hcp *huaweicloudCloudProvider) Cleanup() error {
 	return nil

cluster-autoscaler/cloudprovider/ionoscloud/ionoscloud_cloud_provider.go

@@ -24,6 +24,7 @@ import (
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/errors"
+	"k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	"k8s.io/klog/v2"
 	schedulerframework "k8s.io/kubernetes/pkg/scheduler/framework"
 )
@@ -279,6 +280,12 @@ func (ic *IonosCloudCloudProvider) GetAvailableGPUTypes() map[string]struct{} {
 	return nil
 }
 
+// GetNodeGpuConfig returns the label, type and resource name for the GPU added to node. If node doesn't have
+// any GPUs, it returns nil.
+func (ic *IonosCloudCloudProvider) GetNodeGpuConfig(node *apiv1.Node) *cloudprovider.GpuConfig {
+	return gpu.GetNodeGPUFromCloudProvider(ic, node)
+}
+
 // Cleanup cleans up read resources before the cloud provider is destroyed,
 // i.e. go routines etc.
 func (ic *IonosCloudCloudProvider) Cleanup() error {