Agents stop update managedcluster status when clock is out of sync.

Signed-off-by: xuezhaojun <[email protected]>

Author: xuezhaojun

pkg/registration/hub/lease/clocksynccontroller.go

@@ -103,8 +103,9 @@ func (c *clockSyncController) sync(ctx context.Context, syncCtx factory.SyncCont
 		return err
 	}
 	// When the agent's lease get renewed, the "now" on hub should close to the RenewTime on agent.
-	// If the two time are not close(over 1 lease duration), we assume the clock is out of sync.
-	oneLeaseDuration := time.Duration(LeaseDurationSeconds) * time.Second
+	// If the two time are not close(the same duration in the lease controller), we assume the clock is out of sync.
+	// Then, if the Clock is out of sync, the agent will not be able to update the status of managed cluster.
+	oneLeaseDuration := time.Duration(LeaseDurationSeconds*leaseDurationTimes) * time.Second
 	if err := c.updateClusterStatusClockSynced(ctx, cluster,
 		now.Sub(observedLease.Spec.RenewTime.Time) < oneLeaseDuration && observedLease.Spec.RenewTime.Time.Sub(now) < oneLeaseDuration); err != nil {
 		return err

pkg/registration/hub/lease/clocksynccontroller_test.go

@@ -63,7 +63,7 @@ func TestClockSyncController(t *testing.T) {
 				testinghelpers.NewManagedCluster(),
 			},
 			leases: []runtime.Object{
-				testinghelpers.NewManagedClusterLease("managed-cluster-lease", now.Add(61*time.Second)),
+				testinghelpers.NewManagedClusterLease("managed-cluster-lease", now.Add(301*time.Second)),
 			},
 			validateActions: func(t *testing.T, leaseActions, clusterActions []clienttesting.Action) {
 				expected := metav1.Condition{

pkg/registration/spoke/managedcluster/status_controller.go

@@ -31,6 +31,7 @@ type managedClusterStatusController struct {
 	patcher          patcher.Patcher[*clusterv1.ManagedCluster, clusterv1.ManagedClusterSpec, clusterv1.ManagedClusterStatus]
 	hubClusterLister clusterv1listers.ManagedClusterLister
 	hubEventRecorder kevents.EventRecorder
+	recorder         events.Recorder
 }
 
 type statusReconcile interface {
@@ -97,6 +98,7 @@ func newManagedClusterStatusController(
 		},
 		hubClusterLister: hubClusterInformer.Lister(),
 		hubEventRecorder: hubEventRecorder,
+		recorder:         recorder,
 	}
 }
 
@@ -121,6 +123,13 @@ func (c *managedClusterStatusController) sync(ctx context.Context, syncCtx facto
 		}
 	}
 
+	// check if managedcluster's clock is out of sync, if so, the agent will not be able to update the status of managed cluster.
+	outOfSynced := meta.IsStatusConditionFalse(newCluster.Status.Conditions, clusterv1.ManagedClusterConditionClockSynced)
+	if outOfSynced {
+		c.recorder.Eventf("ClockOutOfSync", "The managed cluster's clock is out of sync, the agent will not be able to update the status of managed cluster.")
+		return fmt.Errorf("the managed cluster's clock is out of sync, the agent will not be able to update the status of managed cluster.")
+	}
+
 	changed, err := c.patcher.PatchStatus(ctx, newCluster, newCluster.Status, cluster.Status)
 	if err != nil {
 		errs = append(errs, err)