mirrors/velero
1
0
Fork 0
mirror of https://github.com/vmware-tanzu/velero.git synced 2026-01-06 21:36:30 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

data mover restore for Windows

Browse Source 
Signed-off-by: Lyndon-Li <lyonghui@vmware.com>
This commit is contained in:
Lyndon-Li
2025-01-08 17:46:49 +08:00
parent be5f56ab18
commit a8469126d8
19 changed files with 294 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
changelogs/unreleased/8594-Lyndon-Li Normal file
View File

@@ -0,0 +1 @@
Data mover restore for Windows

7
config/crd/v2alpha1/bases/velero.io_datadownloads.yaml
View File

@@ -92,6 +92,13 @@ spec:
DataMover specifies the data mover to be used by the backup.
If DataMover is "" or "velero", the built-in data mover will be used.
type: string
nodeOS:
description: NodeOS is OS of the node where the DataDownload is processed.
enum:
- auto
- linux
- windows
type: string
operationTimeout:
description: |-
OperationTimeout specifies the time used to wait internal operations,

10
config/crd/v2alpha1/bases/velero.io_datauploads.yaml
View File

@@ -144,7 +144,8 @@ spec:
description: DataUploadStatus is the current status of a DataUpload.
properties:
acceptedByNode:
description: Node is name of the node where the DataUpload is prepared.
description: AcceptedByNode is name of the node where the DataUpload
is prepared.
type: string
acceptedTimestamp:
description: |-
@@ -175,6 +176,13 @@ spec:
node:
description: Node is name of the node where the DataUpload is processed.
type: string
nodeOS:
description: NodeOS is OS of the node where the DataUpload is processed.
enum:
- auto
- linux
- windows
type: string
path:
description: Path is the full path of the snapshot volume being backed
up.

4
config/crd/v2alpha1/crds/crds.go
View File

File diff suppressed because one or more lines are too long

4
pkg/apis/velero/v2alpha1/data_download_types.go
View File

@@ -54,6 +54,10 @@ type DataDownloadSpec struct {
// OperationTimeout specifies the time used to wait internal operations,
// before returning error as timeout.
OperationTimeout metav1.Duration `json:"operationTimeout"`
// NodeOS is OS of the node where the DataDownload is processed.
// +optional
NodeOS NodeOS `json:"nodeOS,omitempty"`
}
// TargetVolumeSpec is the specification for a target PVC.

21
pkg/apis/velero/v2alpha1/data_upload_types.go
View File

@@ -96,6 +96,16 @@ const (
DataUploadPhaseFailed DataUploadPhase = "Failed"
)
// NodeOS represents OS of a node.
// +kubebuilder:validation:Enum=auto;linux;windows
type NodeOS string
const (
NodeOSLinux NodeOS = "linux"
NodeOSWindows NodeOS = "windows"
NodeOSAuto NodeOS = "auto"
)
// DataUploadStatus is the current status of a DataUpload.
type DataUploadStatus struct {
// Phase is the current state of the DataUpload.
@@ -144,7 +154,12 @@ type DataUploadStatus struct {
// Node is name of the node where the DataUpload is processed.
// +optional
Node string `json:"node,omitempty"`
// Node is name of the node where the DataUpload is prepared.
// NodeOS is OS of the node where the DataUpload is processed.
// +optional
NodeOS NodeOS `json:"nodeOS,omitempty"`
// AcceptedByNode is name of the node where the DataUpload is prepared.
// +optional
AcceptedByNode string `json:"acceptedByNode,omitempty"`
@@ -221,4 +236,8 @@ type DataUploadResult struct {
// +optional
// +nullable
DataMoverResult *map[string]string `json:"dataMoverResult,omitempty"`
// NodeOS is OS of the node where the DataUpload is processed.
// +optional
NodeOS NodeOS `json:"nodeOS,omitempty"`
}

6
pkg/builder/data_download_builder.go
View File

@@ -148,6 +148,12 @@ func (d *DataDownloadBuilder) Node(node string) *DataDownloadBuilder {
return d
}
// NodeOS sets the DataDownload's Node OS.
func (d *DataDownloadBuilder) NodeOS(nodeOS velerov2alpha1api.NodeOS) *DataDownloadBuilder {
d.object.Spec.NodeOS = nodeOS
return d
}
// AcceptedByNode sets the DataDownload's AcceptedByNode.
func (d *DataDownloadBuilder) AcceptedByNode(node string) *DataDownloadBuilder {
d.object.Status.AcceptedByNode = node

6
pkg/builder/data_upload_builder.go
View File

@@ -151,6 +151,12 @@ func (d *DataUploadBuilder) Node(node string) *DataUploadBuilder {
return d
}
// NodeOS sets the DataUpload's Node OS.
func (d *DataUploadBuilder) NodeOS(nodeOS velerov2alpha1api.NodeOS) *DataUploadBuilder {
d.object.Status.NodeOS = nodeOS
return d
}
// AcceptedByNode sets the DataUpload's AcceptedByNode.
func (d *DataUploadBuilder) AcceptedByNode(node string) *DataUploadBuilder {
d.object.Status.AcceptedByNode = node

