Merge pull request #6447 from Jont828/edge-zone

Azure: add support for edge zones

Author: k8s-ci-robot

charts/cluster-autoscaler/README.md

@@ -382,6 +382,7 @@ vpa:
 | awsSecretAccessKey | string | `""` | AWS access secret key ([if AWS user keys used](https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/cloudprovider/aws/README.md#using-aws-credentials)) |
 | azureClientID | string | `""` | Service Principal ClientID with contributor permission to Cluster and Node ResourceGroup. Required if `cloudProvider=azure` |
 | azureClientSecret | string | `""` | Service Principal ClientSecret with contributor permission to Cluster and Node ResourceGroup. Required if `cloudProvider=azure` |
+| azureEnableForceDelete | bool | `false` | Whether to force delete VMs or VMSS instances when scaling down. |
 | azureResourceGroup | string | `""` | Azure resource group that the cluster is located. Required if `cloudProvider=azure` |
 | azureSubscriptionID | string | `""` | Azure subscription where the resources are located. Required if `cloudProvider=azure` |
 | azureTenantID | string | `""` | Azure tenant where the resources are located. Required if `cloudProvider=azure` |

charts/cluster-autoscaler/templates/deployment.yaml

@@ -166,6 +166,8 @@ spec:
                 secretKeyRef:
                   key: VMType
                   name: {{ default (include "cluster-autoscaler.fullname" .) .Values.secretKeyRefNameOverride }}
+            - name: AZURE_ENABLE_FORCE_DELETE
+              value: "{{ .Values.azureEnableForceDelete }}"
             {{- if .Values.azureUseWorkloadIdentityExtension }}
             - name: ARM_USE_WORKLOAD_IDENTITY_EXTENSION
               value: "true"

charts/cluster-autoscaler/values.yaml

@@ -96,6 +96,9 @@ azureUseWorkloadIdentityExtension: false
 # azureVMType -- Azure VM type.
 azureVMType: "vmss"
 
+# azureEnableForceDelete -- Whether to force delete VMs or VMSS instances when scaling down.
+azureEnableForceDelete: false
+
 # cloudConfigPath -- Configuration file for cloud provider.
 cloudConfigPath: ""
 

cluster-autoscaler/cloudprovider/azure/azure_cloud_provider_test.go

@@ -53,6 +53,7 @@ func newTestAzureManager(t *testing.T) *AzureManager {
 			VMType:              vmTypeVMSS,
 			MaxDeploymentsCount: 2,
 			Deployment:          "deployment",
+			EnableForceDelete:   true,
 		},
 		azClient: &azClient{
 			virtualMachineScaleSetsClient:   mockVMSSClient,

cluster-autoscaler/cloudprovider/azure/azure_config.go

@@ -130,6 +130,9 @@ type Config struct {
 	CloudProviderBackoffDuration int     `json:"cloudProviderBackoffDuration,omitempty" yaml:"cloudProviderBackoffDuration,omitempty"`
 	CloudProviderBackoffJitter   float64 `json:"cloudProviderBackoffJitter,omitempty" yaml:"cloudProviderBackoffJitter,omitempty"`
 
+	// EnableForceDelete defines whether to enable force deletion on the APIs
+	EnableForceDelete bool `json:"enableForceDelete,omitempty" yaml:"enableForceDelete,omitempty"`
+
 	// EnableDynamicInstanceList defines whether to enable dynamic instance workflow for instance information check
 	EnableDynamicInstanceList bool `json:"enableDynamicInstanceList,omitempty" yaml:"enableDynamicInstanceList,omitempty"`
 
@@ -303,6 +306,13 @@ func BuildAzureConfig(configReader io.Reader) (*Config, error) {
 		}
 	}
 
