diff --git a/pkg/controller/deployment/recreate.go b/pkg/controller/deployment/recreate.go
index f6225249418..8b8b80a0905 100644
--- a/pkg/controller/deployment/recreate.go
+++ b/pkg/controller/deployment/recreate.go
@@ -83,7 +83,7 @@ func (dc *DeploymentController) scaleDownOldReplicaSetsForRecreate(ctx context.C
 		if *(rs.Spec.Replicas) == 0 {
 			continue
 		}
-		scaledRS, updatedRS, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, rs, 0, deployment)
+		scaledRS, updatedRS, err := dc.scaleReplicaSet(ctx, rs, 0, deployment, false)
 		if err != nil {
 			return false, err
 		}
@@ -127,6 +127,6 @@ func oldPodsRunning(newRS *apps.ReplicaSet, oldRSs []*apps.ReplicaSet, podMap ma
 
 // scaleUpNewReplicaSetForRecreate scales up new replica set when deployment strategy is "Recreate".
 func (dc *DeploymentController) scaleUpNewReplicaSetForRecreate(ctx context.Context, newRS *apps.ReplicaSet, deployment *apps.Deployment) (bool, error) {
-	scaled, _, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, newRS, *(deployment.Spec.Replicas), deployment)
+	scaled, _, err := dc.scaleReplicaSet(ctx, newRS, *(deployment.Spec.Replicas), deployment, false)
 	return scaled, err
 }
diff --git a/pkg/controller/deployment/rolling.go b/pkg/controller/deployment/rolling.go
index b0b833bf0ed..d32584881b1 100644
--- a/pkg/controller/deployment/rolling.go
+++ b/pkg/controller/deployment/rolling.go
@@ -72,14 +72,14 @@ func (dc *DeploymentController) reconcileNewReplicaSet(ctx context.Context, allR
 	}
 	if *(newRS.Spec.Replicas) > *(deployment.Spec.Replicas) {
 		// Scale down.
-		scaled, _, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, newRS, *(deployment.Spec.Replicas), deployment)
+		scaled, _, err := dc.scaleReplicaSet(ctx, newRS, *(deployment.Spec.Replicas), deployment, false)
 		return scaled, err
 	}
 	newReplicasCount, err := deploymentutil.NewRSNewReplicas(deployment, allRSs, newRS)
 	if err != nil {
 		return false, err
 	}
-	scaled, _, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, newRS, newReplicasCount, deployment)
+	scaled, _, err := dc.scaleReplicaSet(ctx, newRS, newReplicasCount, deployment, false)
 	return scaled, err
 }
 
@@ -178,7 +178,7 @@ func (dc *DeploymentController) cleanupUnhealthyReplicas(ctx context.Context, ol
 		if newReplicasCount > *(targetRS.Spec.Replicas) {
 			return nil, 0, fmt.Errorf("when cleaning up unhealthy replicas, got invalid request to scale down %s/%s %d -> %d", targetRS.Namespace, targetRS.Name, *(targetRS.Spec.Replicas), newReplicasCount)
 		}
-		_, updatedOldRS, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, targetRS, newReplicasCount, deployment)
+		_, updatedOldRS, err := dc.scaleReplicaSet(ctx, targetRS, newReplicasCount, deployment, false)
 		if err != nil {
 			return nil, totalScaledDown, err
 		}
@@ -223,7 +223,7 @@ func (dc *DeploymentController) scaleDownOldReplicaSetsForRollingUpdate(ctx cont
 		if newReplicasCount > *(targetRS.Spec.Replicas) {
 			return 0, fmt.Errorf("when scaling down old RS, got invalid request to scale down %s/%s %d -> %d", targetRS.Namespace, targetRS.Name, *(targetRS.Spec.Replicas), newReplicasCount)
 		}
-		_, _, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, targetRS, newReplicasCount, deployment)
+		_, _, err := dc.scaleReplicaSet(ctx, targetRS, newReplicasCount, deployment, false)
 		if err != nil {
 			return totalScaledDown, err
 		}
diff --git a/pkg/controller/deployment/sync.go b/pkg/controller/deployment/sync.go
index 2112f2b8467..d2700981b51 100644
--- a/pkg/controller/deployment/sync.go
+++ b/pkg/controller/deployment/sync.go
@@ -311,7 +311,7 @@ func (dc *DeploymentController) scale(ctx context.Context, deployment *apps.Depl
 		if *(activeOrLatest.Spec.Replicas) == *(deployment.Spec.Replicas) {
 			return nil
 		}
-		_, _, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, activeOrLatest, *(deployment.Spec.Replicas), deployment)
+		_, _, err := dc.scaleReplicaSet(ctx, activeOrLatest, *(deployment.Spec.Replicas), deployment, false)
 		return err
 	}
 
