fix: Ensure volumes are unpublished before deleting node

kwok/charts/templates/clusterrole.yaml

@@ -32,6 +32,9 @@ rules:
   - apiGroups: ["karpenter.sh"]
     resources: ["nodepools", "nodepools/status", "nodeclaims", "nodeclaims/status"]
     verbs: ["get", "list", "watch"]
+  - apiGroups: [ "storage.k8s.io" ]
+    resources: [ "volumeattachments" ]
+    verbs: [ "get", "list", "watch" ]
   - apiGroups: [""]
     resources: ["pods", "nodes", "persistentvolumes", "persistentvolumeclaims", "replicationcontrollers", "namespaces"]
     verbs: ["get", "list", "watch"]
@@ -62,4 +65,4 @@ rules:
     verbs: ["create"]
   {{- with .Values.additionalClusterRoleRules -}}
   {{ toYaml . | nindent 2 }}
-  {{- end -}}
+  {{- end -}}

pkg/controllers/node/termination/controller.go

@@ -47,6 +47,7 @@ import (
 	"sigs.k8s.io/karpenter/pkg/events"
 	"sigs.k8s.io/karpenter/pkg/metrics"
 	nodeutils "sigs.k8s.io/karpenter/pkg/utils/node"
+	volumeattachmentutils "sigs.k8s.io/karpenter/pkg/utils/volumeattachment"
 )
 
 // Controller for the resource
@@ -107,6 +108,15 @@ func (c *Controller) finalize(ctx context.Context, node *v1.Node) (reconcile.Res
 		}
 		return reconcile.Result{RequeueAfter: 1 * time.Second}, nil
 	}
+	// In order for stateful pods to smoothly migrate from the terminating Node, we wait for VolumeAttachments
+	// of drain-able pods to be cleaned up before terminating the node and removing it from the cluster.
+	isNoVolumeAttachments, err := c.ensureNoVolumeAttachments(ctx, node)
+	if err != nil {
+		return reconcile.Result{}, fmt.Errorf("ensuring no volume attachments, %w", err)
+	}
+	if !isNoVolumeAttachments {
+		return reconcile.Result{RequeueAfter: 1 * time.Second}, nil
+	}
 	nodeClaims, err := nodeutils.GetNodeClaims(ctx, node, c.kubeClient)
 	if err != nil {
 		return reconcile.Result{}, fmt.Errorf("deleting nodeclaims, %w", err)
@@ -150,6 +160,22 @@ func (c *Controller) deleteAllNodeClaims(ctx context.Context, node *v1.Node) err
 	return nil
 }
 
+func (c *Controller) ensureNoVolumeAttachments(ctx context.Context, node *v1.Node) (bool, error) {
+	volumeAttachments, err := nodeutils.GetVolumeAttachments(ctx, c.kubeClient, node)
+	if err != nil {
+		return false, err
+	}
+	// Filter out volume attachments associated with non-drain-able nodes or multi-attachable volumes
+	filteredVolumeAttachments, err := volumeattachmentutils.FilterVolumeAttachments(ctx, c.kubeClient, node, volumeAttachments)
+	if err != nil {
+		return false, err
+	}
+	if len(filteredVolumeAttachments) > 0 {
+		return false, nil
+	}
+	return true, nil
+}
+
 func (c *Controller) removeFinalizer(ctx context.Context, n *v1.Node) error {
 	stored := n.DeepCopy()
 	controllerutil.RemoveFinalizer(n, v1beta1.TerminationFinalizer)

pkg/operator/operator.go

@@ -26,6 +26,8 @@ import (
 	"sync"
 	"time"
 
+	storagev1 "k8s.io/api/storage/v1"
+
 	"github.com/awslabs/operatorpkg/controller"
 	opmetrics "github.com/awslabs/operatorpkg/metrics"
 	"github.com/prometheus/client_golang/prometheus"
@@ -204,6 +206,9 @@ func NewOperator() (context.Context, *Operator) {
 	lo.Must0(mgr.GetFieldIndexer().IndexField(ctx, &v1beta1.NodeClaim{}, "spec.nodeClassRef.name", func(o client.Object) []string {
 		return []string{o.(*v1beta1.NodeClaim).Spec.NodeClassRef.Name}
 	}), "failed to setup nodeclaim nodeclassref name indexer")
+	lo.Must0(mgr.GetFieldIndexer().IndexField(ctx, &storagev1.VolumeAttachment{}, "spec.nodeName", func(o client.Object) []string {
+		return []string{o.(*storagev1.VolumeAttachment).Spec.NodeName}
+	}), "failed to setup volumeattachment indexer")
 
 	lo.Must0(mgr.AddReadyzCheck("manager", func(req *http.Request) error {
 		return lo.Ternary(mgr.GetCache().WaitForCacheSync(req.Context()), nil, fmt.Errorf("failed to sync caches"))

pkg/utils/node/node.go

@@ -20,12 +20,13 @@ import (
 	"context"
 	"fmt"
 