+	if enableForceDelete := os.Getenv("AZURE_ENABLE_FORCE_DELETE"); enableForceDelete != "" {
+		cfg.EnableForceDelete, err = strconv.ParseBool(enableForceDelete)
+		if err != nil {
+			return nil, fmt.Errorf("failed to parse AZURE_ENABLE_FORCE_DELETE: %q, %v", enableForceDelete, err)
+		}
+	}
+
 	err = initializeCloudProviderRateLimitConfig(&cfg.CloudProviderRateLimitConfig)
 	if err != nil {
 		return nil, err

cluster-autoscaler/cloudprovider/azure/azure_manager_test.go

@@ -679,6 +679,7 @@ func TestGetFilteredAutoscalingGroupsVmss(t *testing.T) {
 		minSize:                minVal,
 		maxSize:                maxVal,
 		manager:                manager,
+		enableForceDelete:      manager.config.EnableForceDelete,
 		curSize:                3,
 		sizeRefreshPeriod:      manager.azureCache.refreshInterval,
 		instancesRefreshPeriod: defaultVmssInstancesRefreshPeriod,

cluster-autoscaler/cloudprovider/azure/azure_scale_set.go

@@ -58,6 +58,8 @@ type ScaleSet struct {
 	minSize int
 	maxSize int
 
+	enableForceDelete bool
+
 	sizeMutex sync.Mutex
 	curSize   int64
 
@@ -87,6 +89,7 @@ func NewScaleSet(spec *dynamic.NodeGroupSpec, az *AzureManager, curSize int64) (
 		sizeRefreshPeriod:         az.azureCache.refreshInterval,
 		enableDynamicInstanceList: az.config.EnableDynamicInstanceList,
 		instancesRefreshJitter:    az.config.VmssVmsCacheJitter,
+		enableForceDelete:         az.config.EnableForceDelete,
 	}
 
 	if az.config.VmssVmsCacheTTL != 0 {
@@ -251,6 +254,16 @@ func (scaleSet *ScaleSet) SetScaleSetSize(size int64) error {
 		Sku:      vmssInfo.Sku,
 		Location: vmssInfo.Location,
 	}
+
+	if vmssInfo.ExtendedLocation != nil {
+		op.ExtendedLocation = &compute.ExtendedLocation{
+			Name: vmssInfo.ExtendedLocation.Name,
+			Type: vmssInfo.ExtendedLocation.Type,
+		}
+
+		klog.V(3).Infof("Passing ExtendedLocation information if it is not nil, with Edge Zone name:(%s)", *op.ExtendedLocation.Name)
+	}
+
 	ctx, cancel := getContextWithTimeout(vmssContextTimeout)
 	defer cancel()
 	klog.V(3).Infof("Waiting for virtualMachineScaleSetsClient.CreateOrUpdateAsync(%s)", scaleSet.Name)
@@ -437,8 +450,15 @@ func (scaleSet *ScaleSet) DeleteInstances(instances []*azureRef, hasUnregistered
 	resourceGroup := scaleSet.manager.config.ResourceGroup
 
 	scaleSet.instanceMutex.Lock()
-	klog.V(3).Infof("Calling virtualMachineScaleSetsClient.DeleteInstancesAsync(%v)", requiredIds.InstanceIds)
-	future, rerr := scaleSet.manager.azClient.virtualMachineScaleSetsClient.DeleteInstancesAsync(ctx, resourceGroup, commonAsg.Id(), *requiredIds, false)
+	klog.V(3).Infof("Calling virtualMachineScaleSetsClient.DeleteInstancesAsync(%v), force delete set to %v", requiredIds.InstanceIds, scaleSet.enableForceDelete)
+	future, rerr := scaleSet.manager.azClient.virtualMachineScaleSetsClient.DeleteInstancesAsync(ctx, resourceGroup, commonAsg.Id(), *requiredIds, scaleSet.enableForceDelete)
+
+	if scaleSet.enableForceDelete && isOperationNotAllowed(rerr) {
+		klog.Infof("falling back to normal delete for instances %v for %s", requiredIds.InstanceIds, scaleSet.Name)
+		future, rerr = scaleSet.manager.azClient.virtualMachineScaleSetsClient.DeleteInstancesAsync(ctx, resourceGroup,
+			commonAsg.Id(), *requiredIds, false)
+	}
+
 	scaleSet.instanceMutex.Unlock()
 	if rerr != nil {
 		klog.Errorf("virtualMachineScaleSetsClient.DeleteInstancesAsync for instances %v failed: %v", requiredIds.InstanceIds, rerr)
@@ -746,3 +766,7 @@ func (scaleSet *ScaleSet) getOrchestrationMode() (compute.OrchestrationMode, err
 	}
 	return vmss.OrchestrationMode, nil
 }
+
+func isOperationNotAllowed(rerr *retry.Error) bool {
+	return rerr != nil && rerr.ServiceErrorCode() == retry.OperationNotAllowed
+}