Files
velero/pkg/podvolume/restorer.go
T
b7d83a6f2b Cherry pick the in-place restore implementation PRs from feature branch to main (#10415)
* Update CRDs and CLI to support in-place restore (#10038)

Update CRDs(Restore, DataDownload, PodVolumeRestore) and restore create CLI to support in-place restore

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>

* Update Kopia(filesystem) uploader to support incremental and deleteExtraFile during restore (#10066)

Update Kopia(filesystem) uploader to support incremental and deleteExtraFile during restore

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>

* Update Restore Exposer and PVC CSI to support in-place restore (#10104)

1. Update Restore Exposer to support exposing with existing PV for in-place restore
2. Update PVC CSI RIA to continue the restore process for in-place restore

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>

* Update Block uploader to support increase restore (#10244)

Update Block uploader to support increase restore

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>

* Update Exposer to recreate the target PV if the volume mode is different with the restore PVC (#10257)

Update Exposer to recreate the target PV if the volume mode is different with t
he restore PVC

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>

* Preserve PVC selected-node annotation via carrier annotation for in-place restore

For in-place volume data restore, the existing PVC is deleted and
recreated. For StorageClasses with the WaitForFirstConsumer volume
binding mode, losing the volume.kubernetes.io/selected-node annotation
could let the scheduler place the recreated workload Pod in a different
zone than the original PV, leaving it stuck in ContainerCreating.

Instead of relying on RestoreItemAction execution order (the generic
PVC RIA unconditionally strips the selected-node annotation), the PVC
CSI RIA now captures the annotation from the existing PVC right before
deleting it and carries it on the target PVC via the Velero-internal
restore.velero.io/inplace-restore-selected-node annotation. The restore
engine translates the carrier back to the Kubernetes annotation after
all RestoreItemActions have run and always strips the carrier so it
never lands on the cluster.

This makes the behavior independent of RIA ordering: the Kubernetes
annotation is stripped by default on every path (including when the
target PVC does not exist and Velero falls back to provisioning a new
PVC), and preservation only happens when the CSI RIA explicitly
captured a value from the existing PVC.

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

* Update the control path to make the in-place incremental restore with block data mover work E2E (#10410)

Update the control path to make the in-place incremental restore with block data mover work E2E

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>

---------

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>
Signed-off-by: chlins <chlins.zhang@gmail.com>
Co-authored-by: chlins <chlins.zhang@gmail.com>
2026-08-26 10:21:32 -04:00

