feat: Implement early frequent polling for CSI snapshots

Co-authored-by: aider (gemini/gemini-2.5-pro) <aider@aider.chat>
Signed-off-by: Scott Seago <sseago@redhat.com>

pkg/cmd/cli/install/install.go

@@ -81,6 +81,7 @@ type Options struct {
 	DefaultVolumesToFsBackup        bool
 	UploaderType                    string
 	DefaultSnapshotMoveData         bool
+	CSISnapshotEarlyFrequentPolling bool
 	DisableInformerCache            bool
 	ScheduleSkipImmediately         bool
 	PodResources                    kubeutil.PodResources
@@ -141,6 +142,7 @@ func (o *Options) BindFlags(flags *pflag.FlagSet) {
 	flags.BoolVar(&o.DefaultVolumesToFsBackup, "default-volumes-to-fs-backup", o.DefaultVolumesToFsBackup, "Bool flag to configure Velero server to use pod volume file system backup by default for all volumes on all backups. Optional.")
 	flags.StringVar(&o.UploaderType, "uploader-type", o.UploaderType, fmt.Sprintf("The type of uploader to transfer the data of pod volumes, supported value: '%s'", uploader.KopiaType))
 	flags.BoolVar(&o.DefaultSnapshotMoveData, "default-snapshot-move-data", o.DefaultSnapshotMoveData, "Bool flag to configure Velero server to move data by default for all snapshots supporting data movement. Optional.")
+	flags.BoolVar(&o.CSISnapshotEarlyFrequentPolling, "csi-snapshot-early-frequent-polling", o.CSISnapshotEarlyFrequentPolling, "Bool flag to configure Velero server to use early frequent polling by default for all CSI snapshots. Optional.")
 	flags.BoolVar(&o.DisableInformerCache, "disable-informer-cache", o.DisableInformerCache, "Disable informer cache for Get calls on restore. With this enabled, it will speed up restore in cases where there are backup resources which already exist in the cluster, but for very large clusters this will increase velero memory usage. Default is false (don't disable). Optional.")
 	flags.BoolVar(&o.ScheduleSkipImmediately, "schedule-skip-immediately", o.ScheduleSkipImmediately, "Skip the first scheduled backup immediately after creating a schedule. Default is false (don't skip).")
 	flags.BoolVar(&o.NodeAgentDisableHostPath, "node-agent-disable-host-path", o.NodeAgentDisableHostPath, "Don't mount the pod volume host path to node-agent. Optional. Pod volume host path mount is required by fs-backup but could be disabled for other backup methods.")
@@ -238,16 +240,17 @@ func NewInstallOptions() *Options {
 		NodeAgentPodCPULimit:      install.DefaultNodeAgentPodCPULimit,
 		NodeAgentPodMemLimit:      install.DefaultNodeAgentPodMemLimit,
 		// Default to creating a VSL unless we're told otherwise
-		UseVolumeSnapshots:       true,
-		NoDefaultBackupLocation:  false,
-		CRDsOnly:                 false,
-		DefaultVolumesToFsBackup: false,
-		UploaderType:             uploader.KopiaType,
-		DefaultSnapshotMoveData:  false,
-		DisableInformerCache:     false,
-		ScheduleSkipImmediately:  false,
-		kubeletRootDir:           install.DefaultKubeletRootDir,
-		NodeAgentDisableHostPath: false,
+		UseVolumeSnapshots:              true,
+		NoDefaultBackupLocation:         false,
+		CRDsOnly:                        false,
+		DefaultVolumesToFsBackup:        false,
+		UploaderType:                    uploader.KopiaType,
+		DefaultSnapshotMoveData:         false,
+		CSISnapshotEarlyFrequentPolling: false,
+		DisableInformerCache:            false,
+		ScheduleSkipImmediately:         false,
+		kubeletRootDir:                  install.DefaultKubeletRootDir,
+		NodeAgentDisableHostPath:        false,
 	}
 }
 
