fix(snapshot): add checkpoint and restore timing summaries (#7949)

Signed-off-by: Schwinn Saereesitthipitak <schwinns@nvidia.com>

deploy/snapshot/cmd/nsrestore/main.go

@@ -31,14 +31,10 @@ func main() {
 		CgroupRoot:     *cgroupRoot,
 	}
 
-	restoredPID, err := executor.RestoreInNamespace(context.Background(), opts, log)
+	result, err := executor.RestoreInNamespace(context.Background(), opts, log)
 	if err != nil {
 		fatal(log, err, "restore failed")
 	}
-
-	result := struct {
-		RestoredPID int `json:"restoredPID"`
-	}{RestoredPID: restoredPID}
 	if err := json.NewEncoder(os.Stdout).Encode(result); err != nil {
 		fatal(log, err, "Failed to write restore result")
 	}

deploy/snapshot/internal/controller/controller.go

@@ -225,11 +225,12 @@ func (w *NodeController) reconcileCheckpointPod(ctx context.Context, pod *corev1
 		return
 	}
 
-	w.log.Info("Pod ready, triggering checkpoint", "pod", podKey, "checkpoint_id", checkpointID)
+	startedAt := time.Now()
+	w.log.Info("Checkpoint target detected, triggering checkpoint", "pod", podKey, "checkpoint_id", checkpointID)
 	emitPodEvent(ctx, w.clientset, w.log, pod, "snapshot", corev1.EventTypeNormal, "CheckpointRequested", fmt.Sprintf("Checkpoint requested: %s", checkpointID))
 
 	go func() {
-		if err := w.runCheckpoint(ctx, pod, job, checkpointID, checkpointLocation, podKey); err != nil {
+		if err := w.runCheckpoint(ctx, pod, job, checkpointID, checkpointLocation, podKey, startedAt); err != nil {
 			opLog := w.log.WithValues("pod", podKey, "checkpoint_id", checkpointID)
 			opLog.Error(err, "Checkpoint controller worker failed")
 			emitPodEvent(ctx, w.clientset, opLog, pod, "snapshot", corev1.EventTypeWarning, "CheckpointWorkerFailed", err.Error())
@@ -299,11 +300,12 @@ func (w *NodeController) reconcileRestorePod(ctx context.Context, pod *corev1.Po
 		return
 	}
 
-	w.log.Info("Restore pod running, triggering external restore", "pod", podKey, "checkpoint_id", checkpointID)
+	startedAt := time.Now()
+	w.log.Info("Restore target detected, triggering external restore", "pod", podKey, "checkpoint_id", checkpointID)
 	emitPodEvent(ctx, w.clientset, w.log, pod, "snapshot", corev1.EventTypeNormal, "RestoreRequested", fmt.Sprintf("Restore requested from checkpoint %s", checkpointID))
 
 	go func() {
-		if err := w.runRestore(ctx, pod, containerName, containerID, checkpointID, checkpointLocation, restoreAttemptKey); err != nil {
+		if err := w.runRestore(ctx, pod, containerName, containerID, checkpointID, checkpointLocation, restoreAttemptKey, startedAt); err != nil {
 			opLog := w.log.WithValues("pod", podKey, "checkpoint_id", checkpointID)
 			opLog.Error(err, "Restore controller worker failed")
 			emitPodEvent(ctx, w.clientset, opLog, pod, "snapshot", corev1.EventTypeWarning, "RestoreWorkerFailed", err.Error())
@@ -317,7 +319,7 @@ func (w *NodeController) reconcileRestorePod(ctx context.Context, pod *corev1.Po
 //  3. Call executor.Checkpoint (inspect → configure → CUDA lock/checkpoint → CRIU dump → rootfs diff)
 //  4. SIGUSR1 the process on success (notify workload), SIGKILL on failure (terminate immediately)
 //  5. Mark job as completed or failed
-func (w *NodeController) runCheckpoint(ctx context.Context, pod *corev1.Pod, job *batchv1.Job, checkpointID, checkpointLocation, podKey string) error {
+func (w *NodeController) runCheckpoint(ctx context.Context, pod *corev1.Pod, job *batchv1.Job, checkpointID, checkpointLocation, podKey string, startedAt time.Time) error {
 	releasePodOnExit := true
 	defer func() {
 		if releasePodOnExit {
@@ -396,6 +398,7 @@ func (w *NodeController) runCheckpoint(ctx context.Context, pod *corev1.Pod, job
 		ContainerName:      containerName,
 		CheckpointID:       checkpointID,
 		CheckpointLocation: checkpointLocation,
+		StartedAt:          startedAt,
 		NodeName:           w.config.NodeName,
 		PodName:            pod.Name,
 		PodNamespace:       pod.Namespace,
@@ -458,7 +461,7 @@ func (w *NodeController) runCheckpoint(ctx context.Context, pod *corev1.Pod, job
 //  3. SIGCONT the restored process to wake it up
 //  4. Wait for the pod to become Ready
 //  5. Mark the container instance as completed
-func (w *NodeController) runRestore(ctx context.Context, pod *corev1.Pod, containerName, containerID, checkpointID, checkpointLocation, restoreAttemptKey string) error {
+func (w *NodeController) runRestore(ctx context.Context, pod *corev1.Pod, containerName, containerID, checkpointID, checkpointLocation, restoreAttemptKey string, startedAt time.Time) error {
 	releaseOnExit := true
 	defer func() {
 		if releaseOnExit {
@@ -493,6 +496,7 @@ func (w *NodeController) runRestore(ctx context.Context, pod *corev1.Pod, contai
 	req := executor.RestoreRequest{
 		CheckpointID:       checkpointID,
 		CheckpointLocation: checkpointLocation,
+		StartedAt:          startedAt,
 		NSRestorePath:      w.config.Restore.NSRestorePath,
 		PodName:            pod.Name,
 		PodNamespace:       pod.Namespace,

