addressed feedback. added opt-in cmd line flag

cmd/kube-controller-manager/app/controllermanager.go

@@ -60,6 +60,7 @@ type CMServer struct {
 	ResourceQuotaSyncPeriod time.Duration
 	NamespaceSyncPeriod     time.Duration
 	PVClaimBinderSyncPeriod time.Duration
+	EnablePVCClaimBinder    bool
 	RegisterRetryCount      int
 	MachineList             util.StringList
 	SyncNodeList            bool
@@ -117,6 +118,7 @@ func (s *CMServer) AddFlags(fs *pflag.FlagSet) {
 	fs.DurationVar(&s.ResourceQuotaSyncPeriod, "resource_quota_sync_period", s.ResourceQuotaSyncPeriod, "The period for syncing quota usage status in the system")
 	fs.DurationVar(&s.NamespaceSyncPeriod, "namespace_sync_period", s.NamespaceSyncPeriod, "The period for syncing namespace life-cycle updates")
 	fs.DurationVar(&s.PVClaimBinderSyncPeriod, "pvclaimbinder_sync_period", s.PVClaimBinderSyncPeriod, "The period for syncing persistent volumes and persistent volume claims")
+	fs.BoolVar(&s.EnablePVCClaimBinder, "enable_alpha_pvclaimbinder", s.EnablePVCClaimBinder, "Optionally enable persistent volume claim binding.  This feature is experimental and expected to change.")
 	fs.DurationVar(&s.PodEvictionTimeout, "pod_eviction_timeout", s.PodEvictionTimeout, "The grace peroid for deleting pods on failed nodes.")
 	fs.Float32Var(&s.DeletingPodsQps, "deleting_pods_qps", 0.1, "Number of nodes per second on which pods are deleted in case of node failure.")
 	fs.IntVar(&s.DeletingPodsBurst, "deleting_pods_burst", 10, "Number of nodes on which pods are bursty deleted in case of node failure. For more details look into RateLimiter.")
@@ -235,8 +237,10 @@ func (s *CMServer) Run(_ []string) error {
 	namespaceManager := namespace.NewNamespaceManager(kubeClient, s.NamespaceSyncPeriod)
 	namespaceManager.Run()
 
-	pvclaimBinder := volumeclaimbinder.NewPersistentVolumeClaimBinder(kubeClient, s.PVClaimBinderSyncPeriod)
-	pvclaimBinder.Run()
+	if s.EnablePVCClaimBinder {
+		pvclaimBinder := volumeclaimbinder.NewPersistentVolumeClaimBinder(kubeClient, s.PVClaimBinderSyncPeriod)
+		pvclaimBinder.Run()
+	}
 
 	select {}
 	return nil

pkg/api/testing/fuzzer.go

@@ -203,11 +203,13 @@ func FuzzerFor(t *testing.T, version string, src rand.Source) *fuzz.Fuzzer {
 		},
 		func(pv *api.PersistentVolume, c fuzz.Continue) {
 			c.FuzzNoCustom(pv) // fuzz self without calling this function again
-			pv.Status.Phase = api.VolumePending
+			types := []api.PersistentVolumePhase{api.VolumePending, api.VolumeBound, api.VolumeReleased, api.VolumeAvailable}
+			pv.Status.Phase = types[c.Rand.Intn(len(types))]
 		},
 		func(pvc *api.PersistentVolumeClaim, c fuzz.Continue) {
 			c.FuzzNoCustom(pvc) // fuzz self without calling this function again
-			pvc.Status.Phase = api.ClaimPending
+			types := []api.PersistentVolumeClaimPhase{api.ClaimBound, api.ClaimPending}
+			pvc.Status.Phase = types[c.Rand.Intn(len(types))]
 		},
 		func(s *api.NamespaceSpec, c fuzz.Continue) {
 			s.Finalizers = []api.FinalizerName{api.FinalizerKubernetes}

pkg/volumeclaimbinder/persistent_volume_claim_binder.go

@@ -84,6 +84,8 @@ func NewPersistentVolumeClaimBinder(kubeClient client.Interface, syncPeriod time
 		framework.ResourceEventHandlerFuncs{
 			AddFunc:    binder.addClaim,
 			UpdateFunc: binder.updateClaim,
+			// no DeleteFunc needed.  a claim requires no clean-up.
+			// the missing claim itself is the release of the resource.
 		},
 	)
 
@@ -177,6 +179,8 @@ func syncVolume(volumeIndex *persistentVolumeOrderedIndex, binderClient binderCl
 			if err == nil {
 				// bound and active.  Build claim status as needed.
 				if claim.Status.VolumeRef == nil {
+					// syncClaimStatus sets VolumeRef, attempts to persist claim status,
+					// and does a rollback as needed on claim.Status
 					syncClaimStatus(binderClient, volume, claim)
 				}
 			} else {
@@ -198,6 +202,9 @@ func syncVolume(volumeIndex *persistentVolumeOrderedIndex, binderClient binderCl
 
 	if currentPhase != nextPhase {
 		volume.Status.Phase = nextPhase
+
+		// a change in state will trigger another update through this controller.
+		// each pass through this controller evaluates current phase and decides whether or not to change to the next phase
 		volume, err := binderClient.UpdatePersistentVolumeStatus(volume)
 		if err != nil {
 			// Rollback to previous phase
@@ -220,7 +227,6 @@ func syncClaim(volumeIndex *persistentVolumeOrderedIndex, binderClient binderCli
 	nextPhase := currentPhase
 
 	switch currentPhase {
-
 	// pending claims await a matching volume
 	case api.ClaimPending:
 		volume, err := volumeIndex.FindBestMatchForClaim(claim)