19
pkg/cmd/cli/datamover/restore.go
View File

@@ -160,7 +160,24 @@ func newdataMoverRestore(logger logrus.FieldLogger, factory client.Factory, conf
return nil, errors.Wrap(err, "error to create client")
}
cache, err := ctlcache.New(clientConfig, cacheOption)
var cache ctlcache.Cache
retry := 10
for {
cache, err = ctlcache.New(clientConfig, cacheOption)
if err == nil {
break
}
retry--
if retry == 0 {
break
}
logger.WithError(err).Warn("Failed to create client cache, need retry")
time.Sleep(time.Second)
}
if err != nil {
cancelFunc()
return nil, errors.Wrap(err, "error to create client cache")

59
pkg/controller/data_download_controller.go
View File

@@ -183,28 +183,15 @@ func (r *DataDownloadReconciler) Reconcile(ctx context.Context, req ctrl.Request
return ctrl.Result{}, nil
}
hostingPodLabels := map[string]string{velerov1api.DataDownloadLabel: dd.Name}
for _, k := range util.ThirdPartyLabels {
if v, err := nodeagent.GetLabelValue(ctx, r.kubeClient, dd.Namespace, k, kube.NodeOSLinux); err != nil {
if err != nodeagent.ErrNodeAgentLabelNotFound {
log.WithError(err).Warnf("Failed to check node-agent label, skip adding host pod label %s", k)
}
} else {
hostingPodLabels[k] = v
}
exposeParam, err := r.setupExposeParam(dd)
if err != nil {
return r.errorOut(ctx, dd, err, "failed to set exposer parameters", log)
}
// Expose() will trigger to create one pod whose volume is restored by a given volume snapshot,
// but the pod maybe is not in the same node of the current controller, so we need to return it here.
// And then only the controller who is in the same node could do the rest work.
err = r.restoreExposer.Expose(ctx, getDataDownloadOwnerObject(dd), exposer.GenericRestoreExposeParam{
TargetPVCName: dd.Spec.TargetVolume.PVC,
SourceNamespace: dd.Spec.TargetVolume.Namespace,
HostingPodLabels: hostingPodLabels,
Resources: r.podResources,
ExposeTimeout: dd.Spec.OperationTimeout.Duration,
RestorePVCConfig: r.restorePVCConfig,
})
err = r.restoreExposer.Expose(ctx, getDataDownloadOwnerObject(dd), exposeParam)
if err != nil {
if err := r.client.Get(ctx, req.NamespacedName, dd); err != nil {
if !apierrors.IsNotFound(err) {
@@ -243,7 +230,7 @@ func (r *DataDownloadReconciler) Reconcile(ctx context.Context, req ctrl.Request
log.Debugf("Data download is been canceled %s in Phase %s", dd.GetName(), dd.Status.Phase)
r.tryCancelAcceptedDataDownload(ctx, dd, "")
} else if peekErr := r.restoreExposer.PeekExposed(ctx, getDataDownloadOwnerObject(dd)); peekErr != nil {
r.tryCancelAcceptedDataDownload(ctx, dd, fmt.Sprintf("found a dataupload %s/%s with expose error: %s. mark it as cancel", dd.Namespace, dd.Name, peekErr))
r.tryCancelAcceptedDataDownload(ctx, dd, fmt.Sprintf("found a datadownload %s/%s with expose error: %s. mark it as cancel", dd.Namespace, dd.Name, peekErr))
log.Errorf("Cancel dd %s/%s because of expose error %s", dd.Namespace, dd.Name, peekErr)
} else if dd.Status.AcceptedTimestamp != nil {
if time.Since(dd.Status.AcceptedTimestamp.Time) >= r.preparingTimeout {
@@ -737,6 +724,42 @@ func (r *DataDownloadReconciler) closeDataPath(ctx context.Context, ddName strin
r.dataPathMgr.RemoveAsyncBR(ddName)
}
func (r *DataDownloadReconciler) setupExposeParam(dd *velerov2alpha1api.DataDownload) (exposer.GenericRestoreExposeParam, error) {
log := r.logger.WithField("datadownload", dd.Name)
nodeOS := string(dd.Spec.NodeOS)
if nodeOS == "" {
log.Info("nodeOS is empty in DD, fallback to linux")
nodeOS = kube.NodeOSLinux
}
if err := kube.HasNodeWithOS(context.Background(), nodeOS, r.kubeClient.CoreV1()); err != nil {
return exposer.GenericRestoreExposeParam{}, errors.Wrapf(err, "no appropriate node to run datadownload %s/%s", dd.Namespace, dd.Name)
}
hostingPodLabels := map[string]string{velerov1api.DataDownloadLabel: dd.Name}
for _, k := range util.ThirdPartyLabels {
if v, err := nodeagent.GetLabelValue(context.Background(), r.kubeClient, dd.Namespace, k, nodeOS); err != nil {
if err != nodeagent.ErrNodeAgentLabelNotFound {
log.WithError(err).Warnf("Failed to check node-agent label, skip adding host pod label %s", k)
}
} else {
hostingPodLabels[k] = v
}
}
return exposer.GenericRestoreExposeParam{
TargetPVCName: dd.Spec.TargetVolume.PVC,
TargetNamespace: dd.Spec.TargetVolume.Namespace,
HostingPodLabels: hostingPodLabels,
Resources: r.podResources,
OperationTimeout: dd.Spec.OperationTimeout.Duration,
ExposeTimeout: r.preparingTimeout,
NodeOS: nodeOS,
RestorePVCConfig: r.restorePVCConfig,
}, nil
}
func getDataDownloadOwnerObject(dd *velerov2alpha1api.DataDownload) v1.ObjectReference {
return v1.ObjectReference{
Kind: dd.Kind,

