Merge branch 'main' into data-mover-ms-smoking-test

This commit is contained in:
Lyndon-Li
2024-08-14 00:15:56 +08:00
33 changed files with 1276 additions and 616 deletions
+3 -1
View File
@@ -364,8 +364,10 @@ func (o *Options) Validate(c *cobra.Command, args []string, f client.Factory) er
return err
}
if err := uploader.ValidateUploaderType(o.UploaderType); err != nil {
if msg, err := uploader.ValidateUploaderType(o.UploaderType); err != nil {
return err
} else if msg != "" {
fmt.Printf("⚠️ %s\n", msg)
}
// If we're only installing CRDs, we can skip the rest of the validation.
+14 -27
View File
@@ -293,17 +293,29 @@ func (s *nodeAgentServer) run() {
loadAffinity = s.dataPathConfigs.LoadAffinity[0]
}
dataUploadReconciler := controller.NewDataUploadReconciler(s.mgr.GetClient(), s.mgr, s.kubeClient, s.csiSnapshotClient.SnapshotV1(), s.dataPathMgr, loadAffinity, repoEnsurer, clock.RealClock{}, credentialGetter, s.nodeName, s.fileSystem, s.config.dataMoverPrepareTimeout, s.logger, s.metrics)
s.attemptDataUploadResume(dataUploadReconciler)
if err = dataUploadReconciler.SetupWithManager(s.mgr); err != nil {
s.logger.WithError(err).Fatal("Unable to create the data upload controller")
}
dataDownloadReconciler := controller.NewDataDownloadReconciler(s.mgr.GetClient(), s.mgr, s.kubeClient, s.dataPathMgr, repoEnsurer, credentialGetter, s.nodeName, s.config.dataMoverPrepareTimeout, s.logger, s.metrics)
s.attemptDataDownloadResume(dataDownloadReconciler)
if err = dataDownloadReconciler.SetupWithManager(s.mgr); err != nil {
s.logger.WithError(err).Fatal("Unable to create the data download controller")
}
go func() {
s.mgr.GetCache().WaitForCacheSync(s.ctx)
if err := dataUploadReconciler.AttemptDataUploadResume(s.ctx, s.mgr.GetClient(), s.logger.WithField("node", s.nodeName), s.namespace); err != nil {
s.logger.WithError(errors.WithStack(err)).Error("failed to attempt data upload resume")
}
if err := dataDownloadReconciler.AttemptDataDownloadResume(s.ctx, s.mgr.GetClient(), s.logger.WithField("node", s.nodeName), s.namespace); err != nil {
s.logger.WithError(errors.WithStack(err)).Error("failed to attempt data download resume")
}
s.logger.Info("Attempt complete to resume dataUploads and dataDownloads")
}()
s.logger.Info("Controllers starting...")
if err := s.mgr.Start(ctrl.SetupSignalHandler()); err != nil {
@@ -373,31 +385,6 @@ func (s *nodeAgentServer) markInProgressCRsFailed() {
s.markInProgressPVRsFailed(client)
}
func (s *nodeAgentServer) attemptDataUploadResume(r *controller.DataUploadReconciler) {
// the function is called before starting the controller manager, the embedded client isn't ready to use, so create a new one here
client, err := ctrlclient.New(s.mgr.GetConfig(), ctrlclient.Options{Scheme: s.mgr.GetScheme()})
if err != nil {
s.logger.WithError(errors.WithStack(err)).Error("failed to create client")
return
}
if err := r.AttemptDataUploadResume(s.ctx, client, s.logger.WithField("node", s.nodeName), s.namespace); err != nil {
s.logger.WithError(errors.WithStack(err)).Error("failed to attempt data upload resume")
}
}
func (s *nodeAgentServer) attemptDataDownloadResume(r *controller.DataDownloadReconciler) {
// the function is called before starting the controller manager, the embedded client isn't ready to use, so create a new one here
client, err := ctrlclient.New(s.mgr.GetConfig(), ctrlclient.Options{Scheme: s.mgr.GetScheme()})
if err != nil {
s.logger.WithError(errors.WithStack(err)).Error("failed to create client")
return
}
if err := r.AttemptDataDownloadResume(s.ctx, client, s.logger.WithField("node", s.nodeName), s.namespace); err != nil {
s.logger.WithError(errors.WithStack(err)).Error("failed to attempt data download resume")
}
}
func (s *nodeAgentServer) markInProgressPVBsFailed(client ctrlclient.Client) {
pvbs := &velerov1api.PodVolumeBackupList{}
if err := client.List(s.ctx, pvbs, &ctrlclient.ListOptions{Namespace: s.namespace}); err != nil {
+23 -6
View File
@@ -140,6 +140,7 @@ type serverConfig struct {
disableInformerCache bool
scheduleSkipImmediately bool
maintenanceCfg repository.MaintenanceConfig
backukpRepoConfig string
}
func NewCommand(f client.Factory) *cobra.Command {
@@ -253,6 +254,8 @@ func NewCommand(f client.Factory) *cobra.Command {
command.Flags().StringVar(&config.maintenanceCfg.CPULimit, "maintenance-job-cpu-limit", config.maintenanceCfg.CPULimit, "CPU limit for maintenance job. Default is no limit.")
