diff --git a/staging/src/k8s.io/kube-aggregator/pkg/controllers/status/available_controller.go b/staging/src/k8s.io/kube-aggregator/pkg/controllers/status/available_controller.go
index 61205ad7d73..918a01c18e7 100644
--- a/staging/src/k8s.io/kube-aggregator/pkg/controllers/status/available_controller.go
+++ b/staging/src/k8s.io/kube-aggregator/pkg/controllers/status/available_controller.go
@@ -321,17 +321,13 @@ func (c *AvailableConditionController) sync(key string) error {
 					// we had trouble with slow dial and DNS responses causing us to wait too long.
 					// we added this as insurance
 				case <-time.After(6 * time.Second):
-        			defer func() {
-            			// errCh needs to have a reader since the above goroutine doing the work 
-            			// needs to send to it. This is defered to ensure it runs
-            			// even if the post timeout work itself panics.
-            			go func() {
-                			res := <-errch
-                			if res != nil {
-                    			fmt.Error("%v", res)
-                			}	   
-            			}() 
-        			}() 
+					// errCh needs to have a reader since the above goroutine doing the work 
+					// needs to send to it. 
+					go func() {
+						if res := <-errch; res != nil {
+							fmt.Errorf("timed out response from %v: %v", discoveryURL, res)
+						}
+					}()
 					results <- fmt.Errorf("timed out waiting for %v", discoveryURL)
 					return
 				}