332 lines
11 KiB
Go

/*
Copyright 2018 the Velero contributors.
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 podvolume
import (
"context"
"sync"
"time"
"github.com/vmware-tanzu/velero/internal/volume"
"github.com/cockroachdb/errors"
"github.com/sirupsen/logrus"
corev1api "k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/util/wait"
"k8s.io/client-go/kubernetes"
"k8s.io/client-go/tools/cache"
ctrlcache "sigs.k8s.io/controller-runtime/pkg/cache"
ctrlclient "sigs.k8s.io/controller-runtime/pkg/client"
velerov1api "github.com/vmware-tanzu/velero/pkg/apis/velero/v1"
veleroclient "github.com/vmware-tanzu/velero/pkg/client"
"github.com/vmware-tanzu/velero/pkg/label"
"github.com/vmware-tanzu/velero/pkg/nodeagent"
"github.com/vmware-tanzu/velero/pkg/repository"
uploaderutil "github.com/vmware-tanzu/velero/pkg/uploader/util"
"github.com/vmware-tanzu/velero/pkg/util/boolptr"
"github.com/vmware-tanzu/velero/pkg/util/kube"
)
type RestoreData struct {
Restore *velerov1api.Restore
Pod *corev1api.Pod
PodVolumeBackups []*velerov1api.PodVolumeBackup
SourceNamespace, BackupLocation string
}
// Restorer can execute pod volume restores of volumes in a pod.
type Restorer interface {
// RestorePodVolumes restores all annotated volumes in a pod.
RestorePodVolumes(RestoreData, *volume.RestoreVolumeInfoTracker) []error
}
type restorer struct {
ctx context.Context
repoLocker *repository.RepoLocker
repoEnsurer *repository.Ensurer
kubeClient kubernetes.Interface
crClient ctrlclient.Client
resultsLock sync.Mutex
results map[string]chan *velerov1api.PodVolumeRestore
log logrus.FieldLogger
}
func newRestorer(
ctx context.Context,
repoLocker *repository.RepoLocker,
repoEnsurer *repository.Ensurer,
pvrInformer ctrlcache.Informer,
kubeClient kubernetes.Interface,
crClient ctrlclient.Client,
restore *velerov1api.Restore,
log logrus.FieldLogger,
) *restorer {
r := &restorer{
ctx: ctx,
repoLocker: repoLocker,
repoEnsurer: repoEnsurer,
kubeClient: kubeClient,
crClient: crClient,
results: make(map[string]chan *velerov1api.PodVolumeRestore),
log: log,
}
_, _ = pvrInformer.AddEventHandler(
cache.ResourceEventHandlerFuncs{
UpdateFunc: func(oldObj, newObj any) {
pvr := newObj.(*velerov1api.PodVolumeRestore)
pvrOld := oldObj.(*velerov1api.PodVolumeRestore)
if pvr.GetLabels()[velerov1api.RestoreUIDLabel] != string(restore.UID) {
return
}
if pvr.Status.Phase == pvrOld.Status.Phase {
return
}
if pvr.Status.Phase == velerov1api.PodVolumeRestorePhaseCompleted || pvr.Status.Phase == velerov1api.PodVolumeRestorePhaseFailed || pvr.Status.Phase == velerov1api.PodVolumeRestorePhaseCanceled {
r.resultsLock.Lock()
resChan, ok := r.results[resultsKey(pvr.Spec.Pod.Namespace, pvr.Spec.Pod.Name)]
r.resultsLock.Unlock()
if !ok {
log.Errorf("No results channel found for pod %s/%s to send pod volume restore %s/%s on", pvr.Spec.Pod.Namespace, pvr.Spec.Pod.Name, pvr.Namespace, pvr.Name)
return
}
resChan <- pvr
}
},
},
)
return r
}
func (r *restorer) RestorePodVolumes(data RestoreData, tracker *volume.RestoreVolumeInfoTracker) []error {
volumesToRestore := getVolumeBackupInfoForPod(data.PodVolumeBackups, data.Pod, data.SourceNamespace)
if len(volumesToRestore) == 0 {
return nil
}
if err := nodeagent.IsRunningOnLinux(r.ctx, r.kubeClient, data.Restore.Namespace); err != nil {
return []error{errors.Wrapf(err, "error to check node agent status")}
}
repositoryType, err := getVolumesRepositoryType(volumesToRestore)
if err != nil {
return []error{err}
}
repo, err := r.repoEnsurer.EnsureRepo(r.ctx, data.Restore.Namespace, data.SourceNamespace, data.BackupLocation, repositoryType)
if err != nil {
return []error{err}
}
// get a single non-exclusive lock since we'll wait for all individual
// restores to be complete before releasing it.
r.repoLocker.Lock(repo.Name)
defer r.repoLocker.Unlock(repo.Name)
resultsChan := make(chan *velerov1api.PodVolumeRestore, len(volumesToRestore))
r.resultsLock.Lock()
r.results[resultsKey(data.Pod.Namespace, data.Pod.Name)] = resultsChan
r.resultsLock.Unlock()
nodeAgentCheck := make(chan error)
var (
errs []error
numRestores int
podVolumes = make(map[string]corev1api.Volume)
)
// put the pod's volumes in a map for efficient lookup below
for _, podVolume := range data.Pod.Spec.Volumes {
podVolumes[podVolume.Name] = podVolume
}
for volume, backupInfo := range volumesToRestore {
volumeObj, ok := podVolumes[volume]
var pvc *corev1api.PersistentVolumeClaim
if ok {
if volumeObj.PersistentVolumeClaim != nil {
pvc = new(corev1api.PersistentVolumeClaim)
err := r.crClient.Get(context.TODO(), ctrlclient.ObjectKey{Namespace: data.Pod.Namespace, Name: volumeObj.PersistentVolumeClaim.ClaimName}, pvc)
if err != nil {
errs = append(errs, errors.Wrap(err, "error getting persistent volume claim for volume"))
continue
}
}
}
volumeRestore := newPodVolumeRestore(data.Restore, data.Pod, data.BackupLocation, volume, backupInfo.snapshotID, backupInfo.snapshotSize, "", backupInfo.uploaderType, data.SourceNamespace, pvc)
