diff --git a/pkg/kubelet/kubelet.go b/pkg/kubelet/kubelet.go
index 421c0c98be..6968783c8a 100644
--- a/pkg/kubelet/kubelet.go
+++ b/pkg/kubelet/kubelet.go
@@ -1091,6 +1091,10 @@ type Kubelet struct {
 	// handlers called during the tryUpdateNodeStatus cycle
 	setNodeStatusFuncs []func(*v1.Node) error
 
+	lastNodeUnschedulableLock sync.Mutex
+	// maintains Node.Spec.Unschedulable value from previous run of tryUpdateNodeStatus()
+	lastNodeUnschedulable bool
+
 	// TODO: think about moving this to be centralized in PodWorkers in follow-on.
 	// the list of handlers to call during pod admission.
 	admitHandlers lifecycle.PodAdmitHandlers
diff --git a/pkg/kubelet/kubelet_node_status.go b/pkg/kubelet/kubelet_node_status.go
index 9a1e6214b6..753913ac6c 100644
--- a/pkg/kubelet/kubelet_node_status.go
+++ b/pkg/kubelet/kubelet_node_status.go
@@ -23,7 +23,6 @@ import (
 	"net"
 	goruntime "runtime"
 	"strings"
-	"sync"
 	"time"
 
 	"github.com/golang/glog"
@@ -1034,24 +1033,17 @@ func (kl *Kubelet) setNodeOODCondition(node *v1.Node) {
 	}
 }
 
-// Maintains Node.Spec.Unschedulable value from previous run of tryUpdateNodeStatus()
-// TODO: why is this a package var?
-var (
-	oldNodeUnschedulable     bool
-	oldNodeUnschedulableLock sync.Mutex
-)
-
 // record if node schedulable change.
 func (kl *Kubelet) recordNodeSchedulableEvent(node *v1.Node) {
-	oldNodeUnschedulableLock.Lock()
-	defer oldNodeUnschedulableLock.Unlock()
-	if oldNodeUnschedulable != node.Spec.Unschedulable {
+	kl.lastNodeUnschedulableLock.Lock()
+	defer kl.lastNodeUnschedulableLock.Unlock()
+	if kl.lastNodeUnschedulable != node.Spec.Unschedulable {
 		if node.Spec.Unschedulable {
 			kl.recordNodeStatusEvent(v1.EventTypeNormal, events.NodeNotSchedulable)
 		} else {
 			kl.recordNodeStatusEvent(v1.EventTypeNormal, events.NodeSchedulable)
 		}
-		oldNodeUnschedulable = node.Spec.Unschedulable
+		kl.lastNodeUnschedulable = node.Spec.Unschedulable
 	}
 }