Hotfix: Ensure multiple Pods don't get created for a GameServer

Since the cache is eventually consistent, there was a race condition
in which a GameServer could create more than one Pod backing it,
as the first one isn't found in the cache yet.

This sync's the cache before checking this value, so the cache
is up to date.

pkg/gameservers/controller.go

@@ -61,13 +61,15 @@ type Controller struct {
 	crdGetter              v1beta1.CustomResourceDefinitionInterface
 	podGetter              typedcorev1.PodsGetter
 	podLister              corelisterv1.PodLister
+	podSynced              cache.InformerSynced
 	gameServerGetter       getterv1alpha1.GameServersGetter
 	gameServerLister       listerv1alpha1.GameServerLister
 	gameServerSynced       cache.InformerSynced
 	nodeLister             corelisterv1.NodeLister
 	portAllocator          *PortAllocator
 	healthController       *HealthController
 	workerqueue            *workerqueue.WorkerQueue
+	stop                   <-chan struct{}
 	recorder               record.EventRecorder
 }
 
@@ -84,6 +86,7 @@ func NewController(
 	agonesClient versioned.Interface,
 	agonesInformerFactory externalversions.SharedInformerFactory) *Controller {
 
+	pods := kubeInformerFactory.Core().V1().Pods()
 	gameServers := agonesInformerFactory.Stable().V1alpha1().GameServers()
 	gsInformer := gameServers.Informer()
 
@@ -92,7 +95,8 @@ func NewController(
 		alwaysPullSidecarImage: alwaysPullSidecarImage,
 		crdGetter:              extClient.ApiextensionsV1beta1().CustomResourceDefinitions(),
 		podGetter:              kubeClient.CoreV1(),
-		podLister:              kubeInformerFactory.Core().V1().Pods().Lister(),
+		podLister:              pods.Lister(),
+		podSynced:              pods.Informer().HasSynced,
 		gameServerGetter:       agonesClient.StableV1alpha1(),
 		gameServerLister:       gameServers.Lister(),
 		gameServerSynced:       gsInformer.HasSynced,
@@ -127,7 +131,7 @@ func NewController(
 	})
 
 	// track pod deletions, for when GameServers are deleted
-	kubeInformerFactory.Core().V1().Pods().Informer().AddEventHandler(cache.ResourceEventHandlerFuncs{
+	pods.Informer().AddEventHandler(cache.ResourceEventHandlerFuncs{
 		DeleteFunc: func(obj interface{}) {
 			// Could be a DeletedFinalStateUnknown, in which case, just ignore it
 			pod, ok := obj.(*corev1.Pod)
@@ -221,6 +225,8 @@ func (c *Controller) creationValidationHandler(review admv1beta1.AdmissionReview
 // Run the GameServer controller. Will block until stop is closed.
 // Runs threadiness number workers to process the rate limited queue
 func (c *Controller) Run(workers int, stop <-chan struct{}) error {
+	c.stop = stop
+
 	err := crd.WaitForEstablishedCRD(c.crdGetter, "gameservers.stable.agones.dev", c.logger)
 	if err != nil {
 		return err
@@ -362,6 +368,11 @@ func (c *Controller) syncGameServerCreatingState(gs *v1alpha1.GameServer) (*v1al
 
 	c.logger.WithField("gs", gs).Info("Syncing Create State")
 
+	// Wait for pod cache sync, so that we don't end up with multiple pods for a GameServer
+	if !(cache.WaitForCacheSync(c.stop, c.podSynced)) {
+		return nil, errors.New("could not sync pod cache state")
+	}
+
 	// Maybe something went wrong, and the pod was created, but the state was never moved to Starting, so let's check
 	ret, err := c.listGameServerPods(gs)
 	if err != nil {

pkg/gameservers/localsdk_test.go

@@ -85,7 +85,7 @@ func TestLocalSDKServerSetLabel(t *testing.T) {
 		},
 	}
 
-	for k,v := range fixtures {
+	for k, v := range fixtures {
 		t.Run(k, func(t *testing.T) {
 			ctx := context.Background()
 			e := &sdk.Empty{}