deploy/snapshot/internal/controller/controller_test.go

@@ -476,7 +476,7 @@ func TestRunCheckpointKeepsLeaseAndInFlightOnTerminalStatusPatchFailure(t *testi
 		stopCh: make(chan struct{}),
 	}
 
-	err := w.runCheckpoint(context.Background(), pod, job, "abc123", filepath.Join(t.TempDir(), "abc123"), "default/test-pod")
+	err := w.runCheckpoint(context.Background(), pod, job, "abc123", filepath.Join(t.TempDir(), "abc123"), "default/test-pod", time.Now())
 	if err == nil {
 		t.Fatal("expected terminal checkpoint status update to fail")
 	}

deploy/snapshot/internal/criu/restore.go

@@ -71,7 +71,7 @@ func ExecuteRestore(
 	defer closeFiles(inheritedFiles)
 
 	notify := &restoreNotify{log: log}
-	log.Info("Executing go-criu Restore call")
+	log.V(1).Info("Executing go-criu Restore call")
 	if err := c.Restore(criuOpts, notify); err != nil {
 		log.Error(err, "go-criu Restore returned error")
 		logging.LogRestoreErrors(checkpointPath, settings.WorkDir, log)
@@ -88,7 +88,7 @@ func BuildRestoreOpts(m *types.CheckpointManifest, checkpointPath string, cgroup
 	if err != nil {
 		return nil, err
 	}
-	log.Info("Generated external mount map set", "ext_mount_count", len(extMounts))
+	log.V(1).Info("Generated external mount map set", "ext_mount_count", len(extMounts))
 
 	settings := m.CRIUDump.CRIU
 	criuOpts := &criurpc.CriuOpts{
@@ -137,7 +137,7 @@ func buildRestoreExtMounts(m *types.CheckpointManifest) ([]*criurpc.ExtMountMap,
 
 func registerInheritFDs(c *criulib.Criu, stdioFDs []string, log logr.Logger) []*os.File {
 	if len(stdioFDs) == 0 {
-		log.Info("No stdio FD descriptors in manifest, skipping inherit-fd setup")
+		log.V(1).Info("No stdio FD descriptors in manifest, skipping inherit-fd setup")
 		return nil
 	}
 
@@ -161,7 +161,7 @@ func registerInheritFDs(c *criulib.Criu, stdioFDs []string, log logr.Logger) []*
 		c.AddInheritFd(target, f)
 	}
 
-	log.Info("Registered inherited stdio pipes", "count", len(openFiles))
+	log.V(1).Info("Registered inherited stdio pipes", "count", len(openFiles))
 	return openFiles
 }
 

deploy/snapshot/internal/cuda/cuda.go

@@ -8,6 +8,7 @@ import (
 	"regexp"
 	"strconv"
 	"strings"
+	"time"
 