@@ -324,6 +327,7 @@ func (o *Options) AsVeleroOptions() (*install.VeleroOptions, error) {
 		DefaultVolumesToFsBackup:        o.DefaultVolumesToFsBackup,
 		UploaderType:                    o.UploaderType,
 		DefaultSnapshotMoveData:         o.DefaultSnapshotMoveData,
+		CSISnapshotEarlyFrequentPolling: o.CSISnapshotEarlyFrequentPolling,
 		DisableInformerCache:            o.DisableInformerCache,
 		ScheduleSkipImmediately:         o.ScheduleSkipImmediately,
 		PodResources:                    o.PodResources,

pkg/install/deployment.go

@@ -50,6 +50,7 @@ type podTemplateConfig struct {
 	serviceAccountName              string
 	uploaderType                    string
 	defaultSnapshotMoveData         bool
+	csiSnapshotEarlyFrequentPolling bool
 	privilegedNodeAgent             bool
 	disableInformerCache            bool
 	scheduleSkipImmediately         bool
@@ -166,6 +167,12 @@ func WithDefaultSnapshotMoveData(b bool) podTemplateOption {
 	}
 }
 
+func WithCSISnapshotEarlyFrequentPolling(b bool) podTemplateOption {
+	return func(c *podTemplateConfig) {
+		c.csiSnapshotEarlyFrequentPolling = b
+	}
+}
+
 func WithDisableInformerCache(b bool) podTemplateOption {
 	return func(c *podTemplateConfig) {
 		c.disableInformerCache = b
@@ -489,6 +496,15 @@ func Deployment(namespace string, opts ...podTemplateOption) *appsv1api.Deployme
 		}...)
 	}
 
+	if c.csiSnapshotEarlyFrequentPolling {
+		deployment.Spec.Template.Spec.Containers[0].Env = append(deployment.Spec.Template.Spec.Containers[0].Env, []corev1api.EnvVar{
+			{
+				Name:  "CSI_SNAPSHOT_EARLY_FREQUENT_POLLING",
+				Value: "true",
+			},
+		}...)
+	}
+
 	deployment.Spec.Template.Spec.Containers[0].Env = append(deployment.Spec.Template.Spec.Containers[0].Env, c.envVars...)
 
 	if len(c.plugins) > 0 {

pkg/install/resources.go

@@ -263,6 +263,7 @@ type VeleroOptions struct {
 	DefaultVolumesToFsBackup        bool
 	UploaderType                    string
 	DefaultSnapshotMoveData         bool
+	CSISnapshotEarlyFrequentPolling bool
 	DisableInformerCache            bool
 	ScheduleSkipImmediately         bool
 	PodResources                    kube.PodResources
@@ -390,6 +391,10 @@ func AllResources(o *VeleroOptions) *unstructured.UnstructuredList {
 		deployOpts = append(deployOpts, WithDefaultSnapshotMoveData(true))
 	}
 
+	if o.CSISnapshotEarlyFrequentPolling {
+		deployOpts = append(deployOpts, WithCSISnapshotEarlyFrequentPolling(true))
+	}
+
 	if o.DisableInformerCache {
 		deployOpts = append(deployOpts, WithDisableInformerCache(true))
 	}

pkg/util/csi/volume_snapshot.go

@@ -20,6 +20,8 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"os"
+	"strconv"
 	"strings"
 	"time"
 
@@ -660,25 +662,30 @@ func WaitUntilVSCHandleIsReady(
 		return true, nil
 	}
 
-	// The short interval for the first ten seconds is due to the fact that
-	// Microsoft VSS backups have a hard-coded unfreeze call after 10 seconds,
-	// so we need to minimize waiting time during the first 10 seconds.
-	// First poll with a short interval and timeout.
-	interval = 1 * time.Second
-	timeout := 10 * time.Second
-	err := wait.PollUntilContextTimeout(
-		context.Background(),
-		interval,
-		timeout,
-		true,
-		pollFunc,
-	)
+	var err error
+	frequentPolling, err := strconv.ParseBool(os.Getenv("CSI_SNAPSHOT_EARLY_FREQUENT_POLLING"))
 
-	if err == nil {
-		return vsc, nil
-	}
-	if !wait.Interrupted(err) {
-		return nil, err
+	if err == nil && frequentPolling {
+		// The short interval for the first ten seconds is due to the fact that
+		// Microsoft VSS backups have a hard-coded unfreeze call after 10 seconds,
+		// so we need to minimize waiting time during the first 10 seconds.
+		// First poll with a short interval and timeout.
+		interval = 1 * time.Second
+		timeout := 10 * time.Second
+		err = wait.PollUntilContextTimeout(
+			context.Background(),
+			interval,
+			timeout,
+			true,
+			pollFunc,
+		)
+
+		if err == nil {
+			return vsc, nil
+		}
+		if !wait.Interrupted(err) {
+			return nil, err
+		}
 	}
 
 	// If the first poll timed out, poll with a longer interval and the full timeout.