17
pkg/controller/data_download_controller_test.go
View File

@@ -53,6 +53,7 @@ import (
"github.com/vmware-tanzu/velero/pkg/nodeagent"
velerotest "github.com/vmware-tanzu/velero/pkg/test"
"github.com/vmware-tanzu/velero/pkg/uploader"
"github.com/vmware-tanzu/velero/pkg/util/kube"
exposermockes "github.com/vmware-tanzu/velero/pkg/exposer/mocks"
)
@@ -67,7 +68,7 @@ func dataDownloadBuilder() *builder.DataDownloadBuilder {
PV: "test-pv",
PVC: "test-pvc",
Namespace: "test-ns",
})
}).NodeOS(velerov2alpha1api.NodeOS("linux"))
}
func initDataDownloadReconciler(objects []runtime.Object, needError ...bool) (*DataDownloadReconciler, error) {
@@ -167,6 +168,8 @@ func TestDataDownloadReconcile(t *testing.T) {
},
}
node := builder.ForNode("fake-node").Labels(map[string]string{kube.NodeOSLabel: kube.NodeOSLinux}).Result()
tests := []struct {
name string
dd *velerov2alpha1api.DataDownload
@@ -326,9 +329,15 @@ func TestDataDownloadReconcile(t *testing.T) {
},
{
name: "Restore is exposed",
dd: dataDownloadBuilder().Result(),
dd: dataDownloadBuilder().NodeOS(velerov2alpha1api.NodeOSLinux).Result(),
targetPVC: builder.ForPersistentVolumeClaim("test-ns", "test-pvc").Result(),
},
{
name: "Expected node doesn't exist",
dd: dataDownloadBuilder().NodeOS(velerov2alpha1api.NodeOSWindows).Result(),
targetPVC: builder.ForPersistentVolumeClaim("test-ns", "test-pvc").Result(),
expectedStatusMsg: "no appropriate node to run datadownload",
},
{
name: "Get empty restore exposer",
dd: dataDownloadBuilder().Phase(velerov2alpha1api.DataDownloadPhasePrepared).Result(),
@@ -388,9 +397,9 @@ func TestDataDownloadReconcile(t *testing.T) {
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
var objs []runtime.Object
objs := []runtime.Object{daemonSet, node}
if test.targetPVC != nil {
objs = []runtime.Object{test.targetPVC, daemonSet}
objs = append(objs, test.targetPVC)
}
r, err := initDataDownloadReconciler(objs, test.needErrs...)
require.NoError(t, err)

12
pkg/controller/data_upload_controller.go
View File

@@ -285,6 +285,10 @@ func (r *DataUploadReconciler) Reconcile(ctx context.Context, req ctrl.Request)
return ctrl.Result{}, nil
}
if res.ByPod.NodeOS == nil {
return r.errorOut(ctx, du, errors.New("unsupported ambiguous node OS"), "invalid expose result", log)
}
log.Info("Exposed snapshot is ready and creating data path routine")
// Need to first create file system BR and get data path instance then update data upload status
@@ -317,6 +321,7 @@ func (r *DataUploadReconciler) Reconcile(ctx context.Context, req ctrl.Request)
original := du.DeepCopy()
du.Status.Phase = velerov2alpha1api.DataUploadPhaseInProgress
du.Status.StartTimestamp = &metav1.Time{Time: r.Clock.Now()}
du.Status.NodeOS = velerov2alpha1api.NodeOS(*res.ByPod.NodeOS)
if err := r.client.Patch(ctx, du, client.MergeFrom(original)); err != nil {
log.WithError(err).Warnf("Failed to update dataupload %s to InProgress, will data path close and retry", du.Name)
@@ -792,6 +797,8 @@ func (r *DataUploadReconciler) closeDataPath(ctx context.Context, duName string)
}
func (r *DataUploadReconciler) setupExposeParam(du *velerov2alpha1api.DataUpload) (interface{}, error) {
log := r.logger.WithField("dataupload", du.Name)
if du.Spec.SnapshotType == velerov2alpha1api.SnapshotTypeCSI {
pvc := &corev1.PersistentVolumeClaim{}
err := r.client.Get(context.Background(), types.NamespacedName{
@@ -803,7 +810,7 @@ func (r *DataUploadReconciler) setupExposeParam(du *velerov2alpha1api.DataUpload
return nil, errors.Wrapf(err, "failed to get PVC %s/%s", du.Spec.SourceNamespace, du.Spec.SourcePVC)
}
nodeOS, err := kube.GetPVCAttachingNodeOS(pvc, r.kubeClient.CoreV1(), r.kubeClient.StorageV1(), r.logger)
nodeOS, err := kube.GetPVCAttachingNodeOS(pvc, r.kubeClient.CoreV1(), r.kubeClient.StorageV1(), log)
if err != nil {
return nil, errors.Wrapf(err, "failed to get attaching node OS for PVC %s/%s", du.Spec.SourceNamespace, du.Spec.SourcePVC)
}
@@ -821,7 +828,7 @@ func (r *DataUploadReconciler) setupExposeParam(du *velerov2alpha1api.DataUpload
for _, k := range util.ThirdPartyLabels {
if v, err := nodeagent.GetLabelValue(context.Background(), r.kubeClient, du.Namespace, k, nodeOS); err != nil {
if err != nodeagent.ErrNodeAgentLabelNotFound {
r.logger.WithError(err).Warnf("Failed to check node-agent label, skip adding host pod label %s", k)
log.WithError(err).Warnf("Failed to check node-agent label, skip adding host pod label %s", k)
}
} else {
hostingPodLabels[k] = v
@@ -843,6 +850,7 @@ func (r *DataUploadReconciler) setupExposeParam(du *velerov2alpha1api.DataUpload
NodeOS: nodeOS,
}, nil
}
return nil, nil
}