 	"github.com/go-logr/logr"
 	"google.golang.org/grpc"
@@ -25,6 +26,14 @@ var podResourcesSocketPath = "/var/lib/kubelet/pod-resources/kubelet.sock"
 
 var gpuUUIDPattern = regexp.MustCompile(`^GPU-[a-fA-F0-9]{8}-[a-fA-F0-9]{4}-[a-fA-F0-9]{4}-[a-fA-F0-9]{4}-[a-fA-F0-9]{12}$`)
 
+type CheckpointPhaseTimings struct {
+	TotalDuration time.Duration
+}
+
+type RestorePhaseTimings struct {
+	TotalDuration time.Duration
+}
+
 // GetPodGPUUUIDs resolves GPU UUIDs for a pod/container from kubelet
 // PodResources (nvidia.com/gpu entries in GetDevices()).
 func GetPodGPUUUIDs(ctx context.Context, podName, podNamespace, containerName string) ([]string, error) {
@@ -174,38 +183,52 @@ func BuildDeviceMap(sourceUUIDs, targetUUIDs []string, log logr.Logger) (string,
 
 // LockAndCheckpointProcessTree locks and checkpoints CUDA state for all given PIDs.
 // On failure, the caller is expected to fail the operation and terminate the workload.
-func LockAndCheckpointProcessTree(ctx context.Context, cudaPIDs []int, log logr.Logger) error {
+func LockAndCheckpointProcessTree(ctx context.Context, cudaPIDs []int, log logr.Logger) (CheckpointPhaseTimings, error) {
+	var timings CheckpointPhaseTimings
+
+	start := time.Now()
 	for _, pid := range cudaPIDs {
 		if err := lock(ctx, pid, log); err != nil {
-			return err
+			timings.TotalDuration = time.Since(start)
+			return timings, err
 		}
 	}
 
 	for _, pid := range cudaPIDs {
 		if err := checkpoint(ctx, pid, log); err != nil {
-			return err
+			timings.TotalDuration = time.Since(start)
+			return timings, err
 		}
 	}
+	timings.TotalDuration = time.Since(start)
 
-	return nil
+	return timings, nil
 }
 
 // RestoreAndUnlockProcessTree restores and unlocks CUDA state for the given PIDs.
-func RestoreAndUnlockProcessTree(ctx context.Context, cudaPIDs []int, deviceMap string, log logr.Logger) error {
+func RestoreAndUnlockProcessTree(ctx context.Context, cudaPIDs []int, deviceMap string, log logr.Logger) (RestorePhaseTimings, error) {
+	var timings RestorePhaseTimings
+
+	start := time.Now()
 	for _, pid := range cudaPIDs {
 		if err := restoreProcess(ctx, pid, deviceMap, log); err != nil {
-			return err
+			timings.TotalDuration = time.Since(start)
+			return timings, err
 		}
 	}
+
 	for _, pid := range cudaPIDs {
 		if err := unlock(ctx, pid, log); err != nil {
+			timings.TotalDuration = time.Since(start)
 			state, stateErr := getState(ctx, pid)
 			if stateErr == nil && state == "running" {
 				log.Info("cuda-checkpoint-helper unlock returned error but process is already running", "pid", pid)
 				continue
 			}
-			return err
+			return timings, err
 		}
 	}
-	return nil
+	timings.TotalDuration = time.Since(start)
+
+	return timings, nil
 }

deploy/snapshot/internal/cuda/shim.go

@@ -82,7 +82,7 @@ func runAction(ctx context.Context, pid int, action, deviceMap string, log logr.
 		)
 		return fmt.Errorf("cuda-checkpoint-helper %v failed for pid %d after %s: %w (output: %s)", args, pid, duration, err, out)
 	}
-	log.Info("cuda-checkpoint-helper command succeeded",
+	log.V(1).Info("cuda-checkpoint-helper command succeeded",
 		"pid", pid,
 		"outermost_pid", details.OutermostPID,
 		"innermost_pid", details.InnermostPID,

deploy/snapshot/internal/executor/checkpoint.go

@@ -28,12 +28,21 @@ type CheckpointRequest struct {
 	ContainerName      string
 	CheckpointID       string
 	CheckpointLocation string
+	StartedAt          time.Time
 	NodeName           string
 	PodName            string
 	PodNamespace       string
 	Clientset          kubernetes.Interface
 }
 
