Update cluster-api provider to use machineTemplate.status.nodeInfo for architecture-aware autoscale from zero

kubernetes-sigs/cluster-api#11962 introduced the nodeInfo field for MachineTemplates. Providers can reconcile this field in the status subresource to inform the autoscaler about the architecture and operating system that the MachineTemplate's nodes will run.

Previously, we have been implementing this behavior in the cluster autoscaler by leveraging the labels capacity annotation and, as a fallback, default values set in environment variables at cluster-autoscaler deployment time.

With this commit, the cluster autoscaler computes the future architecture of a node with the following priority order:

- Labels set in existing nodes for not-autoscale-from-zero cases
- Labels set in the labels capacity annotation of machine template, machine set, and machine deployment.
- Values in the status.nodeSystemInfo of MachineTemplates
- Generic/default labels set in the environment of the cluster autoscaler

# Conflicts:
#	cluster-autoscaler/cloudprovider/clusterapi/clusterapi_unstructured.go

Author: aleskandro

cluster-autoscaler/cloudprovider/clusterapi/clusterapi_nodegroup.go

@@ -361,12 +361,17 @@ func (ng *nodegroup) TemplateNodeInfo() (*framework.NodeInfo, error) {
 		},
 	}
 
+	nsi := ng.scalableResource.InstanceSystemInfo()
+	if nsi != nil {
+		node.Status.NodeInfo = *nsi
+	}
+
 	node.Status.Capacity = capacity
 	node.Status.Allocatable = capacity
 	node.Status.Conditions = cloudprovider.BuildReadyConditions()
 	node.Spec.Taints = ng.scalableResource.Taints()
 
-	node.Labels, err = ng.buildTemplateLabels(nodeName)
+	node.Labels, err = ng.buildTemplateLabels(nodeName, nsi)
 	if err != nil {
 		return nil, err
 	}
@@ -380,8 +385,19 @@ func (ng *nodegroup) TemplateNodeInfo() (*framework.NodeInfo, error) {
 	return nodeInfo, nil
 }
 
-func (ng *nodegroup) buildTemplateLabels(nodeName string) (map[string]string, error) {
-	labels := cloudprovider.JoinStringMaps(buildGenericLabels(nodeName), ng.scalableResource.Labels())
+func (ng *nodegroup) buildTemplateLabels(nodeName string, nsi *corev1.NodeSystemInfo) (map[string]string, error) {
+	nsiLabels := make(map[string]string)
+	if nsi != nil {
+		nsiLabels[corev1.LabelArchStable] = nsi.Architecture
+		nsiLabels[corev1.LabelOSStable] = nsi.OperatingSystem
+	}
+
+	// The order of priority is:
+	// - Labels set in existing nodes for not-autoscale-from-zero cases
+	// - Labels set in the labels capacity annotation of machine template, machine set, and machine deployment.
+	// - Values in the status.nodeSystemInfo of MachineTemplates
+	// - Generic/default labels set in the environment of the cluster autoscaler
+	labels := cloudprovider.JoinStringMaps(buildGenericLabels(nodeName), nsiLabels, ng.scalableResource.Labels())
 
 	nodes, err := ng.Nodes()
 	if err != nil {

cluster-autoscaler/cloudprovider/clusterapi/clusterapi_unstructured.go

@@ -23,6 +23,7 @@ import (
 	"path"
 	"strconv"
 	"strings"
+	"sync"
 	"time"
 
 	"github.com/pkg/errors"
@@ -42,6 +43,8 @@ import (
 type unstructuredScalableResource struct {
 	controller         *machineController
 	unstructured       *unstructured.Unstructured
+	infraObj           *unstructured.Unstructured
+	infraMutex         sync.RWMutex
 	maxSize            int
 	minSize            int
 	autoscalingOptions map[string]string
@@ -321,6 +324,17 @@ func (r unstructuredScalableResource) InstanceCapacity() (map[corev1.ResourceNam
 	return capacity, nil
 }
 
+// InstanceSystemInfo sets the nodeSystemInfo from the infrastructure reference resource.
+// If the infrastructure reference resource is not found, returns nil.
+func (r unstructuredScalableResource) InstanceSystemInfo() *apiv1.NodeSystemInfo {
+	infraObj, err := r.readInfrastructureReferenceResource()
+	if err != nil || infraObj == nil {
+		return nil
+	}
+	nsiObj := systemInfoFromInfrastructureObject(infraObj)
+	return &nsiObj
+}
+
 func (r unstructuredScalableResource) InstanceResourceSlices(nodeName string) ([]*resourceapi.ResourceSlice, error) {
 	var result []*resourceapi.ResourceSlice
 	driver := r.InstanceDRADriver()
@@ -392,6 +406,17 @@ func (r unstructuredScalableResource) InstanceDRADriver() string {
 }
 
 func (r unstructuredScalableResource) readInfrastructureReferenceResource() (*unstructured.Unstructured, error) {
+	// Cache w/ lazy loading of the infrastructure reference resource.
+	r.infraMutex.RLock()
+	if r.infraObj != nil {
+		defer r.infraMutex.RUnlock()
+		return r.infraObj, nil
+	}
+	r.infraMutex.RUnlock()
+
+	r.infraMutex.Lock()
+	defer r.infraMutex.Unlock()
+
 	obKind := r.unstructured.GetKind()
 	obName := r.unstructured.GetName()
 
@@ -442,6 +467,8 @@ func (r unstructuredScalableResource) readInfrastructureReferenceResource() (*un
 		return nil, err
 	}
 
+	r.infraObj = infra
+
 	return infra, nil
 }
 
@@ -479,6 +506,25 @@ func resourceCapacityFromInfrastructureObject(infraobj *unstructured.Unstructure
 	return capacity
 }
 
+func systemInfoFromInfrastructureObject(infraobj *unstructured.Unstructured) apiv1.NodeSystemInfo {
+	nsi := apiv1.NodeSystemInfo{}
+	infransi, found, err := unstructured.NestedStringMap(infraobj.Object, "status", "nodeInfo")
+	if !found || err != nil {
+		return nsi
+	}
+
+	for k, v := range infransi {
+		switch k {
+		case "architecture":
+			nsi.Architecture = v
+		case "operatingSystem":
+			nsi.OperatingSystem = v
+		}
+	}
+
+	return nsi
+}
+
 // adapted from https://github.com/kubernetes/kubernetes/blob/release-1.25/pkg/util/taints/taints.go#L39
 func parseTaint(st string) (apiv1.Taint, error) {
 	var taint apiv1.Taint