@@ -319,7 +319,7 @@ func (dc *DeploymentController) scale(ctx context.Context, deployment *apps.Depl
 	// This case handles replica set adoption during a saturated new replica set.
 	if deploymentutil.IsSaturated(deployment, newRS) {
 		for _, old := range controller.FilterActiveReplicaSets(oldRSs) {
-			if _, _, err := dc.scaleReplicaSetWithLazyAnnotationUpdate(ctx, old, 0, deployment); err != nil {
+			if _, _, err := dc.scaleReplicaSet(ctx, old, 0, deployment, false); err != nil {
 				return err
 			}
 		}
@@ -391,7 +391,7 @@ func (dc *DeploymentController) scale(ctx context.Context, deployment *apps.Depl
 			}
 
 			// TODO: Use transactions when we have them.
-			if _, _, err := dc.scaleReplicaSet(ctx, rs, nameToSize[rs.Name], deployment); err != nil {
+			if _, _, err := dc.scaleReplicaSet(ctx, rs, nameToSize[rs.Name], deployment, true); err != nil {
 				// Return as soon as we fail, the deployment is requeued
 				return err
 			}
@@ -400,23 +400,18 @@ func (dc *DeploymentController) scale(ctx context.Context, deployment *apps.Depl
 	return nil
 }
 
-// scaleReplicaSetWithLazyAnnotationUpdate does not update the replica set annotations (DesiredReplicasAnnotation and
-// MaxReplicasAnnotation) if no replica set scaling is requested.
-// IMPORTANT: This method should not be called when an annotation update is necessary (e.g. scale during a rolling update)
-// and scaleReplicaSet should be used instead.
-func (dc *DeploymentController) scaleReplicaSetWithLazyAnnotationUpdate(ctx context.Context, rs *apps.ReplicaSet, newScale int32, deployment *apps.Deployment) (bool, *apps.ReplicaSet, error) {
-	// No need to scale
-	if *(rs.Spec.Replicas) == newScale {
+// scaleReplicaSet acts in two modes:
+//  1. lazy update - when forceUpdate is false, it will only act when the replicas
+//     where changed;
+//  2. normal - either replicaset annotations (DesiredReplicasAnnotation and
+//     MaxReplicasAnnotation) require an update or the replicas have changed.
+func (dc *DeploymentController) scaleReplicaSet(ctx context.Context, rs *apps.ReplicaSet, newScale int32, deployment *apps.Deployment, forceUpdate bool) (bool, *apps.ReplicaSet, error) {
+	// Don't scale, unless it's a forced update or the replicas actually differ
+	if !forceUpdate && *(rs.Spec.Replicas) == newScale {
 		return false, rs, nil
 	}
-	scaled, newRS, err := dc.scaleReplicaSet(ctx, rs, newScale, deployment)
-	return scaled, newRS, err
-}
-
-func (dc *DeploymentController) scaleReplicaSet(ctx context.Context, rs *apps.ReplicaSet, newScale int32, deployment *apps.Deployment) (bool, *apps.ReplicaSet, error) {
 
 	sizeNeedsUpdate := *(rs.Spec.Replicas) != newScale
-
 	annotationsNeedUpdate := deploymentutil.ReplicasAnnotationsNeedUpdate(rs, *(deployment.Spec.Replicas), *(deployment.Spec.Replicas)+deploymentutil.MaxSurge(*deployment))
 
 	scaled := false