+type checkpointPhaseTimings struct {
+	PrepareDuration        time.Duration
+	CUDADuration           time.Duration
+	CRIUDumpDuration       time.Duration
+	OverlayCaptureDuration time.Duration
+	FinalizeDuration       time.Duration
+}
+
 // Checkpoint performs a CRIU dump of a container.
 // The operation has three phases: inspect, configure, capture.
 //
@@ -42,6 +51,8 @@ type CheckpointRequest struct {
 // renamed into place at the base path root.
 func Checkpoint(ctx context.Context, ctrd *containerd.Client, log logr.Logger, req CheckpointRequest, cfg *types.AgentConfig) error {
 	checkpointStart := time.Now()
+	phaseTimings := checkpointPhaseTimings{}
+	prepareStart := time.Now()
 	log.Info("=== Starting checkpoint operation ===")
 
 	if strings.TrimSpace(req.CheckpointID) == "" {
@@ -73,26 +84,46 @@ func Checkpoint(ctx context.Context, ctrd *containerd.Client, log logr.Logger, r
 	if err != nil {
 		return err
 	}
+	phaseTimings.PrepareDuration = time.Since(prepareStart)
 
 	// Phase 3: Capture — CRIU dump, rootfs diff
-	criuDumpDuration, err := captureCheckpoint(ctx, criuOpts, &cfg.CRIU, data, state, tmpDir, log)
+	captureTimings, err := captureCheckpoint(ctx, criuOpts, &cfg.CRIU, data, state, tmpDir, log)
 	if err != nil {
 		return err
 	}
+	phaseTimings.CUDADuration = captureTimings.CUDADuration
+	phaseTimings.CRIUDumpDuration = captureTimings.CRIUDumpDuration
+	phaseTimings.OverlayCaptureDuration = captureTimings.OverlayCaptureDuration
 
-	// Remove any previous checkpoint with the same identity hash before finalizing
+	// Remove any previous checkpoint with the same identity hash, then
+	// promote the staged checkpoint directory into place.
+	finalizeStart := time.Now()
 	if err := os.RemoveAll(finalDir); err != nil {
 		return fmt.Errorf("failed to remove previous checkpoint directory: %w", err)
 	}
 	if err := os.Rename(tmpDir, finalDir); err != nil {
 		return fmt.Errorf("failed to finalize checkpoint directory: %w", err)
 	}
+	phaseTimings.FinalizeDuration = time.Since(finalizeStart)
 
 	totalDuration := time.Since(checkpointStart)
-	log.Info("=== Checkpoint operation completed ===",
-		"total_duration", totalDuration,
-		"criu_dump_duration", criuDumpDuration,
+	log.Info("Checkpoint timing summary",
+		"checkpoint", map[string]any{
+			"duration": totalDuration.String(),
+			"phases": map[string]string{
+				"prepare_duration":         phaseTimings.PrepareDuration.String(),
+				"cuda_duration":            phaseTimings.CUDADuration.String(),
+				"criu_dump_duration":       phaseTimings.CRIUDumpDuration.String(),
+				"overlay_capture_duration": phaseTimings.OverlayCaptureDuration.String(),
+				"finalize_duration":        phaseTimings.FinalizeDuration.String(),
+			},
+		},
 	)
+	if !req.StartedAt.IsZero() {
+		log.Info("Checkpoint wall time from agent detection",
+			"started_to_checkpoint_complete", time.Since(req.StartedAt),
+		)
+	}
 
 	return nil
 }
@@ -156,7 +187,7 @@ func inspectContainer(ctx context.Context, ctrd *containerd.Client, log logr.Log
 		cudaNamespacePIDs = append(cudaNamespacePIDs, process.InnermostPID)
 	}
 	if len(cudaHostPIDs) > 0 {
-		log.Info("Resolved checkpoint CUDA PID mapping", "host_pids", cudaHostPIDs, "namespace_pids", cudaNamespacePIDs)
+		log.V(1).Info("Resolved checkpoint CUDA PID mapping", "host_pids", cudaHostPIDs, "namespace_pids", cudaNamespacePIDs)
 	}
 	var gpuUUIDs []string
 	if len(cudaHostPIDs) > 0 {
@@ -218,30 +249,37 @@ func configureCheckpoint(
 	return criuOpts, m, nil
 }
 
