Allow deprecated beta topology labels to be applied for those not rea…

…dy to migrate

This reverts commit 67c2407.
It then introduces a new flag to allow users to opt-in to syncing older, beta
topology labels based on the presence of the GA topolgoy labels.

cmd/cloud-node-manager/app/config/config.go

@@ -81,4 +81,9 @@ type Config struct {
 	// WindowsService should be set to true if cloud-node-manager is running as a service on Windows.
 	// Its corresponding flag only gets registered in Windows builds
 	WindowsService bool
+
+	// EnableDeprecatedBetaTopologyLabels indicates whether the node should apply beta topology labels.
+	// If true, the node will apply beta topology labels.
+	// DEPRECATED: This flag will be removed in a future release.
+	EnableDeprecatedBetaTopologyLabels bool
 }

cmd/cloud-node-manager/app/nodemanager.go

@@ -137,7 +137,8 @@ func startControllers(ctx context.Context, c *cloudnodeconfig.Config, stopCh <-c
 		c.ClientBuilder.ClientOrDie("node-controller"),
 		nodeprovider.NewNodeProvider(ctx, c.UseInstanceMetadata, c.CloudConfigFilePath),
 		c.NodeStatusUpdateFrequency.Duration,
-		c.WaitForRoutes)
+		c.WaitForRoutes,
+		c.EnableDeprecatedBetaTopologyLabels)
 
 	go nodeController.Run(stopCh)
 

cmd/cloud-node-manager/app/options/options.go

@@ -86,6 +86,11 @@ type CloudNodeManagerOptions struct {
 	// WindowsService should be set to true if cloud-node-manager is running as a service on Windows.
 	// Its corresponding flag only gets registered in Windows builds
 	WindowsService bool
+
+	// EnableDeprecatedBetaTopologyLabels indicates whether the node should apply beta topology labels.
+	// If true, the node will apply beta topology labels.
+	// DEPRECATED: This flag will be removed in a future release.
+	EnableDeprecatedBetaTopologyLabels bool
 }
 
 // NewCloudNodeManagerOptions creates a new CloudNodeManagerOptions with a default config.
@@ -131,6 +136,7 @@ func (o *CloudNodeManagerOptions) Flags() cliflag.NamedFlagSets {
 	fs.BoolVar(&o.WaitForRoutes, "wait-routes", false, "Whether the nodes should wait for routes created on Azure route table. It should be set to true when using kubenet plugin.")
 	fs.BoolVar(&o.UseInstanceMetadata, "use-instance-metadata", true, "Should use Instance Metadata Service for fetching node information; if false will use ARM instead.")
 	fs.StringVar(&o.CloudConfigFilePath, "cloud-config", o.CloudConfigFilePath, "The path to the cloud config file to be used when using ARM to fetch node information.")
+	fs.BoolVar(&o.EnableDeprecatedBetaTopologyLabels, "enable-deprecated-beta-topology-labels", o.EnableDeprecatedBetaTopologyLabels, "DEPRECATED: This flag will be removed in a future release. If true, the node will apply beta topology labels.")
 	return fss
 }
 
@@ -193,6 +199,9 @@ func (o *CloudNodeManagerOptions) ApplyTo(c *cloudnodeconfig.Config, userAgent s
 
 	c.WindowsService = o.WindowsService
 
+	// Allow users to choose to apply beta topology labels until they are removed by all cloud providers.
+	c.EnableDeprecatedBetaTopologyLabels = o.EnableDeprecatedBetaTopologyLabels
+
 	return nil
 }
 

pkg/nodemanager/nodemanager.go

@@ -63,17 +63,44 @@ type NodeProvider interface {
 	GetPlatformSubFaultDomain() (string, error)
 }
 
