add cp node deletion param configurable

cluster-autoscaler/config/autoscaling_options.go

@@ -289,6 +289,8 @@ type AutoscalingOptions struct {
 	BypassedSchedulers map[string]bool
 	// ProvisioningRequestEnabled tells if CA processes ProvisioningRequest.
 	ProvisioningRequestEnabled bool
+	// MaxCloudProviderNodeDeletionTime is the maximum time needed by cloud provider to delete a node
+	MaxCloudProviderNodeDeletionTime time.Duration
 }
 
 // KubeClientOptions specify options for kube client

cluster-autoscaler/core/scaledown/actuation/delete_in_batch.go

@@ -41,8 +41,6 @@ import (
 const (
 	// MaxKubernetesEmptyNodeDeletionTime is the maximum time needed by Kubernetes to delete an empty node.
 	MaxKubernetesEmptyNodeDeletionTime = 3 * time.Minute
-	// MaxCloudProviderNodeDeletionTime is the maximum time needed by cloud provider to delete a node.
-	MaxCloudProviderNodeDeletionTime = 5 * time.Minute
 )
 
 // NodeDeletionBatcher batch scale down candidates for one node group and remove them.
@@ -180,9 +178,9 @@ func nodeScaleDownReason(node *apiv1.Node, drain bool) metrics.NodeScaleDownReas
 }
 
 // IsNodeBeingDeleted returns true iff a given node is being deleted.
-func IsNodeBeingDeleted(node *apiv1.Node, timestamp time.Time) bool {
+func IsNodeBeingDeleted(ctx *context.AutoscalingContext, node *apiv1.Node, timestamp time.Time) bool {
 	deleteTime, _ := taints.GetToBeDeletedTime(node)
-	return deleteTime != nil && (timestamp.Sub(*deleteTime) < MaxCloudProviderNodeDeletionTime || timestamp.Sub(*deleteTime) < MaxKubernetesEmptyNodeDeletionTime)
+	return deleteTime != nil && (timestamp.Sub(*deleteTime) < ctx.MaxCloudProviderNodeDeletionTime || timestamp.Sub(*deleteTime) < MaxKubernetesEmptyNodeDeletionTime)
 }
 
 // CleanUpAndRecordFailedScaleDownEvent record failed scale down event and log an error.
@@ -203,7 +201,7 @@ func CleanUpAndRecordFailedScaleDownEvent(ctx *context.AutoscalingContext, node
 func RegisterAndRecordSuccessfulScaleDownEvent(ctx *context.AutoscalingContext, scaleStateNotifier nodegroupchange.NodeGroupChangeObserver, node *apiv1.Node, nodeGroup cloudprovider.NodeGroup, drain bool, nodeDeletionTracker *deletiontracker.NodeDeletionTracker) {
 	ctx.Recorder.Eventf(node, apiv1.EventTypeNormal, "ScaleDown", "nodes removed by cluster autoscaler")
 	currentTime := time.Now()
-	expectedDeleteTime := time.Now().Add(MaxCloudProviderNodeDeletionTime)
+	expectedDeleteTime := time.Now().Add(ctx.MaxCloudProviderNodeDeletionTime)
 	scaleStateNotifier.RegisterScaleDown(nodeGroup, node.Name, currentTime, expectedDeleteTime)
 	gpuConfig := ctx.CloudProvider.GetNodeGpuConfig(node)
 	metricResourceName, metricGpuType := gpu.GetGpuInfoForMetrics(gpuConfig, ctx.CloudProvider.GetAvailableGPUTypes(), node, nodeGroup)

cluster-autoscaler/core/scaledown/eligibility/eligibility.go

@@ -109,7 +109,7 @@ func (c *Checker) FilterOutUnremovable(context *context.AutoscalingContext, scal
 func (c *Checker) unremovableReasonAndNodeUtilization(context *context.AutoscalingContext, timestamp time.Time, nodeInfo *schedulerframework.NodeInfo, utilLogsQuota *klogx.Quota) (simulator.UnremovableReason, *utilization.Info) {
 	node := nodeInfo.Node()
 
-	if actuation.IsNodeBeingDeleted(node, timestamp) {
+	if actuation.IsNodeBeingDeleted(context, node, timestamp) {
 		klog.V(1).Infof("Skipping %s from delete consideration - the node is currently being deleted", node.Name)
 		return simulator.CurrentlyBeingDeleted, nil
 	}

cluster-autoscaler/core/scaledown/eligibility/eligibility_test.go

@@ -164,6 +164,7 @@ func TestFilterOutUnremovable(t *testing.T) {
 					ScaleDownUnreadyTime:             config.DefaultScaleDownUnreadyTime,
 					IgnoreDaemonSetsUtilization:      tc.ignoreDaemonSetsUtilization,
 				},
+				MaxCloudProviderNodeDeletionTime: 5 * time.Minute,
 			}
 			s := nodegroupconfig.NewDefaultNodeGroupConfigProcessor(options.NodeGroupDefaults)
 			c := NewChecker(s)

cluster-autoscaler/core/scaledown/resource/limits.go

@@ -62,7 +62,7 @@ func NoLimits() Limits {
 // LimitsLeft returns the amount of each resource that can be deleted from the
 // cluster without violating any constraints.
 func (lf *LimitsFinder) LimitsLeft(context *context.AutoscalingContext, nodes []*apiv1.Node, resourceLimiter *cloudprovider.ResourceLimiter, timestamp time.Time) Limits {
-	totalCores, totalMem := coresMemoryTotal(nodes, timestamp)
+	totalCores, totalMem := coresMemoryTotal(context, nodes, timestamp)
 
 	var totalResources map[string]int64
 	var totalResourcesErr error
@@ -102,10 +102,10 @@ func computeAboveMin(total int64, min int64) int64 {
 	return 0
 }
 