-func captureCheckpoint(ctx context.Context, criuOpts *criurpc.CriuOpts, criuSettings *types.CRIUSettings, data *types.CheckpointManifest, state *types.CheckpointContainerSnapshot, checkpointDir string, log logr.Logger) (time.Duration, error) {
+func captureCheckpoint(ctx context.Context, criuOpts *criurpc.CriuOpts, criuSettings *types.CRIUSettings, data *types.CheckpointManifest, state *types.CheckpointContainerSnapshot, checkpointDir string, log logr.Logger) (*checkpointPhaseTimings, error) {
+	timings := &checkpointPhaseTimings{}
+
 	// CUDA lock+checkpoint must happen before CRIU dump
 	if len(state.CUDAHostPIDs) > 0 {
-		if err := cuda.LockAndCheckpointProcessTree(ctx, state.CUDAHostPIDs, log); err != nil {
-			return 0, fmt.Errorf("CUDA checkpoint failed: %w", err)
+		cudaTimings, err := cuda.LockAndCheckpointProcessTree(ctx, state.CUDAHostPIDs, log)
+		if err != nil {
+			return nil, fmt.Errorf("CUDA checkpoint failed: %w", err)
 		}
+		timings.CUDADuration = cudaTimings.TotalDuration
 	}
 
 	criuDumpDuration, err := criu.ExecuteDump(criuOpts, checkpointDir, criuSettings, log)
 	if err != nil {
-		return 0, err
+		return nil, err
 	}
+	timings.CRIUDumpDuration = criuDumpDuration
 
 	// Overlay rootfs diff capture is best-effort. Failures are logged but not
 	// propagated — a checkpoint without overlay diffs is still valid for restore
 	// (the base container image provides the filesystem).
 	if state.UpperDir != "" {
+		overlayCaptureStart := time.Now()
 		if _, err := snapshotruntime.CaptureRootfsDiff(state.UpperDir, checkpointDir, data.Overlay.Exclusions, data.Overlay.BindMountDests); err != nil {
 			log.Error(err, "Failed to capture rootfs diff")
 		}
 		if _, err := snapshotruntime.CaptureDeletedFiles(state.UpperDir, checkpointDir); err != nil {
 			log.Error(err, "Failed to capture deleted files")
 		}
+		timings.OverlayCaptureDuration = time.Since(overlayCaptureStart)
 	}
 
-	return criuDumpDuration, nil
+	return timings, nil
 }

deploy/snapshot/internal/executor/nsrestore.go

@@ -22,8 +22,15 @@ type RestoreOptions struct {
 	CgroupRoot     string
 }
 
+type RestoreInNamespaceResult struct {
+	RestoredPID            int           `json:"restoredPID"`
+	NSRestoreSetupDuration time.Duration `json:"nsrestoreSetupDuration"`
+	CRIURestoreDuration    time.Duration `json:"criuRestoreDuration"`
+	CUDADuration           time.Duration `json:"cudaDuration"`
+}
+
 // RestoreInNamespace performs a full restore from inside the target container's namespaces.
-func RestoreInNamespace(ctx context.Context, opts RestoreOptions, log logr.Logger) (int, error) {
+func RestoreInNamespace(ctx context.Context, opts RestoreOptions, log logr.Logger) (*RestoreInNamespaceResult, error) {
 	restoreStart := time.Now()
 	log.Info("Starting nsrestore workflow",
 		"checkpoint_path", opts.CheckpointPath,
@@ -31,49 +38,77 @@ func RestoreInNamespace(ctx context.Context, opts RestoreOptions, log logr.Logge
 		"cgroup_root", opts.CgroupRoot,
 	)
 
+	manifestReadStart := time.Now()
 	m, err := types.ReadManifest(opts.CheckpointPath)
 	if err != nil {
-		return 0, fmt.Errorf("failed to read manifest: %w", err)
+		return nil, fmt.Errorf("failed to read manifest: %w", err)
 	}
-	log.Info("Loaded checkpoint manifest",
+	manifestReadDuration := time.Since(manifestReadStart)
+	log.V(1).Info("Loaded checkpoint manifest",
 		"ext_mounts", len(m.CRIUDump.ExtMnt),
 		"criu_log_level", m.CRIUDump.CRIU.LogLevel,
 		"manage_cgroups_mode", m.CRIUDump.CRIU.ManageCgroupsMode,
 		"checkpoint_has_cuda", !m.CUDA.IsEmpty(),
 	)
+
 	// Phase 1: Configure — build CRIU opts from manifest
+	configureStart := time.Now()
 	criuOpts, err := criu.BuildRestoreOpts(m, opts.CheckpointPath, opts.CgroupRoot, log)
 	if err != nil {
-		return 0, err
+		return nil, err
 	}
+	configureDuration := time.Since(configureStart)
 