67
pkg/controller/data_upload_controller_test.go
View File

@@ -166,6 +166,7 @@ func initDataUploaderReconcilerWithError(needError ...error) (*DataUploadReconci
RestoreSize: &restoreSize,
},
}
daemonSet := &appsv1.DaemonSet{
ObjectMeta: metav1.ObjectMeta{
Namespace: "velero",
@@ -265,9 +266,10 @@ func dataUploadBuilder() *builder.DataUploadBuilder {
}
type fakeSnapshotExposer struct {
kubeClient kbclient.Client
clock clock.WithTickerAndDelayedExecution
peekErr error
kubeClient kbclient.Client
clock clock.WithTickerAndDelayedExecution
ambiguousNodeOS bool
peekErr error
}
func (f *fakeSnapshotExposer) Expose(ctx context.Context, ownerObject corev1.ObjectReference, param interface{}) error {
@@ -296,7 +298,13 @@ func (f *fakeSnapshotExposer) GetExposed(ctx context.Context, du corev1.ObjectRe
if err != nil {
return nil, err
}
return &exposer.ExposeResult{ByPod: exposer.ExposeByPod{HostingPod: pod, VolumeName: dataUploadName}}, nil
nodeOS := "linux"
pNodeOS := &nodeOS
if f.ambiguousNodeOS {
pNodeOS = nil
}
return &exposer.ExposeResult{ByPod: exposer.ExposeByPod{HostingPod: pod, VolumeName: dataUploadName, NodeOS: pNodeOS}}, nil
}
func (f *fakeSnapshotExposer) PeekExposed(ctx context.Context, ownerObject corev1.ObjectReference) error {
@@ -350,6 +358,8 @@ func TestReconcile(t *testing.T) {
expectedRequeue ctrl.Result
expectedErrMsg string
needErrs []bool
removeNode bool
ambiguousNodeOS bool
peekErr error
notCreateFSBR bool
fsBRInitErr error
@@ -359,25 +369,29 @@ func TestReconcile(t *testing.T) {
name: "Dataupload is not initialized",
du: builder.ForDataUpload("unknown-ns", "unknown-name").Result(),
expectedRequeue: ctrl.Result{},
}, {
},
{
name: "Error get Dataupload",
du: builder.ForDataUpload(velerov1api.DefaultNamespace, "unknown-name").Result(),
expectedRequeue: ctrl.Result{},
expectedErrMsg: "getting DataUpload: Get error",
needErrs: []bool{true, false, false, false},
}, {
},
{
name: "Unsupported data mover type",
du: dataUploadBuilder().DataMover("unknown type").Result(),
expected: dataUploadBuilder().Phase("").Result(),
expectedRequeue: ctrl.Result{},
}, {
},
{
name: "Unknown type of snapshot exposer is not initialized",
du: dataUploadBuilder().SnapshotType("unknown type").Result(),
expectedProcessed: true,
expected: dataUploadBuilder().Phase(velerov2alpha1api.DataUploadPhaseFailed).Result(),
expectedRequeue: ctrl.Result{},
expectedErrMsg: "unknown type type of snapshot exposer is not exist",
}, {
},
{
name: "Dataupload should be accepted",
du: dataUploadBuilder().Result(),
pod: builder.ForPod("fake-ns", dataUploadName).Volumes(&corev1.Volume{Name: "test-pvc"}).Result(),
@@ -394,6 +408,27 @@ func TestReconcile(t *testing.T) {
expectedRequeue: ctrl.Result{},
expectedErrMsg: "failed to get PVC",
},
{
name: "Dataupload should fail to get PVC attaching node",
du: dataUploadBuilder().Result(),
pod: builder.ForPod("fake-ns", dataUploadName).Volumes(&corev1.Volume{Name: "test-pvc"}).Result(),
pvc: builder.ForPersistentVolumeClaim("fake-ns", "test-pvc").StorageClass("fake-sc").Result(),
expectedProcessed: true,
expected: dataUploadBuilder().Phase(velerov2alpha1api.DataUploadPhaseFailed).Result(),
expectedRequeue: ctrl.Result{},
expectedErrMsg: "error to get storage class",
},
{
name: "Dataupload should fail because expected node doesn't exist",
du: dataUploadBuilder().Result(),
pod: builder.ForPod("fake-ns", dataUploadName).Volumes(&corev1.Volume{Name: "test-pvc"}).Result(),
pvc: builder.ForPersistentVolumeClaim("fake-ns", "test-pvc").Result(),
removeNode: true,
expectedProcessed: true,
expected: dataUploadBuilder().Phase(velerov2alpha1api.DataUploadPhaseFailed).Result(),
expectedRequeue: ctrl.Result{},
expectedErrMsg: "no appropriate node to run data upload",
},
{
name: "Dataupload should be prepared",
du: dataUploadBuilder().SnapshotType(fakeSnapshotType).Result(),
@@ -407,6 +442,15 @@ func TestReconcile(t *testing.T) {
expected: dataUploadBuilder().Phase(velerov2alpha1api.DataUploadPhaseInProgress).Result(),
expectedRequeue: ctrl.Result{},
},
{
name: "Dataupload should fail if expose returns ambiguous nodeOS",
pod: builder.ForPod(velerov1api.DefaultNamespace, dataUploadName).Volumes(&corev1.Volume{Name: "dataupload-1"}).Result(),
du: dataUploadBuilder().Phase(velerov2alpha1api.DataUploadPhasePrepared).SnapshotType(fakeSnapshotType).Result(),
ambiguousNodeOS: true,
expectedProcessed: true,
expected: dataUploadBuilder().Phase(velerov2alpha1api.DataUploadPhaseFailed).Result(),
expectedErrMsg: "unsupported ambiguous node OS",
},
{
name: "Dataupload with not enabled cancel",
pod: builder.ForPod(velerov1api.DefaultNamespace, dataUploadName).Volumes(&corev1.Volume{Name: "dataupload-1"}).Result(),
@@ -557,6 +601,11 @@ func TestReconcile(t *testing.T) {
require.NoError(t, err)
}
if test.removeNode {
err = r.kubeClient.CoreV1().Nodes().Delete(ctx, "fake-node", metav1.DeleteOptions{})
require.NoError(t, err)
}
if test.dataMgr != nil {
r.dataPathMgr = test.dataMgr
} else {
@@ -564,7 +613,7 @@ func TestReconcile(t *testing.T) {
}
if test.du.Spec.SnapshotType == fakeSnapshotType {
r.snapshotExposerList = map[velerov2alpha1api.SnapshotType]exposer.SnapshotExposer{fakeSnapshotType: &fakeSnapshotExposer{r.client, r.Clock, test.peekErr}}
r.snapshotExposerList = map[velerov2alpha1api.SnapshotType]exposer.SnapshotExposer{fakeSnapshotType: &fakeSnapshotExposer{r.client, r.Clock, test.ambiguousNodeOS, test.peekErr}}
} else if test.du.Spec.SnapshotType == velerov2alpha1api.SnapshotTypeCSI {
r.snapshotExposerList = map[velerov2alpha1api.SnapshotType]exposer.SnapshotExposer{velerov2alpha1api.SnapshotTypeCSI: exposer.NewCSISnapshotExposer(r.kubeClient, r.csiSnapshotClient, velerotest.NewLogger())}
}

