diff --git a/pkg/drift/defaults.go b/pkg/drift/defaults.go
index b4f82be..9a61d7d 100644
--- a/pkg/drift/defaults.go
+++ b/pkg/drift/defaults.go
@@ -7,5 +7,6 @@ package drift
 func DefaultDetectors() []Detector {
 	return []Detector{
 		NewKubernetesVersionDetector(),
+		NewRebootDetector(),
 	}
 }
diff --git a/pkg/drift/detector.go b/pkg/drift/detector.go
index 2b7608e..400e50d 100644
--- a/pkg/drift/detector.go
+++ b/pkg/drift/detector.go
@@ -25,6 +25,7 @@ type RemediationAction string
 const (
 	RemediationActionUnspecified       RemediationAction = ""
 	RemediationActionKubernetesUpgrade RemediationAction = "kubernetes-upgrade"
+	RemediationActionReboot            RemediationAction = "reboot"
 )
 
 // Remediation describes what the agent should do to address a drift.
diff --git a/pkg/drift/node_reboot.go b/pkg/drift/node_reboot.go
new file mode 100644
index 0000000..9d6874c
--- /dev/null
+++ b/pkg/drift/node_reboot.go
@@ -0,0 +1,51 @@
+package drift
+
+import (
+	"context"
+
+	"github.com/Azure/AKSFlexNode/pkg/config"
+	"github.com/Azure/AKSFlexNode/pkg/spec"
+	"github.com/Azure/AKSFlexNode/pkg/status"
+)
+
+const NodeRebootFindingID = "node-reboot"
+
+type RebootDetector struct{}
+
+func NewRebootDetector() *RebootDetector {
+	return &RebootDetector{}
+}
+
+func (d *RebootDetector) Name() string {
+	return "RebootDetector"
+}
+
+func (d *RebootDetector) Detect(
+	ctx context.Context,
+	_ *config.Config,
+	_ *spec.ManagedClusterSpec,
+	statusSnap *status.NodeStatus,
+) ([]Finding, error) {
+	if ctx.Err() != nil {
+		return nil, ctx.Err()
+	}
+
+	if statusSnap == nil {
+		return nil, nil
+	}
+
+	if !statusSnap.NeedReboot {
+		return nil, nil
+	}
+
+	return []Finding{
+		{
+			ID:      NodeRebootFindingID,
+			Title:   "Node reboot required",
+			Details: "Node status indicates a reboot is needed",
+			Remediation: Remediation{
+				Action: RemediationActionReboot,
+			},
+		},
+	}, nil
+}
diff --git a/pkg/drift/node_reboot_test.go b/pkg/drift/node_reboot_test.go
new file mode 100644
index 0000000..2082c13
--- /dev/null
+++ b/pkg/drift/node_reboot_test.go
@@ -0,0 +1,82 @@
+package drift
+
+import (
+	"context"
+	"testing"
+
+	"github.com/Azure/AKSFlexNode/pkg/status"
+)
+
+func TestRebootDetector_Name(t *testing.T) {
+	t.Parallel()
+	d := NewRebootDetector()
+	if name := d.Name(); name != "RebootDetector" {
+		t.Errorf("expected name %q, got %q", "RebootDetector", name)
+	}
+}
+
+func TestRebootDetector_NilStatus_NoFindings(t *testing.T) {
+	t.Parallel()
+	d := NewRebootDetector()
+	findings, err := d.Detect(context.Background(), nil, nil, nil)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(findings) != 0 {
+		t.Errorf("expected no findings, got %d", len(findings))
+	}
+}
+
+func TestRebootDetector_NeedRebootFalse_NoFindings(t *testing.T) {
+	t.Parallel()
+	d := NewRebootDetector()
+	statusSnap := &status.NodeStatus{
+		NeedReboot: false,
+	}
+	findings, err := d.Detect(context.Background(), nil, nil, statusSnap)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(findings) != 0 {
+		t.Errorf("expected no findings, got %d", len(findings))
+	}
+}
+
+func TestRebootDetector_NeedRebootTrue_ReturnsFinding(t *testing.T) {
+	t.Parallel()
+	d := NewRebootDetector()
+	statusSnap := &status.NodeStatus{
+		NeedReboot: true,
+	}
+	findings, err := d.Detect(context.Background(), nil, nil, statusSnap)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(findings) != 1 {
+		t.Fatalf("expected 1 finding, got %d", len(findings))
+	}
+
+	f := findings[0]
+	if f.ID != NodeRebootFindingID {
+		t.Errorf("expected ID %q, got %q", NodeRebootFindingID, f.ID)
+	}
+	if f.Title != "Node reboot required" {
+		t.Errorf("unexpected title: %q", f.Title)
+	}
+	if f.Remediation.Action != RemediationActionReboot {
+		t.Errorf("expected action %q, got %q", RemediationActionReboot, f.Remediation.Action)
+	}
+}
+
+func TestRebootDetector_CanceledContext_ReturnsError(t *testing.T) {
+	t.Parallel()
+	d := NewRebootDetector()
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+
+	statusSnap := &status.NodeStatus{NeedReboot: true}
+	_, err := d.Detect(ctx, nil, nil, statusSnap)
+	if err == nil {
+		t.Fatal("expected error from canceled context")
+	}
+}
diff --git a/pkg/drift/remediation.go b/pkg/drift/remediation.go
index bcf1bd9..90be5a6 100644
--- a/pkg/drift/remediation.go
+++ b/pkg/drift/remediation.go
@@ -5,6 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"os"
+	"os/exec"
 	"sync/atomic"
 	"time"
 