-	"sigs.k8s.io/karpenter/pkg/apis/v1beta1"
+	storagev1 "k8s.io/api/storage/v1"
 
 	"github.com/samber/lo"
 	v1 "k8s.io/api/core/v1"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
+	"sigs.k8s.io/karpenter/pkg/apis/v1beta1"
 	"sigs.k8s.io/karpenter/pkg/utils/pod"
 )
 
@@ -75,6 +76,32 @@ func GetProvisionablePods(ctx context.Context, kubeClient client.Client) ([]*v1.
 	}), nil
 }
 
+// GetTolerantPods grabs all pods from passed nodes that don't satisfy the ToleratesDisruptionNoScheduleTaint criteria
+func GetTolerantPods(ctx context.Context, kubeClient client.Client, nodes ...*v1.Node) ([]*v1.Pod, error) {
+	pods, err := GetPods(ctx, kubeClient, nodes...)
+	if err != nil {
+		return nil, fmt.Errorf("listing pods, %w", err)
+	}
+	return lo.Reject(pods, func(p *v1.Pod, _ int) bool {
+		return pod.ToleratesDisruptionNoScheduleTaint(p)
+	}), nil
+}
+
+// GetVolumeAttachments grabs all volumeAttachments of passed node
+func GetVolumeAttachments(ctx context.Context, kubeClient client.Client, node *v1.Node) ([]*storagev1.VolumeAttachment, error) {
+	var volumeAttachments []*storagev1.VolumeAttachment
+	var volumeAttachmentList storagev1.VolumeAttachmentList
+	if err := kubeClient.List(ctx, &volumeAttachmentList, client.MatchingFields{"spec.nodeName": node.Name}); err != nil {
+		return nil, fmt.Errorf("listing volumeattachments, %w", err)
+	}
+	for i := range volumeAttachmentList.Items {
+		if volumeAttachmentList.Items[i].Spec.NodeName == node.Name {
+			volumeAttachments = append(volumeAttachments, &volumeAttachmentList.Items[i])
+		}
+	}
+	return volumeAttachments, nil
+}
+
 func GetCondition(n *v1.Node, match v1.NodeConditionType) v1.NodeCondition {
 	for _, condition := range n.Status.Conditions {
 		if condition.Type == match {

pkg/utils/volumeattachment/volumeattachment.go

@@ -0,0 +1,74 @@
+/*
+Copyright The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package volumeattachment
+
+import (
+	"context"
+	"fmt"
+
+	v1 "k8s.io/api/core/v1"
+	storagev1 "k8s.io/api/storage/v1"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+
+	nodeutil "sigs.k8s.io/karpenter/pkg/utils/node"
+	volumeutil "sigs.k8s.io/karpenter/pkg/utils/volume"
+)
+
+// FilterVolumeAttachments filters out volumeAttachments that should not block the termination of the passed node
+func FilterVolumeAttachments(ctx context.Context, kubeClient client.Client, node *v1.Node, volumeAttachments []*storagev1.VolumeAttachment) ([]*storagev1.VolumeAttachment, error) {
+	var filteredVolumeAttachments []*storagev1.VolumeAttachment
+	// No need to filter empty volumeAttachments list
+	if len(volumeAttachments) == 0 {
+		return volumeAttachments, nil
+	}
+	// Filter out non-drain-able pods
+	tolerantPods, err := nodeutil.GetTolerantPods(ctx, kubeClient, node)
+	if err != nil {
+		return nil, fmt.Errorf("listing pods on node, %w", err)
+	}
+	// Filter out Multi-Attach volumes
+	shouldFilterOutVolume := make(map[string]bool)
+	for _, p := range tolerantPods {
+		for _, v := range p.Spec.Volumes {
+			pvc, err := volumeutil.GetPersistentVolumeClaim(ctx, kubeClient, p, v)
+			if err != nil {
+				return nil, fmt.Errorf("getting persistent volume claim, %w", err)
+			}
+			if pvc != nil {
+				shouldFilterOutVolume[pvc.Spec.VolumeName] = CannotMultiAttach(*pvc)
+			}
+		}
+	}
+	for i := range volumeAttachments {
+		pvName := volumeAttachments[i].Spec.Source.PersistentVolumeName
+		if pvName != nil && shouldFilterOutVolume[*pvName] {
+			filteredVolumeAttachments = append(filteredVolumeAttachments, volumeAttachments[i])
+		}
+	}
+	return filteredVolumeAttachments, nil
+}
+
+// CannotMultiAttach returns true if the persistentVolumeClaim's underlying volume cannot be attached to multiple nodes
+// i.e. its access mode is not ReadWriteOnce/ReadWriteOncePod
+func CannotMultiAttach(pvc v1.PersistentVolumeClaim) bool {
+	for _, accessMode := range pvc.Spec.AccessModes {
+		if accessMode == v1.ReadWriteOnce || accessMode == v1.ReadWriteOncePod {
+			return true
+		}
+	}
+	return false
+}