Skip to content

Commit

Permalink
Improve the concurrency for PVBs in different pods
Browse files Browse the repository at this point in the history
Improve the concurrency for PVBs in different pods

Fixes #6676

Signed-off-by: Wenkai Yin(尹文开) <yinw@vmware.com>
  • Loading branch information
ywk253100 committed Mar 29, 2024
1 parent d640cc1 commit e556a70
Show file tree
Hide file tree
Showing 7 changed files with 160 additions and 149 deletions.
1 change: 1 addition & 0 deletions changelogs/unreleased/7571-ywk253100
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Improve the concurrency for PVBs in different pods
3 changes: 3 additions & 0 deletions pkg/backup/backup.go
Original file line number Diff line number Diff line change
Expand Up @@ -419,6 +419,9 @@ func (kb *kubernetesBackupper) BackupWithResolvers(log logrus.FieldLogger,
}
}

processedPVBs := itemBackupper.podVolumeBackupper.WaitAllPodVolumesProcessed(log)
backupRequest.PodVolumeBackups = append(backupRequest.PodVolumeBackups, processedPVBs...)

// do a final update on progress since we may have just added some CRDs and may not have updated
// for the last few processed items.
updated = backupRequest.Backup.DeepCopy()
Expand Down
12 changes: 10 additions & 2 deletions pkg/backup/backup_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -3054,7 +3054,9 @@ func (f *fakePodVolumeBackupperFactory) NewBackupper(context.Context, *velerov1.
return &fakePodVolumeBackupper{}, nil
}

type fakePodVolumeBackupper struct{}
type fakePodVolumeBackupper struct {
pvbs []*velerov1.PodVolumeBackup
}

// BackupPodVolumes returns one pod volume backup per entry in volumes, with namespace "velero"
// and name "pvb-<pod-namespace>-<pod-name>-<volume-name>".
Expand All @@ -3072,9 +3074,15 @@ func (b *fakePodVolumeBackupper) BackupPodVolumes(backup *velerov1.Backup, pod *
res = append(res, pvb)
}

b.pvbs = res

return res, pvcSummary, nil
}

func (b *fakePodVolumeBackupper) WaitAllPodVolumesProcessed(log logrus.FieldLogger) []*velerov1.PodVolumeBackup {
return b.pvbs
}

// TestBackupWithPodVolume runs backups of pods that are annotated for PodVolume backup,
// and ensures that the pod volume backupper is called, that the returned PodVolumeBackups
// are added to the Request object, and that when PVCs are backed up with PodVolume, the
Expand Down Expand Up @@ -3289,7 +3297,7 @@ func newHarness(t *testing.T) *harness {

// unsupported
podCommandExecutor: nil,
podVolumeBackupperFactory: nil,
podVolumeBackupperFactory: new(fakePodVolumeBackupperFactory),
podVolumeTimeout: 0,
},
log: log,
Expand Down
1 change: 0 additions & 1 deletion pkg/backup/item_backupper.go
Original file line number Diff line number Diff line change
Expand Up @@ -270,7 +270,6 @@ func (ib *itemBackupper) backupItemInternal(logger logrus.FieldLogger, obj runti
// even if there are errors.
podVolumeBackups, podVolumePVCBackupSummary, errs := ib.backupPodVolumes(log, pod, pvbVolumes)

ib.backupRequest.PodVolumeBackups = append(ib.backupRequest.PodVolumeBackups, podVolumeBackups...)
backupErrs = append(backupErrs, errs...)