@@ -16,6 +17,7 @@ import (
 	"github.com/Azure/AKSFlexNode/pkg/kube"
 	"github.com/Azure/AKSFlexNode/pkg/spec"
 	"github.com/Azure/AKSFlexNode/pkg/status"
+	"github.com/Azure/AKSFlexNode/pkg/systemd"
 )
 
 const driftKubernetesUpgradeOperation = "drift-kubernetes-upgrade"
@@ -28,6 +30,8 @@ const (
 	upgradeStepUncordon             = "uncordon"
 )
 
+const agentServiceName = "aks-flex-node-agent.service"
+
 // maxManagedClusterSpecAge is a safety guard to avoid acting on very stale spec snapshots.
 // In normal operation we run drift immediately after a successful spec collection, so this
 // should rarely block remediation.
@@ -154,6 +158,13 @@ func detectAndRemediate(
 		logger.Info("Kubernetes upgrade remediation completed successfully")
 		return detectErr
 
+	case RemediationActionReboot:
+		if err := runRebootRemediation(ctx, logger); err != nil {
+			return fmt.Errorf("reboot remediation failed: %w", err)
+		}
+		logger.Info("Reboot remediation completed without error")
+		return detectErr
+
 	default:
 		return fmt.Errorf("unsupported drift remediation action: %q", plan.Action)
 	}
@@ -267,6 +278,67 @@ func runKubernetesUpgradeRemediation(
 	return result, err
 }
 
+func runRebootRemediation(
+	ctx context.Context,
+	logger *logrus.Logger,
+) error {
+	// Key design points:
+	//   - Only reboot if aks-flex-node-agent is running as a systemd service
+	//   - If not running under systemd, skip reboot (agent may be running in development/test mode)
+	//   - Use systemctl reboot for a clean shutdown
+	if logger == nil {
+		logger = logrus.New()
+	}
+
+	// Check if aks-flex-node-agent is managed by systemd.
+	// We use GetUnitStatus to check if the service is active and running under systemd.
+	mgr := systemd.New()
+	status, err := mgr.GetUnitStatus(ctx, agentServiceName)
+	if err != nil {
+		if errors.Is(err, systemd.ErrUnitNotFound) {
+			logger.Warn("aks-flex-node-agent is not running as a systemd service; skipping reboot")
+			// Not running under systemd is an expected scenario (e.g., dev/test); treat as a no-op, not an error.
+			return nil
+		}
+		logger.WithError(err).Warn("Failed to check systemd service status; aborting reboot remediation")
+		return fmt.Errorf("failed to check systemd service status: %w", err)
+	}
+
+	// Only reboot if the service is active
+	if status.ActiveState != systemd.UnitActiveStateActive {
+		logger.Warnf("aks-flex-node-agent service is not active (state: %s); skipping reboot", status.ActiveState)
+		return nil
+	}
+
+	logger.Info("Initiating system reboot via systemctl")
+
+	// Use systemctl reboot for a clean shutdown.
+	// This will gracefully stop services and sync filesystems before rebooting.
+	// To avoid silently ignoring immediate failures (e.g., DBus unavailable), run the
+	// command and check its exit status, using a short timeout if no deadline is set.
+	rebootCtx := ctx
+	if _, hasDeadline := ctx.Deadline(); !hasDeadline {
+		var cancel context.CancelFunc
+		rebootCtx, cancel = context.WithTimeout(ctx, 30*time.Second)
+		defer cancel()
+	}
+
+	cmd := exec.CommandContext(rebootCtx, "systemctl", "reboot")
+	output, err := cmd.CombinedOutput()
+	if err != nil {
+		// If the context was canceled or timed out, surface that information explicitly.
+		if errors.Is(rebootCtx.Err(), context.DeadlineExceeded) {
+			logger.WithError(err).WithField("output", string(output)).
+				Error("systemctl reboot timed out")
+			return fmt.Errorf("systemctl reboot timed out: %w", err)
+		}
+		logger.WithError(err).WithField("output", string(output)).
+			Error("systemctl reboot failed")
+		return fmt.Errorf("systemctl reboot failed: %w", err)
+	}
+	return nil
+}
+
 // handleExecutionResult mirrors main's handleExecutionResult but lives in drift so remediation
 // can share the same logging and error semantics.
 func handleExecutionResult(result *bootstrapper.ExecutionResult, operation string, logger *logrus.Logger) error {
diff --git a/pkg/status/collector.go b/pkg/status/collector.go
index dc4d72b..4cd7b19 100644
--- a/pkg/status/collector.go
+++ b/pkg/status/collector.go
@@ -3,6 +3,7 @@ package status
 import (
 	"context"
 	"encoding/json"
+	"fmt"
 	"os"
 	"os/exec"
 	"path/filepath"
@@ -13,6 +14,7 @@ import (
 	"github.com/Azure/AKSFlexNode/pkg/kube"
 	"github.com/Azure/AKSFlexNode/pkg/utils"
 	"github.com/sirupsen/logrus"
+	"golang.org/x/sys/unix"
 	corev1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 )
@@ -60,6 +62,9 @@ func (c *Collector) CollectStatus(ctx context.Context) (*NodeStatus, error) {
 	}
 	status.ArcStatus = arcStatus
 