6
pkg/exposer/csi_snapshot.go
View File

@@ -281,10 +281,16 @@ func (e *csiSnapshotExposer) GetExposed(ctx context.Context, ownerObject corev1.
curLog.WithField("pod", pod.Name).Infof("Backup volume is found in pod at index %v", i)
var nodeOS *string
if os, found := pod.Spec.NodeSelector[kube.NodeOSLabel]; found {
nodeOS = &os
}
return &ExposeResult{ByPod: ExposeByPod{
HostingPod: pod,
HostingContainer: containerName,
VolumeName: volumeName,
NodeOS: nodeOS,
}}, nil
}

59
pkg/exposer/generic_restore.go
View File

@@ -40,8 +40,8 @@ type GenericRestoreExposeParam struct {
// TargetPVCName is the target volume name to be restored
TargetPVCName string
// SourceNamespace is the original namespace of the volume that the snapshot is taken for
SourceNamespace string
// TargetNamespace is the namespace of the volume to be restored
TargetNamespace string
// HostingPodLabels is the labels that are going to apply to the hosting pod
HostingPodLabels map[string]string
@@ -52,6 +52,12 @@ type GenericRestoreExposeParam struct {
// ExposeTimeout specifies the timeout for the entire expose process
ExposeTimeout time.Duration
// OperationTimeout specifies the time wait for resources operations in Expose
OperationTimeout time.Duration
// NodeOS specifies the OS of node that the volume should be attached
NodeOS string
// RestorePVCConfig is the config for restorePVC (intermediate PVC) of generic restore
RestorePVCConfig nodeagent.RestorePVC
}
@@ -99,21 +105,21 @@ func (e *genericRestoreExposer) Expose(ctx context.Context, ownerObject corev1.O
curLog := e.log.WithFields(logrus.Fields{
"owner": ownerObject.Name,
"target PVC": param.TargetPVCName,
"source namespace": param.SourceNamespace,
"target namespace": param.TargetNamespace,
})
selectedNode, targetPVC, err := kube.WaitPVCConsumed(ctx, e.kubeClient.CoreV1(), param.TargetPVCName, param.SourceNamespace, e.kubeClient.StorageV1(), param.ExposeTimeout, param.RestorePVCConfig.IgnoreDelayBinding)
selectedNode, targetPVC, err := kube.WaitPVCConsumed(ctx, e.kubeClient.CoreV1(), param.TargetPVCName, param.TargetNamespace, e.kubeClient.StorageV1(), param.ExposeTimeout, param.RestorePVCConfig.IgnoreDelayBinding)
if err != nil {
return errors.Wrapf(err, "error to wait target PVC consumed, %s/%s", param.SourceNamespace, param.TargetPVCName)
return errors.Wrapf(err, "error to wait target PVC consumed, %s/%s", param.TargetNamespace, param.TargetPVCName)
}
curLog.WithField("target PVC", param.TargetPVCName).WithField("selected node", selectedNode).Info("Target PVC is consumed")
if kube.IsPVCBound(targetPVC) {
return errors.Errorf("Target PVC %s/%s has already been bound, abort", param.SourceNamespace, param.TargetPVCName)
return errors.Errorf("Target PVC %s/%s has already been bound, abort", param.TargetNamespace, param.TargetPVCName)
}
restorePod, err := e.createRestorePod(ctx, ownerObject, targetPVC, param.ExposeTimeout, param.HostingPodLabels, selectedNode, param.Resources)
restorePod, err := e.createRestorePod(ctx, ownerObject, targetPVC, param.OperationTimeout, param.HostingPodLabels, selectedNode, param.Resources, param.NodeOS)
if err != nil {
return errors.Wrapf(err, "error to create restore pod")
}
@@ -274,19 +280,19 @@ func (e *genericRestoreExposer) CleanUp(ctx context.Context, ownerObject corev1.