-// labelReconcileInfo lists Node labels to reconcile, and how to reconcile them.
+// labelReconcile holds information about a label to reconcile and how to reconcile it.
 // primaryKey and secondaryKey are keys of labels to reconcile.
 // - If both keys exist, but their values don't match. Use the value from the
 // primaryKey as the source of truth to reconcile.
 // - If ensureSecondaryExists is true, and the secondaryKey does not
 // exist, secondaryKey will be added with the value of the primaryKey.
-var labelReconcileInfo = []struct {
+type labelReconcile struct {
 	primaryKey            string
 	secondaryKey          string
 	ensureSecondaryExists bool
-}{}
+}
+
+// betaToplogyLabels lists beta topology labels that are deprecated and will
+// be removed in a future release.
+// For now we reconcile them optionally onto nodes.
+var betaToplogyLabels = []labelReconcile{
+	{
+		// Reconcile the beta and the GA zone label using the GA label as
+		// the source of truth
+		primaryKey:            v1.LabelZoneFailureDomainStable,
+		secondaryKey:          v1.LabelZoneFailureDomain,
+		ensureSecondaryExists: true,
+	},
+	{
+		// Reconcile the beta and the stable region label using the GA label as
+		// the source of truth
+		primaryKey:            v1.LabelZoneRegionStable,
+		secondaryKey:          v1.LabelZoneRegion,
+		ensureSecondaryExists: true,
+	},
+	{
+		// Reconcile the beta and the stable instance-type label using the GA label as
+		// the source of truth
+		primaryKey:            v1.LabelInstanceTypeStable,
+		secondaryKey:          v1.LabelInstanceType,
+		ensureSecondaryExists: true,
+	},
+}
 
 // UpdateNodeSpecBackoff is the back configure for node update.
 var UpdateNodeSpecBackoff = wait.Backoff{
@@ -103,6 +130,10 @@ type CloudNodeController struct {
 	recorder      record.EventRecorder
 
 	nodeStatusUpdateFrequency time.Duration
+
+	labelReconcileInfo []labelReconcile
+
+	enableBetaTopologyLabels bool
 }
 
 // NewCloudNodeController creates a CloudNodeController object
@@ -112,7 +143,7 @@ func NewCloudNodeController(
 	kubeClient clientset.Interface,
 	nodeProvider NodeProvider,
 	nodeStatusUpdateFrequency time.Duration,
-	waitForRoutes bool) *CloudNodeController {
+	waitForRoutes, enableBetaTopologyLabels bool) *CloudNodeController {
 
 	eventBroadcaster := record.NewBroadcaster()
 	recorder := eventBroadcaster.NewRecorder(scheme.Scheme, v1.EventSource{Component: "cloud-node-controller"})
@@ -132,6 +163,12 @@ func NewCloudNodeController(
 		nodeProvider:              nodeProvider,
 		waitForRoutes:             waitForRoutes,
 		nodeStatusUpdateFrequency: nodeStatusUpdateFrequency,
+		enableBetaTopologyLabels:  enableBetaTopologyLabels,
+	}
+
+	// Only reconcile the beta toplogy labels when the feature flag is enabled.
+	if cnc.enableBetaTopologyLabels {
+		cnc.labelReconcileInfo = append(cnc.labelReconcileInfo, betaToplogyLabels...)
 	}
 
 	// Use shared informer to listen to add/update of nodes. Note that any nodes
@@ -190,7 +227,7 @@ func (cnc *CloudNodeController) reconcileNodeLabels(node *v1.Node) error {
 	}
 
 	labelsToUpdate := map[string]string{}
-	for _, r := range labelReconcileInfo {
+	for _, r := range cnc.labelReconcileInfo {
 		primaryValue, primaryExists := node.Labels[r.primaryKey]
 		secondaryValue, secondaryExists := node.Labels[r.secondaryKey]
 
