Merge pull request #8474 from MaximilianoUribe/muribefalcon/cherry-pick-force-delete-failed-instance

chore: cherry-pick forceDeleteFailedNodes into release branch 1.33

Author: k8s-ci-robot

cluster-autoscaler/config/autoscaling_options.go

@@ -305,6 +305,8 @@ type AutoscalingOptions struct {
 	CheckCapacityProvisioningRequestBatchTimebox time.Duration
 	// ForceDeleteLongUnregisteredNodes is used to enable/disable ignoring min size constraints during removal of long unregistered nodes
 	ForceDeleteLongUnregisteredNodes bool
+	// ForceDeleteFailedNodes is used to enable/disable ignoring min size constraints during removal of failed nodes
+	ForceDeleteFailedNodes bool
 	// DynamicResourceAllocationEnabled configures whether logic for handling DRA objects is enabled.
 	DynamicResourceAllocationEnabled bool
 	// ClusterSnapshotParallelism is the maximum parallelism of cluster snapshot creation.

cluster-autoscaler/config/flags/flags.go

@@ -223,6 +223,7 @@ var (
 	checkCapacityProvisioningRequestMaxBatchSize = flag.Int("check-capacity-provisioning-request-max-batch-size", 10, "Maximum number of provisioning requests to process in a single batch.")
 	checkCapacityProvisioningRequestBatchTimebox = flag.Duration("check-capacity-provisioning-request-batch-timebox", 10*time.Second, "Maximum time to process a batch of provisioning requests.")
 	forceDeleteLongUnregisteredNodes             = flag.Bool("force-delete-unregistered-nodes", false, "Whether to enable force deletion of long unregistered nodes, regardless of the min size of the node group the belong to.")
+	forceDeleteFailedNodes                       = flag.Bool("force-delete-failed-nodes", false, "Whether to enable force deletion of failed nodes, regardless of the min size of the node group the belong to.")
 	enableDynamicResourceAllocation              = flag.Bool("enable-dynamic-resource-allocation", false, "Whether logic for handling DRA (Dynamic Resource Allocation) objects is enabled.")
 	clusterSnapshotParallelism                   = flag.Int("cluster-snapshot-parallelism", 16, "Maximum parallelism of cluster snapshot creation.")
 	checkCapacityProcessorInstance               = flag.String("check-capacity-processor-instance", "", "Name of the processor instance. Only ProvisioningRequests that define this name in their parameters with the key \"processorInstance\" will be processed by this CA instance. It only refers to check capacity ProvisioningRequests, but if not empty, best-effort atomic ProvisioningRequests processing is disabled in this instance. Not recommended: Until CA 1.35, ProvisioningRequests with this name as prefix in their class will be also processed.")
@@ -391,6 +392,7 @@ func createAutoscalingOptions() config.AutoscalingOptions {
 		CheckCapacityProvisioningRequestMaxBatchSize: *checkCapacityProvisioningRequestMaxBatchSize,
 		CheckCapacityProvisioningRequestBatchTimebox: *checkCapacityProvisioningRequestBatchTimebox,
 		ForceDeleteLongUnregisteredNodes:             *forceDeleteLongUnregisteredNodes,
+		ForceDeleteFailedNodes:                       *forceDeleteFailedNodes,
 		DynamicResourceAllocationEnabled:             *enableDynamicResourceAllocation,
 		ClusterSnapshotParallelism:                   *clusterSnapshotParallelism,
 		CheckCapacityProcessorInstance:               *checkCapacityProcessorInstance,

cluster-autoscaler/core/static_autoscaler.go

@@ -879,8 +879,15 @@ func (a *StaticAutoscaler) deleteCreatedNodesWithErrors() {
 		nodeGroup := nodeGroups[nodeGroupId]
 		if nodeGroup == nil {
 			err = fmt.Errorf("node group %s not found", nodeGroupId)
-		} else if nodesToDelete, err = overrideNodesToDeleteForZeroOrMax(a.NodeGroupDefaults, nodeGroup, nodesToDelete); err == nil {
-			err = nodeGroup.DeleteNodes(nodesToDelete)
+		} else if nodesToDelete, err = overrideNodesToDeleteForZeroOrMax(a.NodeGroupDefaults, nodeGroup, nodesToDelete); err == nil && len(nodesToDelete) > 0 {
+			if a.ForceDeleteFailedNodes {
+				err = nodeGroup.ForceDeleteNodes(nodesToDelete)
+				if errors.Is(err, cloudprovider.ErrNotImplemented) {
+					err = nodeGroup.DeleteNodes(nodesToDelete)
+				}
+			} else {
+				err = nodeGroup.DeleteNodes(nodesToDelete)
+			}
 		}
 
 		if err != nil {