Add code for reading unstructured clusterapi csidriver annotations

Author: gnufied

cluster-autoscaler/cloudprovider/clusterapi/README.md

@@ -307,6 +307,30 @@ metadata:
 > Please see the [Cluster API Book chapter on Metadata propagation](https://cluster-api.sigs.k8s.io/reference/api/metadata-propagation)
 > for more information.
 
+
+#### Pre-defined csi driver information on nodes scaled from zero
+
+To provide CSI driver information for scale from zero, the optional
+capacity annotation may be supplied as a comma separated list of driver name
+and volume limit key/value pairs, as demonstrated in the example below:
+
+```yaml
+apiVersion: cluster.x-k8s.io/v1alpha4
+kind: MachineDeployment
+metadata:
+  annotations:
+    cluster.x-k8s.io/cluster-api-autoscaler-node-group-max-size: "5"
+    cluster.x-k8s.io/cluster-api-autoscaler-node-group-min-size: "0"
+    capacity.cluster-autoscaler.kubernetes.io/memory: "128G"
+    capacity.cluster-autoscaler.kubernetes.io/cpu: "16"
+    capacity.cluster-autoscaler.kubernetes.io/csi-driver: "ebs.csi.aws.com=25,efs.csi.aws.com=16"
+```
+
+> Note: The CSI driver information supplied through the capacity annotation
+> specifies which CSI drivers will be installed on nodes scaled from zero, along
+> with their respective volume limits. The format is `driver-name=volume-limit`
+> with multiple drivers separated by commas.
+
 #### Per-NodeGroup autoscaling options
 