kube.DeletePVAndPVCIfAny(ctx, e.kubeClient.CoreV1(), restorePVCName, ownerObject.Namespace, 0, e.log)
}
func (e *genericRestoreExposer) RebindVolume(ctx context.Context, ownerObject corev1.ObjectReference, targetPVCName string, sourceNamespace string, timeout time.Duration) error {
func (e *genericRestoreExposer) RebindVolume(ctx context.Context, ownerObject corev1.ObjectReference, targetPVCName string, targetNamespace string, timeout time.Duration) error {
restorePodName := ownerObject.Name
restorePVCName := ownerObject.Name
curLog := e.log.WithFields(logrus.Fields{
"owner": ownerObject.Name,
"target PVC": targetPVCName,
"source namespace": sourceNamespace,
"target namespace": targetNamespace,
})
targetPVC, err := e.kubeClient.CoreV1().PersistentVolumeClaims(sourceNamespace).Get(ctx, targetPVCName, metav1.GetOptions{})
targetPVC, err := e.kubeClient.CoreV1().PersistentVolumeClaims(targetNamespace).Get(ctx, targetPVCName, metav1.GetOptions{})
if err != nil {
return errors.Wrapf(err, "error to get target PVC %s/%s", sourceNamespace, targetPVCName)
return errors.Wrapf(err, "error to get target PVC %s/%s", targetNamespace, targetPVCName)
}
restorePV, err := kube.WaitPVCBound(ctx, e.kubeClient.CoreV1(), e.kubeClient.CoreV1(), restorePVCName, ownerObject.Namespace, timeout)
@@ -368,7 +374,7 @@ func (e *genericRestoreExposer) RebindVolume(ctx context.Context, ownerObject co
}
func (e *genericRestoreExposer) createRestorePod(ctx context.Context, ownerObject corev1.ObjectReference, targetPVC *corev1.PersistentVolumeClaim,
operationTimeout time.Duration, label map[string]string, selectedNode string, resources corev1.ResourceRequirements) (*corev1.Pod, error) {
operationTimeout time.Duration, label map[string]string, selectedNode string, resources corev1.ResourceRequirements, nodeType string) (*corev1.Pod, error) {
restorePodName := ownerObject.Name
restorePVCName := ownerObject.Name
@@ -409,7 +415,28 @@ func (e *genericRestoreExposer) createRestorePod(ctx context.Context, ownerObjec
args = append(args, podInfo.logFormatArgs...)
args = append(args, podInfo.logLevelArgs...)
userID := int64(0)
var securityCtx *corev1.PodSecurityContext
nodeSelector := map[string]string{}
podOS := corev1.PodOS{}
if nodeType == kube.NodeOSWindows {
userID := "ContainerAdministrator"
securityCtx = &corev1.PodSecurityContext{
WindowsOptions: &corev1.WindowsSecurityContextOptions{
RunAsUserName: &userID,
},
}
nodeSelector[kube.NodeOSLabel] = kube.NodeOSWindows
podOS.Name = kube.NodeOSWindows
} else {
userID := int64(0)
securityCtx = &corev1.PodSecurityContext{
RunAsUser: &userID,
}
nodeSelector[kube.NodeOSLabel] = kube.NodeOSLinux
podOS.Name = kube.NodeOSLinux
}
pod := &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{
@@ -427,6 +454,8 @@ func (e *genericRestoreExposer) createRestorePod(ctx context.Context, ownerObjec
Labels: label,
},
Spec: corev1.PodSpec{
NodeSelector: nodeSelector,
OS: &podOS,
Containers: []corev1.Container{
{
Name: containerName,
@@ -450,9 +479,7 @@ func (e *genericRestoreExposer) createRestorePod(ctx context.Context, ownerObjec
Volumes: volumes,
NodeName: selectedNode,
RestartPolicy: corev1.RestartPolicyNever,
SecurityContext: &corev1.PodSecurityContext{
RunAsUser: &userID,
},
SecurityContext: securityCtx,
},
}