 	// Phase 2: Execute — rootfs, CRIU restore, CUDA restore
-	restoredPID, err := executeRestore(ctx, criuOpts, m, opts, log)
+	executeTimings, restoredPID, err := executeRestore(ctx, criuOpts, m, opts, log)
 	if err != nil {
-		return 0, err
+		return nil, err
 	}
 
-	log.Info("nsrestore completed", "restored_pid", restoredPID, "duration", time.Since(restoreStart))
-	return restoredPID, nil
+	result := &RestoreInNamespaceResult{
+		RestoredPID:            restoredPID,
+		NSRestoreSetupDuration: manifestReadDuration + configureDuration + executeTimings.nsrestoreSetupDuration,
+		CRIURestoreDuration:    executeTimings.criuRestoreDuration,
+		CUDADuration:           executeTimings.cudaDuration,
+	}
+	log.V(1).Info("nsrestore timing summary",
+		"restored_pid", restoredPID,
+		"nsrestore_setup_duration", result.NSRestoreSetupDuration,
+		"criu_restore_duration", result.CRIURestoreDuration,
+		"cuda_duration", result.CUDADuration,
+		"total_duration", time.Since(restoreStart),
+	)
+	return result, nil
 }
 
-func executeRestore(ctx context.Context, criuOpts *criurpc.CriuOpts, m *types.CheckpointManifest, opts RestoreOptions, log logr.Logger) (int, error) {
+type nsrestorePhaseTimings struct {
+	nsrestoreSetupDuration time.Duration
+	criuRestoreDuration    time.Duration
+	cudaDuration           time.Duration
+}
+
+func executeRestore(ctx context.Context, criuOpts *criurpc.CriuOpts, m *types.CheckpointManifest, opts RestoreOptions, log logr.Logger) (*nsrestorePhaseTimings, int, error) {
+	timings := &nsrestorePhaseTimings{}
+
 	// Apply rootfs diff inside the namespace (target root is /)
+	nsrestoreSetupStart := time.Now()
 	if err := snapshotruntime.ApplyRootfsDiff(opts.CheckpointPath, "/", log); err != nil {
-		return 0, fmt.Errorf("rootfs diff failed: %w", err)
+		return nil, 0, fmt.Errorf("rootfs diff failed: %w", err)
 	}
+
 	if err := snapshotruntime.ApplyDeletedFiles(opts.CheckpointPath, "/", log); err != nil {
 		log.Error(err, "Failed to apply deleted files")
 	}
 
 	// Unmount placeholder's /dev/shm so CRIU can recreate tmpfs with checkpointed content
 	if err := syscall.Unmount("/dev/shm", 0); err != nil {
-		return 0, fmt.Errorf("failed to unmount /dev/shm before restore: %w", err)
+		return nil, 0, fmt.Errorf("failed to unmount /dev/shm before restore: %w", err)
 	}
 
 	if err := snapshotruntime.RemountProcSys(true); err != nil {
-		return 0, fmt.Errorf("failed to remount /proc/sys read-write for restore: %w", err)
+		return nil, 0, fmt.Errorf("failed to remount /proc/sys read-write for restore: %w", err)
 	}
+	timings.nsrestoreSetupDuration = time.Since(nsrestoreSetupStart)
 	defer func() {
 		if err := snapshotruntime.RemountProcSys(false); err != nil {
 			log.Error(err, "Failed to remount /proc/sys read-only after restore")
@@ -81,22 +116,26 @@ func executeRestore(ctx context.Context, criuOpts *criurpc.CriuOpts, m *types.Ch
 	}()
 