// Mark the volumes that has been processed by pod volume backup as Taken in the tracker.
Expand Down
97 changes: 48 additions & 49 deletions pkg/podvolume/backupper.go
Original file line number Diff line number Diff line change
Expand Up @@ -45,17 +45,19 @@ import (
type Backupper interface {
// BackupPodVolumes backs up all specified volumes in a pod.
BackupPodVolumes(backup *velerov1api.Backup, pod *corev1api.Pod, volumesToBackup []string, resPolicies *resourcepolicies.Policies, log logrus.FieldLogger) ([]*velerov1api.PodVolumeBackup, *PVCBackupSummary, []error)
WaitAllPodVolumesProcessed(log logrus.FieldLogger) []*velerov1api.PodVolumeBackup
}

type backupper struct {
ctx context.Context
repoLocker *repository.RepoLocker
repoEnsurer *repository.Ensurer
crClient ctrlclient.Client
uploaderType string

results map[string]chan *velerov1api.PodVolumeBackup
resultsLock sync.Mutex
ctx context.Context
repoLocker *repository.RepoLocker
repoEnsurer *repository.Ensurer
crClient ctrlclient.Client
uploaderType string
pvbInformer ctrlcache.Informer
handlerRegistration cache.ResourceEventHandlerRegistration
wg sync.WaitGroup
result []*velerov1api.PodVolumeBackup
}

type skippedPVC struct {
Expand Down Expand Up @@ -105,19 +107,19 @@ func newBackupper(
crClient ctrlclient.Client,
uploaderType string,
backup *velerov1api.Backup,
log logrus.FieldLogger,
) *backupper {
b := &backupper{
ctx: ctx,
repoLocker: repoLocker,
repoEnsurer: repoEnsurer,
crClient: crClient,
uploaderType: uploaderType,

results: make(map[string]chan *velerov1api.PodVolumeBackup),
pvbInformer: pvbInformer,
wg: sync.WaitGroup{},
result: []*velerov1api.PodVolumeBackup{},
}

_, _ = pvbInformer.AddEventHandler(
b.handlerRegistration, _ = pvbInformer.AddEventHandler(
cache.ResourceEventHandlerFuncs{
UpdateFunc: func(_, obj interface{}) {
pvb := obj.(*velerov1api.PodVolumeBackup)
Expand All @@ -126,17 +128,13 @@ func newBackupper(
return
}

if pvb.Status.Phase == velerov1api.PodVolumeBackupPhaseCompleted || pvb.Status.Phase == velerov1api.PodVolumeBackupPhaseFailed {
b.resultsLock.Lock()
defer b.resultsLock.Unlock()

resChan, ok := b.results[resultsKey(pvb.Spec.Pod.Namespace, pvb.Spec.Pod.Name)]
if !ok {
log.Errorf("No results channel found for pod %s/%s to send pod volume backup %s/%s on", pvb.Spec.Pod.Namespace, pvb.Spec.Pod.Name, pvb.Namespace, pvb.Name)
return
}
resChan <- pvb
if pvb.Status.Phase != velerov1api.PodVolumeBackupPhaseCompleted &&
pvb.Status.Phase != velerov1api.PodVolumeBackupPhaseFailed {
return

Check warning on line 133 in pkg/podvolume/backupper.go

View check run for this annotation

Codecov / codecov/patch

pkg/podvolume/backupper.go#L133

Added line #L133 was not covered by tests
}

b.result = append(b.result, pvb)
b.wg.Done()
},
},
)
Expand Down Expand Up @@ -217,12 +215,6 @@ func (b *backupper) BackupPodVolumes(backup *velerov1api.Backup, pod *corev1api.
b.repoLocker.Lock(repo.Name)
defer b.repoLocker.Unlock(repo.Name)

resultsChan := make(chan *velerov1api.PodVolumeBackup)

b.resultsLock.Lock()
b.results[resultsKey(pod.Namespace, pod.Name)] = resultsChan
b.resultsLock.Unlock()

var (
podVolumeBackups []*velerov1api.PodVolumeBackup
mountedPodVolumes = sets.Set[string]{}
Expand All @@ -243,7 +235,6 @@ func (b *backupper) BackupPodVolumes(backup *velerov1api.Backup, pod *corev1api.
repoIdentifier = repo.Spec.ResticIdentifier
}

var numVolumeSnapshots int
for _, volumeName := range volumesToBackup {
volume, ok := podVolumes[volumeName]
if !ok {
Expand Down Expand Up @@ -305,32 +296,40 @@ func (b *backupper) BackupPodVolumes(backup *velerov1api.Backup, pod *corev1api.
errs = append(errs, err)
continue
}
b.wg.Add(1)
podVolumeBackups = append(podVolumeBackups, volumeBackup)