-func coresMemoryTotal(nodes []*apiv1.Node, timestamp time.Time) (int64, int64) {
+func coresMemoryTotal(ctx *context.AutoscalingContext, nodes []*apiv1.Node, timestamp time.Time) (int64, int64) {
 	var coresTotal, memoryTotal int64
 	for _, node := range nodes {
-		if actuation.IsNodeBeingDeleted(node, timestamp) {
+		if actuation.IsNodeBeingDeleted(ctx, node, timestamp) {
 			// Nodes being deleted do not count towards total cluster resources
 			continue
 		}
@@ -122,7 +122,7 @@ func (lf *LimitsFinder) customResourcesTotal(context *context.AutoscalingContext
 	result := make(map[string]int64)
 	ngCache := make(map[string][]customresources.CustomResourceTarget)
 	for _, node := range nodes {
-		if actuation.IsNodeBeingDeleted(node, timestamp) {
+		if actuation.IsNodeBeingDeleted(context, node, timestamp) {
 			// Nodes being deleted do not count towards total cluster resources
 			continue
 		}

cluster-autoscaler/core/scaledown/resource/limits_test.go

@@ -21,6 +21,7 @@ import (
 	"testing"
 	"time"
 
+	"k8s.io/autoscaler/cluster-autoscaler/config"
 	. "k8s.io/autoscaler/cluster-autoscaler/core/test"
 	"k8s.io/autoscaler/cluster-autoscaler/core/utils"
 	"k8s.io/autoscaler/cluster-autoscaler/utils/taints"
@@ -55,7 +56,12 @@ func TestCalculateCoresAndMemoryTotal(t *testing.T) {
 		},
 	}
 
-	coresTotal, memoryTotal := coresMemoryTotal(nodes, time.Now())
+	options := config.AutoscalingOptions{
+		MaxCloudProviderNodeDeletionTime: 5 * time.Minute,
+	}
+	context, err := NewScaleTestAutoscalingContext(options, nil, nil, nil, nil, nil)
+	assert.NoError(t, err)
+	coresTotal, memoryTotal := coresMemoryTotal(&context, nodes, time.Now())
 
 	assert.Equal(t, int64(42), coresTotal)
 	assert.Equal(t, int64(44000*utils.MiB), memoryTotal)

cluster-autoscaler/core/static_autoscaler.go

@@ -322,7 +322,7 @@ func (a *StaticAutoscaler) RunOnce(currentTime time.Time) caerrors.AutoscalerErr
 	}
 
 	// Update cluster resource usage metrics
-	coresTotal, memoryTotal := calculateCoresMemoryTotal(allNodes, currentTime)
+	coresTotal, memoryTotal := calculateCoresMemoryTotal(a.AutoscalingContext, allNodes, currentTime)
 	metrics.UpdateClusterCPUCurrentCores(coresTotal)
 	metrics.UpdateClusterMemoryCurrentBytes(memoryTotal)
 
@@ -1048,12 +1048,12 @@ func getUpcomingNodeInfos(upcomingCounts map[string]int, nodeInfos map[string]*s
 	return upcomingNodes
 }
 
-func calculateCoresMemoryTotal(nodes []*apiv1.Node, timestamp time.Time) (int64, int64) {
+func calculateCoresMemoryTotal(context *context.AutoscalingContext, nodes []*apiv1.Node, timestamp time.Time) (int64, int64) {
 	// this function is essentially similar to the calculateScaleDownCoresMemoryTotal
 	// we want to check all nodes, aside from those deleting, to sum the cluster resource usage.
 	var coresTotal, memoryTotal int64
 	for _, node := range nodes {
-		if actuation.IsNodeBeingDeleted(node, timestamp) {
+		if actuation.IsNodeBeingDeleted(context, node, timestamp) {
 			// Nodes being deleted do not count towards total cluster resources
 			continue
 		}

cluster-autoscaler/main.go

@@ -257,7 +257,8 @@ var (
 			"--max-graceful-termination-sec flag should not be set when this flag is set. Not setting this flag will use unordered evictor by default."+
 			"Priority evictor reuses the concepts of drain logic in kubelet(https://github.com/kubernetes/enhancements/tree/master/keps/sig-node/2712-pod-priority-based-graceful-node-shutdown#migration-from-the-node-graceful-shutdown-feature)."+
 			"Eg. flag usage:  '10000:20,1000:100,0:60'")
-	provisioningRequestsEnabled = flag.Bool("enable-provisioning-requests", false, "Whether the clusterautoscaler will be handling the ProvisioningRequest CRs.")
+	provisioningRequestsEnabled      = flag.Bool("enable-provisioning-requests", false, "Whether the clusterautoscaler will be handling the ProvisioningRequest CRs.")
+	maxCloudProviderNodeDeletionTime = flag.Duration("max-cloud-provider-node-deletion-time", 5*time.Minute, "Maximum time needed by cloud provider to delete a node")
 )
 
 func isFlagPassed(name string) bool {