 Custom autoscaling options per node group (MachineDeployment/MachinePool/MachineSet) can be specified as annoations with a common prefix:
@@ -328,14 +352,14 @@ metadata:
     cluster.x-k8s.io/autoscaling-options-maxnodeprovisiontime: "20m0s"
 ```
 
-#### CPU Architecture awareness for single-arch clusters 
+#### CPU Architecture awareness for single-arch clusters
 
-Users of single-arch non-amd64 clusters who are using scale from zero 
+Users of single-arch non-amd64 clusters who are using scale from zero
 support should also set the `CAPI_SCALE_ZERO_DEFAULT_ARCH` environment variable
 to set the architecture of the nodes they want to default the node group templates to.
-The autoscaler will default to `amd64` if it is not set, and the node 
-group templates may not match the nodes' architecture, specifically when 
-the workload triggering the scale-up uses a node affinity predicate checking 
+The autoscaler will default to `amd64` if it is not set, and the node
+group templates may not match the nodes' architecture, specifically when
+the workload triggering the scale-up uses a node affinity predicate checking
 for the node's architecture.
 
 ## Specifying a Custom Resource Group

cluster-autoscaler/cloudprovider/clusterapi/clusterapi_nodegroup.go

@@ -380,8 +380,12 @@ func (ng *nodegroup) TemplateNodeInfo() (*framework.NodeInfo, error) {
 	if err != nil {
 		return nil, err
 	}
+	csiNode := ng.scalableResource.InstanceCSINode()
 
 	nodeInfo := framework.NewNodeInfo(&node, resourceSlices, &framework.PodInfo{Pod: cloudprovider.BuildKubeProxy(ng.scalableResource.Name())})
+	if csiNode != nil {
+		nodeInfo.AddCSINode(csiNode)
+	}
 	return nodeInfo, nil
 }
 

cluster-autoscaler/cloudprovider/clusterapi/clusterapi_nodegroup_test.go

@@ -27,13 +27,15 @@ import (
 
 	"github.com/stretchr/testify/assert"
 	corev1 "k8s.io/api/core/v1"
+	storagev1 "k8s.io/api/storage/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
 	"k8s.io/apimachinery/pkg/util/wait"
 	"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
 	"k8s.io/autoscaler/cluster-autoscaler/config"
 	gpuapis "k8s.io/autoscaler/cluster-autoscaler/utils/gpu"
 	"k8s.io/client-go/tools/cache"
+	"k8s.io/utils/ptr"
 )
 
 const (
@@ -1500,6 +1502,7 @@ func TestNodeGroupTemplateNodeInfo(t *testing.T) {
 		expectedCapacity      map[corev1.ResourceName]int64
 		expectedNodeLabels    map[string]string
 		expectedResourceSlice testResourceSlice
+		expectedCSINode       *storagev1.CSINode
 	}
 
 	testCases := []struct {
@@ -1650,6 +1653,49 @@ func TestNodeGroupTemplateNodeInfo(t *testing.T) {
 				},
 			},
 		},
+		{
+			name: "When the NodeGroup can scale from zero and CSI driver annotations are present, it creates CSINode with driver information",
+			nodeGroupAnnotations: map[string]string{
+				memoryKey:    "2048Mi",
+				cpuKey:       "2",
+				csiDriverKey: "ebs.csi.aws.com=25,efs.csi.aws.com=16",
+			},
+			config: testCaseConfig{
+				expectedErr: nil,
+				nodeLabels: map[string]string{
+					"kubernetes.io/os":   "linux",
+					"kubernetes.io/arch": "amd64",
+				},
+				expectedCapacity: map[corev1.ResourceName]int64{
+					corev1.ResourceCPU:    2,
+					corev1.ResourceMemory: 2048 * 1024 * 1024,
+					corev1.ResourcePods:   110,
+				},
+				expectedNodeLabels: map[string]string{
+					"kubernetes.io/os":       "linux",
+					"kubernetes.io/arch":     "amd64",
+					"kubernetes.io/hostname": "random value",
+				},
+				expectedCSINode: &storagev1.CSINode{
+					Spec: storagev1.CSINodeSpec{
+						Drivers: []storagev1.CSINodeDriver{
+							{
+								Name: "ebs.csi.aws.com",
+								Allocatable: &storagev1.VolumeNodeResources{
+									Count: ptr.To(int32(25)),
+								},
+							},
+							{
+								Name: "efs.csi.aws.com",
+								Allocatable: &storagev1.VolumeNodeResources{
+									Count: ptr.To(int32(16)),
+								},
+							},
+						},
+					},
+				},
+			},
+		},
 	}
 
 	test := func(t *testing.T, testConfig *TestConfig, config testCaseConfig) {
@@ -1726,6 +1772,55 @@ func TestNodeGroupTemplateNodeInfo(t *testing.T) {
 				}
 			}
 		}
+
+		// Validate CSINode if expected
+		if config.expectedCSINode != nil {
+			if nodeInfo.CSINode == nil {
+				t.Errorf("Expected CSINode to be set, but got nil")
+			} else {
+				expectedDrivers := config.expectedCSINode.Spec.Drivers
+				gotDrivers := nodeInfo.CSINode.Spec.Drivers
+				if len(expectedDrivers) != len(gotDrivers) {
+					t.Errorf("Expected %d CSI drivers, but got %d", len(expectedDrivers), len(gotDrivers))
+				} else {
+					for i, expectedDriver := range expectedDrivers {
+						if i >= len(gotDrivers) {
+							t.Errorf("Expected driver at index %d but got only %d drivers", i, len(gotDrivers))
+							break
+						}
+						gotDriver := gotDrivers[i]
+						if expectedDriver.Name != gotDriver.Name {
