Identify the backed-up volume by CSI volume handle in the in-place restore pre-flight check (#10530)

The pre-flight check that verifies the existing PVC is still bound to the
backed-up volume compared PV names. A block data mover restore of a file
system volume recreates the PV under a new name (the volumeMode field is
immutable), so a second in-place restore of the same workload failed the
check even though the PVC was bound to the very same volume.

Record the CSI volume handle in the backup volume info (PVInfo) and
compare handles when both the backup and the bound PV record one; the
PV name remains the fallback for non-CSI volumes and for backups taken
before the handle was recorded. The handle reaches the PVC CSI RIA
through the same carrier annotation mechanism as the source size.

Signed-off-by: chlins <chlins.zhang@gmail.com>

pkg/podvolume/restorer.go

@@ -190,7 +190,11 @@ func (r *restorer) RestorePodVolumes(data RestoreData, tracker *volume.RestoreVo
 		// restore's PodVolumeRestores complete.
 		if data.Restore.IsVolumeDataInplaceRestore() && pvc != nil {
 			pvName := backedUpPVName(data.BackupVolumeInfos, data.SourceNamespace, pvc.Name)
-			if err := inplace.CheckPVCBoundToBackedUpPV(pvc, pvName, data.SourceNamespace); err != nil {
+			var volumeHandle string
+			if info := data.BackupVolumeInfos[pvName].PVInfo; info != nil {
+				volumeHandle = info.VolumeHandle
+			}
+			if err := inplace.CheckPVCBoundToBackedUpVolume(r.ctx, r.crClient, pvc, pvName, volumeHandle, data.SourceNamespace); err != nil {
 				errs = append(errs, err)
 				continue
 			}

pkg/podvolume/restorer_test.go

@@ -367,6 +367,7 @@ func TestRestorePodVolumes(t *testing.T) {
 			inplace: true,
 			kubeClientObj: []runtime.Object{
 				createNodeAgentDaemonset(),
+				createPVObj(1, false),
 				createPVCObj(1),
 				func() *corev1api.Pod {
 					pod := builder.ForPod("fake-ns", "other-pod").
@@ -398,6 +399,7 @@ func TestRestorePodVolumes(t *testing.T) {
 			inplace: true,
 			kubeClientObj: []runtime.Object{
 				createNodeAgentDaemonset(),
+				createPVObj(1, false),
 				createPVCObj(1),
 				createGatedPodObj("old-restore-uid", 1),
 			},
@@ -423,6 +425,7 @@ func TestRestorePodVolumes(t *testing.T) {
 			inplace: true,
 			kubeClientObj: []runtime.Object{
 				createNodeAgentDaemonset(),
+				createPVObj(1, false),
 				createPVCObj(1),
 			},
 			ctlClientObj: []runtime.Object{
@@ -440,6 +443,72 @@ func TestRestorePodVolumes(t *testing.T) {
 				},
 			},
 		},
+		{
+			// The PV was recreated under a new name by a previous in-place restore
+			// (block data mover on a file system volume) but is the same CSI volume.
+			name: "in-place restore proceeds when the PVC is bound to the backed-up volume under a recreated PV name",
+			pvbs: []*velerov1api.PodVolumeBackup{
+				createPVBObj(true, true, 1, "kopia"),
+			},
+			inplace: true,
+			kubeClientObj: []runtime.Object{
+				createNodeAgentDaemonset(),
+				createNodeObj(),
+				func() *corev1api.PersistentVolume {
+					pv := createPVObj(1, false)
+					pv.Spec.CSI = &corev1api.CSIPersistentVolumeSource{Driver: "fake.csi", VolumeHandle: "vol-1"}
+					return pv
+				}(),
+				createPVCObj(1),
+				createGatedPodObj("fake-restore-uid", 1),
+				createNodeAgentPodObj(true),
+			},
+			ctlClientObj: []runtime.Object{
+				createBackupRepoObj(),
+			},
+			restoredPod:     createPodObj(true, true, true, 1),
+			sourceNamespace: "fake-ns",
+			bsl:             "fake-bsl",
+			volumeInfos: map[string]volume.BackupVolumeInfo{
+				"backed-up-pv": {PVCNamespace: "fake-ns", PVCName: "fake-pvc-1", PVInfo: &volume.PVInfo{VolumeHandle: "vol-1"}},
+			},
+			runtimeScheme: scheme,
+			retPVRs: []*velerov1api.PodVolumeRestore{
+				completedPVR,
+			},
+		},
+		{
+			name: "in-place restore blocked when the PVC is bound to a different CSI volume",
+			pvbs: []*velerov1api.PodVolumeBackup{
+				createPVBObj(true, true, 1, "kopia"),
+			},
+			inplace: true,
+			kubeClientObj: []runtime.Object{
+				createNodeAgentDaemonset(),
+				func() *corev1api.PersistentVolume {
+					pv := createPVObj(1, false)
+					pv.Spec.CSI = &corev1api.CSIPersistentVolumeSource{Driver: "fake.csi", VolumeHandle: "vol-other"}
+					return pv
+				}(),
+				createPVCObj(1),
+			},
+			ctlClientObj: []runtime.Object{
+				createBackupRepoObj(),
+			},
+			restoredPod:     createPodObj(true, true, true, 1),
+			sourceNamespace: "fake-ns",
+			bsl:             "fake-bsl",
+			volumeInfos: map[string]volume.BackupVolumeInfo{
+				"fake-pv-1": {PVCNamespace: "fake-ns", PVCName: "fake-pvc-1", PVInfo: &volume.PVInfo{VolumeHandle: "vol-1"}},
+			},
+			runtimeScheme: scheme,
+			errs: []expectError{
+				{
+					err:        "in-place restore pre-flight check failed, skipping volume data restore: PVC fake-ns/fake-pvc-1 is bound to volume vol-other (PV fake-pv-1), but was bound to volume vol-1 (PV fake-pv-1) at backup time",
+					prefixOnly: true,
+				},
+			},
+		},
 		{
 			name: "in-place restore blocked when the PVC is too small for the source volume",
 			pvbs: []*velerov1api.PodVolumeBackup{
@@ -448,6 +517,7 @@ func TestRestorePodVolumes(t *testing.T) {
 			inplace: true,
 			kubeClientObj: []runtime.Object{
 				createNodeAgentDaemonset(),
+				createPVObj(1, false),
 				func() *corev1api.PersistentVolumeClaim {
 					pvc := createPVCObj(1)
 					pvc.Status.Capacity = corev1api.ResourceList{corev1api.ResourceStorage: resource.MustParse("100Mi")}
@@ -479,6 +549,7 @@ func TestRestorePodVolumes(t *testing.T) {
 			inplace: true,
 			kubeClientObj: []runtime.Object{
 				createNodeAgentDaemonset(),
+				createPVObj(1, false),
 				createNodeObj(),
 				createPVCObj(1),
 				createGatedPodObj("fake-restore-uid", 1),