+	// Check if reboot is needed node condition
+	status.NeedReboot = c.checkRebootNeeded(ctx)
+
 	return status, nil
 }
 
@@ -297,3 +302,71 @@ func GetStatusFilePath() string {
 	// Fallback to temp directory for testing/development
 	return filepath.Join("/tmp/aks-flex-node", "status.json")
 }
+
+func getBootTime() (time.Time, error) {
+	var sysinfo unix.Sysinfo_t
+	if err := unix.Sysinfo(&sysinfo); err != nil {
+		return time.Time{}, fmt.Errorf("failed to get system info: %w", err)
+	}
+
+	// Calculate boot time: current time - uptime
+	// Sysinfo.Uptime is in seconds since boot
+	uptime := time.Duration(sysinfo.Uptime) * time.Second
+	bootTime := time.Now().Add(-uptime)
+	return bootTime, nil
+}
+
+func getNodeName() (string, error) {
+	host, err := os.Hostname()
+	if err != nil {
+		return "", fmt.Errorf("failed to get hostname: %w", err)
+	}
+
+	nodeName := strings.ToLower(strings.TrimSpace(host))
+	if nodeName == "" {
+		return "", fmt.Errorf("node name is empty")
+	}
+
+	return nodeName, nil
+}
+
+func (c *Collector) checkRebootNeeded(ctx context.Context) bool {
+	hostBootTime, err := getBootTime()
+	if err != nil {
+		c.logger.Warnf("Failed to get boot time: %v", err)
+		return false
+	}
+	nodeName, err := getNodeName()
+	if err != nil {
+		c.logger.Errorf("failed to get node name: %s", err.Error())
+		return false
+	}
+
+	clientset, err := kube.KubeletClientset()
+	if err != nil {
+		c.logger.Errorf("failed to get kubelet clientset: %s", err.Error())
+		return false
+	}
+
+	// Get the node with a timeout and respecting the passed-in context
+	ctxWithTimeout, cancel := context.WithTimeout(ctx, 10*time.Second)
+	defer cancel()
+
+	node, err := clientset.CoreV1().Nodes().Get(ctxWithTimeout, nodeName, metav1.GetOptions{})
+	if err != nil {
+		c.logger.Errorf("failed to get node: %s", err.Error())
+		return false
+	}
+
+	for _, condition := range node.Status.Conditions {
+		switch condition.Type {
+		case "KernelDeadlock":
+			if condition.Status == corev1.ConditionTrue && condition.LastTransitionTime.After(hostBootTime) {
+				c.logger.Infof("Node has a kernel deadlock since %s, rebooting...",
+					condition.LastTransitionTime.Format("2006-01-02 15:04:05"))
+				return true
+			}
+		}
+	}
+	return false
+}
diff --git a/pkg/status/types.go b/pkg/status/types.go
index ea32d81..c2c1aab 100644
--- a/pkg/status/types.go
+++ b/pkg/status/types.go
@@ -41,6 +41,8 @@ type NodeStatus struct {
 	LastUpdatedBy     LastUpdatedBy     `json:"lastUpdatedBy,omitempty"`
 	LastUpdatedReason LastUpdatedReason `json:"lastUpdatedReason,omitempty"`
 	AgentVersion      string            `json:"agentVersion"`
+
+	NeedReboot bool `json:"needReboot,omitempty"`
 }
 
 // ArcStatus contains Azure Arc machine registration and connection status