+							t.Errorf("Expected CSI driver name at index %d to be %s, but got %s", i, expectedDriver.Name, gotDriver.Name)
+						}
+						if expectedDriver.Allocatable == nil {
+							if gotDriver.Allocatable != nil {
+								t.Errorf("Expected CSI driver Allocatable at index %d to be nil, but got non-nil", i)
+							}
+						} else {
+							if gotDriver.Allocatable == nil {
+								t.Errorf("Expected CSI driver Allocatable at index %d to be non-nil, but got nil", i)
+							} else {
+								if expectedDriver.Allocatable.Count == nil {
+									if gotDriver.Allocatable.Count != nil {
+										t.Errorf("Expected CSI driver Count at index %d to be nil, but got %d", i, *gotDriver.Allocatable.Count)
+									}
+								} else {
+									if gotDriver.Allocatable.Count == nil {
+										t.Errorf("Expected CSI driver Count at index %d to be %d, but got nil", i, *expectedDriver.Allocatable.Count)
+									} else if *expectedDriver.Allocatable.Count != *gotDriver.Allocatable.Count {
+										t.Errorf("Expected CSI driver Count at index %d to be %d, but got %d", i, *expectedDriver.Allocatable.Count, *gotDriver.Allocatable.Count)
+									}
+								}
+							}
+						}
+					}
+				}
+			}
+		} else {
+			if nodeInfo.CSINode != nil {
+				t.Errorf("Expected CSINode to be nil, but got non-nil with %d drivers", len(nodeInfo.CSINode.Spec.Drivers))
+			}
+		}
 	}
 
 	for _, tc := range testCases {

cluster-autoscaler/cloudprovider/clusterapi/clusterapi_unstructured.go

@@ -30,6 +30,7 @@ import (
 	apiv1 "k8s.io/api/core/v1"
 	corev1 "k8s.io/api/core/v1"
 	resourceapi "k8s.io/api/resource/v1"
+	storagev1 "k8s.io/api/storage/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
@@ -413,6 +414,27 @@ func (r unstructuredScalableResource) InstanceDRADriver() string {
 	return parseDRADriver(r.unstructured.GetAnnotations())
 }
 
+// InstanceCSINode parses CSI driver information from annotations and returns
+// a CSINode object with the list of installed drivers and their volume limits.
+// The annotation format is "driver-name=volume-limit,driver-name2=volume-limit2".
+// Returns nil if the annotation is not present or empty.
+func (r unstructuredScalableResource) InstanceCSINode() *storagev1.CSINode {
+	annotations := r.unstructured.GetAnnotations()
+	// annotation value of the form "driver1=limit1,driver2=limit2"
+	if val, found := annotations[csiDriverKey]; found && val != "" {
+		drivers := parseCSIDriverAnnotation(val)
+		if len(drivers) == 0 {
+			return nil
+		}
+		return &storagev1.CSINode{
+			Spec: storagev1.CSINodeSpec{
+				Drivers: drivers,
+			},
+		}
+	}
+	return nil
+}
+
 func (r unstructuredScalableResource) readInfrastructureReferenceResource() (*unstructured.Unstructured, error) {
 	// Cache w/ lazy loading of the infrastructure reference resource.
 	r.infraMutex.RLock()
@@ -533,6 +555,66 @@ func systemInfoFromInfrastructureObject(infraobj *unstructured.Unstructured) api
 	return nsi
 }
 
+// parseCSIDriverAnnotation parses a comma-separated list of CSI driver name and volume limit
+// key/value pairs in the format "driver-name=volume-limit,driver-name2=volume-limit2".
+// Returns a slice of CSINodeDriver objects with Name and Allocatable.Count set.
+func parseCSIDriverAnnotation(annotationValue string) []storagev1.CSINodeDriver {
+	drivers := []storagev1.CSINodeDriver{}
+	if annotationValue == "" {
+		return drivers
+	}
+
+	driverSpecs := strings.Split(annotationValue, ",")
+	for _, driverSpec := range driverSpecs {
+		driverSpec = strings.TrimSpace(driverSpec)
+		if driverSpec == "" {
+			continue
+		}
+
+		// Split on "=" to get driver name and volume limit
+		parts := strings.SplitN(driverSpec, "=", 2)
+		if len(parts) != 2 {
+			klog.V(4).Infof("Invalid CSI driver spec format (expected driver-name=volume-limit): %s", driverSpec)
+			continue
+		}
+
+		driverName := strings.TrimSpace(parts[0])
+		volumeLimitStr := strings.TrimSpace(parts[1])
+
+		if driverName == "" {
+			klog.V(4).Infof("Empty driver name in CSI driver spec: %s", driverSpec)
+			continue
+		}
+
+		// Parse volume limit as integer
+		volumeLimit, err := strconv.ParseInt(volumeLimitStr, 10, 32)
+		if err != nil {
+			klog.V(4).Infof("Invalid volume limit value (expected integer) in CSI driver spec %s: %v", driverSpec, err)
+			continue
+		}
+
+		if volumeLimit < 0 {
+			klog.V(4).Infof("Volume limit must be non-negative in CSI driver spec: %s", driverSpec)
+			continue
+		}
+
+		// Create CSINodeDriver with Name and optionally Allocatable.Count
+		// If volume limit is 0, Allocatable is not set
+		driver := storagev1.CSINodeDriver{
+			Name: driverName,
+		}
+		if volumeLimit > 0 {
+			limit := int32(volumeLimit)
+			driver.Allocatable = &storagev1.VolumeNodeResources{
+				Count: &limit,
+			}
+		}
+		drivers = append(drivers, driver)
+	}
+
+	return drivers
+}
+
 // 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