95
pkg/exposer/generic_restore_test.go
View File

@@ -31,7 +31,6 @@ import (
velerotest "github.com/vmware-tanzu/velero/pkg/test"
appsv1 "k8s.io/api/apps/v1"
corev1 "k8s.io/api/core/v1"
corev1api "k8s.io/api/core/v1"
clientTesting "k8s.io/client-go/testing"
)
@@ -76,9 +75,9 @@ func TestRestoreExpose(t *testing.T) {
APIVersion: appsv1.SchemeGroupVersion.String(),
},
Spec: appsv1.DaemonSetSpec{
Template: corev1.PodTemplateSpec{
Spec: corev1.PodSpec{
Containers: []corev1.Container{
Template: corev1api.PodTemplateSpec{
Spec: corev1api.PodSpec{
Containers: []corev1api.Container{
{
Image: "fake-image",
},
@@ -93,21 +92,21 @@ func TestRestoreExpose(t *testing.T) {
kubeClientObj []runtime.Object
ownerRestore *velerov1.Restore
targetPVCName string
sourceNamespace string
targetNamespace string
kubeReactors []reactor
err string
}{
{
name: "wait target pvc consumed fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
err: "error to wait target PVC consumed, fake-ns/fake-target-pvc: error to wait for PVC: error to get pvc fake-ns/fake-target-pvc: persistentvolumeclaims \"fake-target-pvc\" not found",
},
{
name: "target pvc is already bound",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObjBound,
@@ -117,7 +116,7 @@ func TestRestoreExpose(t *testing.T) {
{
name: "create restore pod fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -137,7 +136,7 @@ func TestRestoreExpose(t *testing.T) {
{
name: "create restore pvc fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -182,9 +181,9 @@ func TestRestoreExpose(t *testing.T) {
err := exposer.Expose(context.Background(), ownerObject, GenericRestoreExposeParam{
TargetPVCName: test.targetPVCName,
SourceNamespace: test.sourceNamespace,
TargetNamespace: test.targetNamespace,
HostingPodLabels: map[string]string{},
Resources: corev1.ResourceRequirements{},
Resources: corev1api.ResourceRequirements{},
ExposeTimeout: time.Millisecond})
assert.EqualError(t, err, test.err)
})
@@ -244,21 +243,21 @@ func TestRebindVolume(t *testing.T) {
kubeClientObj []runtime.Object
ownerRestore *velerov1.Restore
targetPVCName string
sourceNamespace string
targetNamespace string
kubeReactors []reactor
err string
}{
{
name: "get target pvc fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
err: "error to get target PVC fake-ns/fake-target-pvc: persistentvolumeclaims \"fake-target-pvc\" not found",
},
{
name: "wait restore pvc bound fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -268,7 +267,7 @@ func TestRebindVolume(t *testing.T) {
{
name: "retain target pv fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -289,7 +288,7 @@ func TestRebindVolume(t *testing.T) {
{
name: "delete restore pod fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -311,7 +310,7 @@ func TestRebindVolume(t *testing.T) {
{
name: "delete restore pvc fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -333,7 +332,7 @@ func TestRebindVolume(t *testing.T) {
{
name: "rebind target pvc fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -355,7 +354,7 @@ func TestRebindVolume(t *testing.T) {
{
name: "reset pv binding fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -382,7 +381,7 @@ func TestRebindVolume(t *testing.T) {
{
name: "wait restore PV bound fail",
targetPVCName: "fake-target-pvc",
sourceNamespace: "fake-ns",
targetNamespace: "fake-ns",
ownerRestore: restore,
kubeClientObj: []runtime.Object{
targetPVCObj,
@@ -420,7 +419,7 @@ func TestRebindVolume(t *testing.T) {
hookCount = 0
err := exposer.RebindVolume(context.Background(), ownerObject, test.targetPVCName, test.sourceNamespace, time.Millisecond)
err := exposer.RebindVolume(context.Background(), ownerObject, test.targetPVCName, test.targetNamespace, time.Millisecond)
assert.EqualError(t, err, test.err)
})
}
@@ -526,7 +525,7 @@ func Test_ReastoreDiagnoseExpose(t *testing.T) {
},
}