command.Flags().StringVar(&config.maintenanceCfg.MemLimit, "maintenance-job-mem-limit", config.maintenanceCfg.MemLimit, "Memory limit for maintenance job. Default is no limit.")
command.Flags().StringVar(&config.backukpRepoConfig, "backup-repository-config", config.backukpRepoConfig, "The name of configMap containing backup repository configurations.")
// maintenance job log setting inherited from velero server
config.maintenanceCfg.FormatFlag = config.formatFlag
config.maintenanceCfg.LogLevelFlag = logLevelFlag
@@ -288,8 +291,10 @@ type server struct {
}
func newServer(f client.Factory, config serverConfig, logger *logrus.Logger) (*server, error) {
if err := uploader.ValidateUploaderType(config.uploaderType); err != nil {
if msg, err := uploader.ValidateUploaderType(config.uploaderType); err != nil {
return nil, err
} else if msg != "" {
logger.Warn(msg)
}
if config.clientQPS < 0.0 {
@@ -876,7 +881,7 @@ func (s *server) runControllers(defaultVolumeSnapshotLocations map[string]string
}
if _, ok := enabledRuntimeControllers[controller.BackupRepo]; ok {
if err := controller.NewBackupRepoReconciler(s.namespace, s.logger, s.mgr.GetClient(), s.config.repoMaintenanceFrequency, s.repoManager).SetupWithManager(s.mgr); err != nil {
if err := controller.NewBackupRepoReconciler(s.namespace, s.logger, s.mgr.GetClient(), s.config.repoMaintenanceFrequency, s.config.backukpRepoConfig, s.repoManager).SetupWithManager(s.mgr); err != nil {
s.logger.Fatal(err, "unable to create controller", "controller", controller.BackupRepo)
}
}
@@ -1148,9 +1153,15 @@ func markDataUploadsCancel(ctx context.Context, client ctrlclient.Client, backup
du.Status.Phase == velerov2alpha1api.DataUploadPhaseNew ||
du.Status.Phase == "" {
err := controller.UpdateDataUploadWithRetry(ctx, client, types.NamespacedName{Namespace: du.Namespace, Name: du.Name}, log.WithField("dataupload", du.Name),
func(dataUpload *velerov2alpha1api.DataUpload) {
func(dataUpload *velerov2alpha1api.DataUpload) bool {
if dataUpload.Spec.Cancel {
return false
}
dataUpload.Spec.Cancel = true
dataUpload.Status.Message = fmt.Sprintf("found a dataupload with status %q during the velero server starting, mark it as cancel", du.Status.Phase)
dataUpload.Status.Message = fmt.Sprintf("Dataupload is in status %q during the velero server starting, mark it as cancel", du.Status.Phase)
return true
})
if err != nil {
@@ -1183,9 +1194,15 @@ func markDataDownloadsCancel(ctx context.Context, client ctrlclient.Client, rest
dd.Status.Phase == velerov2alpha1api.DataDownloadPhaseNew ||
dd.Status.Phase == "" {
err := controller.UpdateDataDownloadWithRetry(ctx, client, types.NamespacedName{Namespace: dd.Namespace, Name: dd.Name}, log.WithField("datadownload", dd.Name),
func(dataDownload *velerov2alpha1api.DataDownload) {
func(dataDownload *velerov2alpha1api.DataDownload) bool {
if dataDownload.Spec.Cancel {
return false
}
dataDownload.Spec.Cancel = true
dataDownload.Status.Message = fmt.Sprintf("found a datadownload with status %q during the velero server starting, mark it as cancel", dd.Status.Phase)
dataDownload.Status.Message = fmt.Sprintf("Datadownload is in status %q during the velero server starting, mark it as cancel", dd.Status.Phase)
return true
})
if err != nil {
+7
View File
@@ -203,6 +203,13 @@ func Test_newServer(t *testing.T) {
}, logger)
assert.Error(t, err)
// invalid clientQPS Restic uploader
_, err = newServer(factory, serverConfig{
uploaderType: uploader.ResticType,
clientQPS: -1,
}, logger)
assert.Error(t, err)
// invalid clientBurst
factory.On("SetClientQPS", mock.Anything).Return()
_, err = newServer(factory, serverConfig{