@@ -445,53 +482,51 @@ func (cnc *CloudNodeController) getNodeModifiersFromCloudProvider(ctx context.Co
 	if instanceType, err := cnc.getInstanceTypeByName(ctx, node); err != nil {
 		return nil, err
 	} else if instanceType != "" {
-		klog.V(2).Infof("Adding node label from cloud provider: %s=%s", v1.LabelInstanceTypeStable, instanceType)
-		nodeModifiers = append(nodeModifiers, func(n *v1.Node) {
-			if n.Labels == nil {
-				n.Labels = map[string]string{}
-			}
-			n.Labels[v1.LabelInstanceTypeStable] = instanceType
-		})
+		nodeModifiers = append(nodeModifiers, addCloudNodeLabel(v1.LabelInstanceTypeStable, instanceType))
+		if cnc.enableBetaTopologyLabels {
+			nodeModifiers = append(nodeModifiers, addCloudNodeLabel(v1.LabelInstanceType, instanceType))
+		}
 	}
+
 	zone, err := cnc.getZoneByName(ctx, node)
 	if err != nil {
 		return nil, fmt.Errorf("failed to get zone from cloud provider: %w", err)
 	}
 	if zone.FailureDomain != "" {
-		klog.V(2).Infof("Adding node label from cloud provider: %s=%s", v1.LabelZoneFailureDomainStable, zone.FailureDomain)
-		nodeModifiers = append(nodeModifiers, func(n *v1.Node) {
-			if n.Labels == nil {
-				n.Labels = map[string]string{}
-			}
-			n.Labels[v1.LabelZoneFailureDomainStable] = zone.FailureDomain
-		})
+		nodeModifiers = append(nodeModifiers, addCloudNodeLabel(v1.LabelZoneFailureDomainStable, zone.FailureDomain))
+		if cnc.enableBetaTopologyLabels {
+			nodeModifiers = append(nodeModifiers, addCloudNodeLabel(v1.LabelZoneFailureDomain, zone.FailureDomain))
+		}
 	}
 	if zone.Region != "" {
-		klog.V(2).Infof("Adding node label from cloud provider: %s=%s", v1.LabelZoneRegionStable, zone.Region)
-		nodeModifiers = append(nodeModifiers, func(n *v1.Node) {
-			if n.Labels == nil {
-				n.Labels = map[string]string{}
-			}
-			n.Labels[v1.LabelZoneRegionStable] = zone.Region
-		})
+		nodeModifiers = append(nodeModifiers, addCloudNodeLabel(v1.LabelZoneRegionStable, zone.Region))
+		if cnc.enableBetaTopologyLabels {
+			nodeModifiers = append(nodeModifiers, addCloudNodeLabel(v1.LabelZoneRegion, zone.Region))
+		}
 	}
+
 	platformSubFaultDomain, err := cnc.getPlatformSubFaultDomain()
 	if err != nil {
 		return nil, fmt.Errorf("failed to get platformSubFaultDomain: %w", err)
 	}
 	if platformSubFaultDomain != "" {
-		klog.V(2).Infof("Adding node label from cloud provider: %s=%s", consts.LabelPlatformSubFaultDomain, platformSubFaultDomain)
-		nodeModifiers = append(nodeModifiers, func(n *v1.Node) {
-			if n.Labels == nil {
-				n.Labels = map[string]string{}
-			}
-			n.Labels[consts.LabelPlatformSubFaultDomain] = platformSubFaultDomain
-		})
+		nodeModifiers = append(nodeModifiers, addCloudNodeLabel(consts.LabelPlatformSubFaultDomain, platformSubFaultDomain))
 	}
 
 	return nodeModifiers, nil
 }
 
+// addCloudNodeLabel creates a nodeModifier that adds a label to a node.
+func addCloudNodeLabel(key, value string) func(*v1.Node) {
+	klog.V(2).Infof("Adding node label from cloud provider: %s=%s", key, value)
+	return func(node *v1.Node) {
+		if node.Labels == nil {
+			node.Labels = map[string]string{}
+		}
+		node.Labels[key] = value
+	}
+}
+
 func GetCloudTaint(taints []v1.Taint) *v1.Taint {
 	for _, taint := range taints {
 		taint := taint