restorePodWithoutNodeName := corev1.Pod{
restorePodWithoutNodeName := corev1api.Pod{
ObjectMeta: metav1.ObjectMeta{
Namespace: velerov1.DefaultNamespace,
Name: "fake-restore",
@@ -539,19 +538,19 @@ func Test_ReastoreDiagnoseExpose(t *testing.T) {
},
},
},
Status: corev1.PodStatus{
Phase: corev1.PodPending,
Conditions: []corev1.PodCondition{
Status: corev1api.PodStatus{
Phase: corev1api.PodPending,
Conditions: []corev1api.PodCondition{
{
Type: corev1.PodInitialized,
Status: corev1.ConditionTrue,
Type: corev1api.PodInitialized,
Status: corev1api.ConditionTrue,
Message: "fake-pod-message",
},
},
},
}
restorePodWithNodeName := corev1.Pod{
restorePodWithNodeName := corev1api.Pod{
ObjectMeta: metav1.ObjectMeta{
Namespace: velerov1.DefaultNamespace,
Name: "fake-restore",
@@ -564,22 +563,22 @@ func Test_ReastoreDiagnoseExpose(t *testing.T) {
},
},
},
Spec: corev1.PodSpec{
Spec: corev1api.PodSpec{
NodeName: "fake-node",
},
Status: corev1.PodStatus{
Phase: corev1.PodPending,
Conditions: []corev1.PodCondition{
Status: corev1api.PodStatus{
Phase: corev1api.PodPending,
Conditions: []corev1api.PodCondition{
{
Type: corev1.PodInitialized,
Status: corev1.ConditionTrue,
Type: corev1api.PodInitialized,
Status: corev1api.ConditionTrue,
Message: "fake-pod-message",
},
},
},
}
restorePVCWithoutVolumeName := corev1.PersistentVolumeClaim{
restorePVCWithoutVolumeName := corev1api.PersistentVolumeClaim{
ObjectMeta: metav1.ObjectMeta{
Namespace: velerov1.DefaultNamespace,
Name: "fake-restore",
@@ -592,12 +591,12 @@ func Test_ReastoreDiagnoseExpose(t *testing.T) {
},
},
},
Status: corev1.PersistentVolumeClaimStatus{
Phase: corev1.ClaimPending,
Status: corev1api.PersistentVolumeClaimStatus{
Phase: corev1api.ClaimPending,
},
}
restorePVCWithVolumeName := corev1.PersistentVolumeClaim{
restorePVCWithVolumeName := corev1api.PersistentVolumeClaim{
ObjectMeta: metav1.ObjectMeta{
Namespace: velerov1.DefaultNamespace,
Name: "fake-restore",
@@ -610,35 +609,35 @@ func Test_ReastoreDiagnoseExpose(t *testing.T) {
},
},
},
Spec: corev1.PersistentVolumeClaimSpec{
Spec: corev1api.PersistentVolumeClaimSpec{
VolumeName: "fake-pv",
},
Status: corev1.PersistentVolumeClaimStatus{
Phase: corev1.ClaimPending,
Status: corev1api.PersistentVolumeClaimStatus{
Phase: corev1api.ClaimPending,
},
}
restorePV := corev1.PersistentVolume{
restorePV := corev1api.PersistentVolume{
ObjectMeta: metav1.ObjectMeta{
Name: "fake-pv",
},
Status: corev1.PersistentVolumeStatus{
Phase: corev1.VolumePending,
Status: corev1api.PersistentVolumeStatus{
Phase: corev1api.VolumePending,
Message: "fake-pv-message",
},
}
nodeAgentPod := corev1.Pod{
nodeAgentPod := corev1api.Pod{
ObjectMeta: metav1.ObjectMeta{
Namespace: velerov1.DefaultNamespace,
Name: "node-agent-pod-1",
Labels: map[string]string{"role": "node-agent"},
},
Spec: corev1.PodSpec{
Spec: corev1api.PodSpec{
NodeName: "fake-node",
},
Status: corev1.PodStatus{
Phase: corev1.PodRunning,
Status: corev1api.PodStatus{
Phase: corev1api.PodRunning,
},
}

1
pkg/exposer/types.go
View File

@@ -38,4 +38,5 @@ type ExposeByPod struct {
HostingPod *corev1.Pod
HostingContainer string
VolumeName string
NodeOS *string
}

1
pkg/restore/actions/csi/pvc_action.go
View File

@@ -478,6 +478,7 @@ func newDataDownload(
SnapshotID: dataUploadResult.SnapshotID,
SourceNamespace: dataUploadResult.SourceNamespace,
OperationTimeout: backup.Spec.CSISnapshotTimeout,
NodeOS: dataUploadResult.NodeOS,
},
}
if restore.Spec.UploaderConfig != nil {

1
pkg/restore/actions/dataupload_retrieve_action.go
View File

@@ -78,6 +78,7 @@ func (d *DataUploadRetrieveAction) Execute(input *velero.RestoreItemActionExecut
SnapshotID: dataUpload.Status.SnapshotID,
SourceNamespace: dataUpload.Spec.SourceNamespace,
DataMoverResult: dataUpload.Status.DataMoverResult,
NodeOS: dataUpload.Status.NodeOS,
}
jsonBytes, err := json.Marshal(dataUploadResult)