Check warning on line 300 in pkg/podvolume/backupper.go

View check run for this annotation

Codecov / codecov/patch

pkg/podvolume/backupper.go#L299-L300

Added lines #L299 - L300 were not covered by tests
pvcSummary.addBackedup(volumeName)
numVolumeSnapshots++
}

ForEachVolume:
for i, count := 0, numVolumeSnapshots; i < count; i++ {
select {
case <-b.ctx.Done():
errs = append(errs, errors.New("timed out waiting for all PodVolumeBackups to complete"))
break ForEachVolume
case res := <-resultsChan:
switch res.Status.Phase {
case velerov1api.PodVolumeBackupPhaseCompleted:
podVolumeBackups = append(podVolumeBackups, res)
case velerov1api.PodVolumeBackupPhaseFailed:
errs = append(errs, errors.Errorf("pod volume backup failed: %s", res.Status.Message))
podVolumeBackups = append(podVolumeBackups, res)
return podVolumeBackups, pvcSummary, errs
}

func (b *backupper) WaitAllPodVolumesProcessed(log logrus.FieldLogger) []*velerov1api.PodVolumeBackup {
defer func() {
if err := b.pvbInformer.RemoveEventHandler(b.handlerRegistration); err != nil {
log.Debugf("failed to remove the event handler for PVB: %v", err)
}

Check warning on line 311 in pkg/podvolume/backupper.go

View check run for this annotation

Codecov / codecov/patch

pkg/podvolume/backupper.go#L310-L311

Added lines #L310 - L311 were not covered by tests
}()

done := make(chan struct{})
go func() {
defer close(done)
b.wg.Wait()
}()

var podVolumeBackups []*velerov1api.PodVolumeBackup
select {
case <-b.ctx.Done():
log.Error("timed out waiting for all PodVolumeBackups to complete")
case <-done:
for _, pvb := range b.result {
podVolumeBackups = append(podVolumeBackups, pvb)
if pvb.Status.Phase == velerov1api.PodVolumeBackupPhaseFailed {
log.Errorf("pod volume backup failed: %s", pvb.Status.Message)
}
}
}

b.resultsLock.Lock()
delete(b.results, resultsKey(pod.Namespace, pod.Name))
b.resultsLock.Unlock()

return podVolumeBackups, pvcSummary, errs
return podVolumeBackups
}

func skipAllPodVolumes(pod *corev1api.Pod, volumesToBackup []string, err error, pvcSummary *PVCBackupSummary, log logrus.FieldLogger) {
Expand Down
2 changes: 1 addition & 1 deletion pkg/podvolume/backupper_factory.go
Original file line number Diff line number Diff line change
Expand Up @@ -60,7 +60,7 @@ type backupperFactory struct {
}

func (bf *backupperFactory) NewBackupper(ctx context.Context, backup *velerov1api.Backup, uploaderType string) (Backupper, error) {
b := newBackupper(ctx, bf.repoLocker, bf.repoEnsurer, bf.pvbInformer, bf.crClient, uploaderType, backup, bf.log)
b := newBackupper(ctx, bf.repoLocker, bf.repoEnsurer, bf.pvbInformer, bf.crClient, uploaderType, backup)

if !cache.WaitForCacheSync(ctx.Done(), bf.pvbInformer.HasSynced) {
return nil, errors.New("timed out waiting for caches to sync")
Expand Down
Loading

0 comments on commit e556a70

Please sign in to comment.