if err := veleroclient.CreateRetryGenerateName(r.crClient, r.ctx, volumeRestore); err != nil {
errs = append(errs, errors.WithStack(err))
continue
}
numRestores++
}
checkCtx, checkCancel := context.WithCancel(context.Background())
go func() {
nodeName := ""
checkFunc := func(ctx context.Context) (bool, error) {
newObj, err := r.kubeClient.CoreV1().Pods(data.Pod.Namespace).Get(ctx, data.Pod.Name, metav1.GetOptions{})
if err != nil {
return false, err
}
nodeName = newObj.Spec.NodeName
err = kube.IsPodScheduled(newObj)
if err != nil {
r.log.WithField("error", err).Debugf("Pod %s/%s is not scheduled yet", newObj.GetNamespace(), newObj.GetName())
return false, nil
}
return true, nil
}
err := wait.PollUntilContextTimeout(checkCtx, time.Millisecond*500, time.Minute*10, true, checkFunc)
if wait.Interrupted(err) {
r.log.WithError(err).Error("Restoring pod is not scheduled until timeout or cancel, disengage")
} else if err != nil {
r.log.WithError(err).Error("Failed to check node-agent pod status, disengage")
} else {
err = nodeagent.IsRunningInNode(checkCtx, data.Restore.Namespace, nodeName, r.crClient)
if err != nil {
r.log.WithField("node", nodeName).WithError(err).Error("node-agent pod is not running in node, abort the restore")
nodeAgentCheck <- errors.Wrapf(err, "node-agent pod is not running in node %s", nodeName)
}
}
}()
ForEachVolume:
for i := 0; i < numRestores; i++ {
select {
case <-r.ctx.Done():
errs = append(errs, errors.New("timed out waiting for all PodVolumeRestores to complete"))
break ForEachVolume
case res := <-resultsChan:
if res.Status.Phase == velerov1api.PodVolumeRestorePhaseFailed {
errs = append(errs, errors.Errorf("pod volume restore failed: %s", res.Status.Message))
} else if res.Status.Phase == velerov1api.PodVolumeRestorePhaseCanceled {
errs = append(errs, errors.Errorf("pod volume restore canceled: %s", res.Status.Message))
}
tracker.TrackPodVolume(res)
case err := <-nodeAgentCheck:
errs = append(errs, err)
break ForEachVolume
}
}
// This is to prevent the case that resultsChan is signaled before nodeAgentCheck though this is unlikely possible.
// One possible case is that the CR is edited and set to an ending state manually, either completed or failed.
// In this case, we must notify the check routine to stop.
checkCancel()
r.resultsLock.Lock()
delete(r.results, resultsKey(data.Pod.Namespace, data.Pod.Name))
r.resultsLock.Unlock()
return errs
}
func newPodVolumeRestore(restore *velerov1api.Restore, pod *corev1api.Pod, backupLocation, volume, snapshot string, size int64, repoIdentifier, uploaderType, sourceNamespace string, pvc *corev1api.PersistentVolumeClaim) *velerov1api.PodVolumeRestore {
pvr := &velerov1api.PodVolumeRestore{
ObjectMeta: metav1.ObjectMeta{
Namespace: restore.Namespace,
GenerateName: restore.Name + "-",
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: velerov1api.SchemeGroupVersion.String(),
Kind: "Restore",
Name: restore.Name,
UID: restore.UID,
Controller: boolptr.True(),
},
},
Labels: map[string]string{
velerov1api.RestoreNameLabel: label.GetValidName(restore.Name),
velerov1api.RestoreUIDLabel: string(restore.UID),
velerov1api.PodUIDLabel: string(pod.UID),
},
},
Spec: velerov1api.PodVolumeRestoreSpec{
Pod: corev1api.ObjectReference{
Kind: "Pod",
Namespace: pod.Namespace,
Name: pod.Name,
UID: pod.UID,
},
Volume: volume,
SnapshotID: snapshot,
SnapshotSize: size,
BackupStorageLocation: backupLocation,
RepoIdentifier: repoIdentifier,
UploaderType: uploaderType,
SourceNamespace: sourceNamespace,
},
}
if pvc != nil {
// this label is not used by velero, but useful for debugging.
pvr.Labels[velerov1api.PVCUIDLabel] = string(pvc.UID)
}
if restore.Spec.UploaderConfig != nil {
pvr.Spec.UploaderSettings = uploaderutil.StoreRestoreConfig(restore.Spec.UploaderConfig)
}
if restore.IsVolumeDataInplaceRestore() {
pvr.Spec.RestoreType = string(restore.Spec.ExistingVolumeDataPolicy)
}
return pvr
}
func getVolumesRepositoryType(volumes map[string]volumeBackupInfo) (string, error) {
if len(volumes) == 0 {
return "", errors.New("empty volume list")
}
// the podVolumeBackups list come from one backup. In one backup, it is impossible that volumes are
// backed up by different uploaders or to different repositories. Asserting this ensures one repo only,
// which will simplify the following logics
repositoryType := ""
for _, backupInfo := range volumes {
if backupInfo.repositoryType == "" {
return "", errors.Errorf("empty repository type found among volume snapshots, snapshot ID %s, uploader %s",
backupInfo.snapshotID, backupInfo.uploaderType)
}
if repositoryType == "" {
repositoryType = backupInfo.repositoryType
} else if repositoryType != backupInfo.repositoryType {
return "", errors.Errorf("multiple repository type in one backup, current type %s, differential one [type %s, snapshot ID %s, uploader %s]",
repositoryType, backupInfo.repositoryType, backupInfo.snapshotID, backupInfo.uploaderType)
}
}
return repositoryType, nil
}