 	// CRIU restore
+	criuRestoreStart := time.Now()
 	restoredPID, err := criu.ExecuteRestore(criuOpts, m, opts.CheckpointPath, log)
 	if err != nil {
-		return 0, err
+		return nil, 0, err
 	}
+	timings.criuRestoreDuration = time.Since(criuRestoreStart)
+
+	cudaStart := time.Now()
 	processes, err := snapshotruntime.ReadProcessTable("/proc")
 	if err != nil {
-		return 0, fmt.Errorf("failed to read restored process table: %w", err)
+		return nil, 0, fmt.Errorf("failed to read restored process table: %w", err)
 	}
-	log.Info("Restored process table snapshot",
+	log.V(1).Info("Restored process table snapshot",
 		"proc_root", "/proc",
 		"criu_callback_pid", restoredPID,
 		"process_count", len(processes),
 		"manifest_cuda_pids", m.CUDA.PIDs,
 	)
 	for _, process := range processes {
-		log.Info("Restored process entry",
+		log.V(1).Info("Restored process entry",
 			"observed_pid", process.ObservedPID,
 			"parent_pid", process.ParentPID,
 			"outermost_pid", process.OutermostPID,
@@ -111,17 +150,19 @@ func executeRestore(ctx context.Context, criuOpts *criurpc.CriuOpts, m *types.Ch
 	if !m.CUDA.IsEmpty() {
 		restorePIDs, err := snapshotruntime.ResolveManifestPIDsToObservedPIDs(processes, int(restoredPID), m.CUDA.PIDs)
 		if err != nil {
-			return 0, fmt.Errorf("failed to resolve restored CUDA PIDs: %w", err)
+			return nil, 0, fmt.Errorf("failed to resolve restored CUDA PIDs: %w", err)
 		}
-		log.Info("Resolved manifest CUDA PIDs to current restore PIDs",
+		log.V(1).Info("Resolved manifest CUDA PIDs to current restore PIDs",
 			"manifest_cuda_pids", m.CUDA.PIDs,
 			"restored_cuda_pids", restorePIDs,
 			"criu_callback_pid", restoredPID,
 		)
-		if err := cuda.RestoreAndUnlockProcessTree(ctx, restorePIDs, opts.CUDADeviceMap, log); err != nil {
-			return 0, fmt.Errorf("CUDA restore failed: %w", err)
+		_, err = cuda.RestoreAndUnlockProcessTree(ctx, restorePIDs, opts.CUDADeviceMap, log)
+		if err != nil {
+			return nil, 0, fmt.Errorf("CUDA restore failed: %w", err)
 		}
 	}
+	timings.cudaDuration = time.Since(cudaStart)
 
-	return int(restoredPID), nil
+	return timings, int(restoredPID), nil
 }

deploy/snapshot/internal/executor/restore.go

@@ -27,6 +27,7 @@ import (
 type RestoreRequest struct {
 	CheckpointID       string
 	CheckpointLocation string
+	StartedAt          time.Time
 	NSRestorePath      string
 	PodName            string
 	PodNamespace       string
@@ -47,30 +48,53 @@ func Restore(ctx context.Context, ctrd *containerd.Client, log logr.Logger, req
 		"container", req.ContainerName,
 	)
 
-	// Phase 1: Inspect — resolve placeholder, discover target GPUs, build device map
+	// Phase 1: Host inspect — resolve placeholder, discover target GPUs, build device map
+	hostInspectStart := time.Now()
 	snap, err := inspectRestore(ctx, ctrd, log, req)
 	if err != nil {
 		return 0, err
 	}
+	hostInspectDuration := time.Since(hostInspectStart)
 
 	// Phase 2: Execute — nsrestore handles rootfs, CRIU restore, and CUDA restore inside namespace
-	restoredPID, err := execNSRestore(ctx, log, req, snap)
+	result, err := execNSRestore(ctx, log, req, snap)
 	if err != nil {
 		return 0, fmt.Errorf("nsrestore failed: %w", err)
 	}
-	log.Info("nsrestore completed", "restored_pid", restoredPID)
+	restoreDuration := hostInspectDuration + result.NSRestoreSetupDuration + result.CRIURestoreDuration + result.CUDADuration
+	log.Info("Restore timing summary",
+		"restore", map[string]any{
+			"duration": restoreDuration.String(),
+			"phases": map[string]string{
+				"host_inspect_duration":    hostInspectDuration.String(),
+				"nsrestore_setup_duration": result.NSRestoreSetupDuration.String(),
+				"criu_restore_duration":    result.CRIURestoreDuration.String(),
+				"cuda_duration":            result.CUDADuration.String(),
+			},
+		},
+	)
+	if !req.StartedAt.IsZero() {
+		log.Info("Restore wall time from agent detection",
+			"started_to_restore_complete", time.Since(req.StartedAt),
+		)
+	}
 
 	// Validate restored process from the host side
+	validationStart := time.Now()
 	procRoot := filepath.Join(snap.TargetRoot, "proc")
-	if err := snapshotruntime.ValidateProcessState(procRoot, restoredPID); err != nil {
+	if err := snapshotruntime.ValidateProcessState(procRoot, result.RestoredPID); err != nil {
 		restoreLogPath := filepath.Join(snap.TargetRoot, "var", "criu-work", criu.RestoreLogFilename)
-		logging.LogProcessDiagnostics(procRoot, restoredPID, restoreLogPath, log)
+		logging.LogProcessDiagnostics(procRoot, result.RestoredPID, restoreLogPath, log)
 		return 0, fmt.Errorf("restored process failed post-restore validation: %w", err)
 	}
 
-	log.Info("=== External restore completed ===", "total_duration", time.Since(restoreStart))
+	log.Info("=== External restore completed ===",
+		"restored_pid", result.RestoredPID,
+		"validation_duration", time.Since(validationStart),
+		"total_duration", time.Since(restoreStart),
+	)
 
-	return restoredPID, nil
+	return result.RestoredPID, nil
 }
 
 func inspectRestore(ctx context.Context, ctrd *containerd.Client, log logr.Logger, req RestoreRequest) (*types.RestoreContainerSnapshot, error) {
@@ -105,7 +129,7 @@ func inspectRestore(ctx context.Context, ctrd *containerd.Client, log logr.Logge
 	if err != nil {
 		return nil, fmt.Errorf("failed to resolve placeholder container: %w", err)
 	}
-	log.Info("Resolved placeholder container", "pid", placeholderPID)
+	log.V(1).Info("Resolved placeholder container", "pid", placeholderPID)
 
 	cgroupRoot, err := snapshotruntime.ResolveCgroupRootFromHostPID(placeholderPID)
 	if err != nil {
@@ -137,7 +161,7 @@ func inspectRestore(ctx context.Context, ctrd *containerd.Client, log logr.Logge
 		if err != nil {
 			return nil, fmt.Errorf("failed to build CUDA device map: %w", err)
 		}
-		log.Info("GPU UUIDs for device map",
+		log.V(1).Info("GPU UUIDs for device map",
 			"source_uuids", m.CUDA.SourceGPUUUIDs,
 			"target_uuids", targetGPUUUIDs,
 			"device_map", cudaDeviceMap,
@@ -155,7 +179,7 @@ func inspectRestore(ctx context.Context, ctrd *containerd.Client, log logr.Logge
 
 // execNSRestore launches the nsrestore binary inside the placeholder container's
 // namespaces via nsenter and parses the restored PID from stdout JSON.
-func execNSRestore(ctx context.Context, log logr.Logger, req RestoreRequest, snap *types.RestoreContainerSnapshot) (int, error) {
+func execNSRestore(ctx context.Context, log logr.Logger, req RestoreRequest, snap *types.RestoreContainerSnapshot) (*RestoreInNamespaceResult, error) {
 	args := []string{
 		"-t", strconv.Itoa(snap.PlaceholderPID),
 		// Intentionally exclude cgroup namespace (-C): CRIU must manage cgroups
@@ -181,18 +205,16 @@ func execNSRestore(ctx context.Context, log logr.Logger, req RestoreRequest, sna
 	cmd.Stderr = os.Stderr
 
 	if err := cmd.Run(); err != nil {
-		return 0, fmt.Errorf("nsrestore failed: %w\nstdout: %s", err, stdout.String())
+		return nil, fmt.Errorf("nsrestore failed: %w\nstdout: %s", err, stdout.String())
 	}
 
-	var result struct {
-		RestoredPID int `json:"restoredPID"`
-	}
+	var result RestoreInNamespaceResult
 	if err := json.Unmarshal(stdout.Bytes(), &result); err != nil {
-		return 0, fmt.Errorf("failed to parse nsrestore result: %w\nstdout: %s", err, stdout.String())
+		return nil, fmt.Errorf("failed to parse nsrestore result: %w\nstdout: %s", err, stdout.String())
 	}
 	if result.RestoredPID <= 0 {
-		return 0, fmt.Errorf("nsrestore returned invalid PID %d", result.RestoredPID)
+		return nil, fmt.Errorf("nsrestore returned invalid PID %d", result.RestoredPID)
 	}
 
-	return result.RestoredPID, nil
+	return &result, nil
 }