prevent RC hotloop on denied pods

This commit is contained in:
deads2k 2016-08-15 10:36:46 -04:00
parent 9696a27aa0
commit 7cd51b4610
2 changed files with 101 additions and 81 deletions

View File

@ -112,7 +112,7 @@ type ReplicationManager struct {
lookupCache *controller.MatchingCache
// Controllers that need to be synced
queue *workqueue.Type
queue workqueue.RateLimitingInterface
// garbageCollectorEnabled denotes if the garbage collector is enabled. RC
// manager behaves differently if GC is enabled.
@ -143,7 +143,7 @@ func newReplicationManager(eventRecorder record.EventRecorder, podInformer frame
},
burstReplicas: burstReplicas,
expectations: controller.NewUIDTrackingControllerExpectations(controller.NewControllerExpectations()),
queue: workqueue.New(),
queue: workqueue.NewRateLimitingQueue(workqueue.DefaultControllerRateLimiter()),
garbageCollectorEnabled: garbageCollectorEnabled,
}
@ -464,10 +464,15 @@ func (rm *ReplicationManager) worker() {
return true
}
defer rm.queue.Done(key)
err := rm.syncHandler(key.(string))
if err != nil {
glog.Errorf("Error syncing replication controller: %v", err)
if err == nil {
rm.queue.Forget(key)
return false
}
rm.queue.AddRateLimited(key)
utilruntime.HandleError(err)
return false
}
for {
@ -480,13 +485,16 @@ func (rm *ReplicationManager) worker() {
// manageReplicas checks and updates replicas for the given replication controller.
// Does NOT modify <filteredPods>.
func (rm *ReplicationManager) manageReplicas(filteredPods []*api.Pod, rc *api.ReplicationController) {
func (rm *ReplicationManager) manageReplicas(filteredPods []*api.Pod, rc *api.ReplicationController) error {
diff := len(filteredPods) - int(rc.Spec.Replicas)
rcKey, err := controller.KeyFunc(rc)
if err != nil {
glog.Errorf("Couldn't get key for replication controller %#v: %v", rc, err)
return
return err
}
if diff == 0 {
return nil
}
if diff < 0 {
diff *= -1
if diff > rm.burstReplicas {
@ -497,6 +505,7 @@ func (rm *ReplicationManager) manageReplicas(filteredPods []*api.Pod, rc *api.Re
// UID, which would require locking *across* the create, which will turn
// into a performance bottleneck. We should generate a UID for the pod
// beforehand and store it via ExpectCreations.
errCh := make(chan error, diff)
rm.expectations.ExpectCreations(rcKey, diff)
var wg sync.WaitGroup
wg.Add(diff)
@ -522,55 +531,79 @@ func (rm *ReplicationManager) manageReplicas(filteredPods []*api.Pod, rc *api.Re
// Decrement the expected number of creates because the informer won't observe this pod
glog.V(2).Infof("Failed creation, decrementing expectations for controller %q/%q", rc.Namespace, rc.Name)
rm.expectations.CreationObserved(rcKey)
rm.enqueueController(rc)
errCh <- err
utilruntime.HandleError(err)
}
}()
}
wg.Wait()
} else if diff > 0 {
if diff > rm.burstReplicas {
diff = rm.burstReplicas
select {
case err := <-errCh:
// all errors have been reported before and they're likely to be the same, so we'll only return the first one we hit.
if err != nil {
return err
}
default:
}
glog.V(2).Infof("Too many %q/%q replicas, need %d, deleting %d", rc.Namespace, rc.Name, rc.Spec.Replicas, diff)
// No need to sort pods if we are about to delete all of them
if rc.Spec.Replicas != 0 {
// Sort the pods in the order such that not-ready < ready, unscheduled
// < scheduled, and pending < running. This ensures that we delete pods
// in the earlier stages whenever possible.
sort.Sort(controller.ActivePods(filteredPods))
}
// Snapshot the UIDs (ns/name) of the pods we're expecting to see
// deleted, so we know to record their expectations exactly once either
// when we see it as an update of the deletion timestamp, or as a delete.
// Note that if the labels on a pod/rc change in a way that the pod gets
// orphaned, the rs will only wake up after the expectations have
// expired even if other pods are deleted.
deletedPodKeys := []string{}
for i := 0; i < diff; i++ {
deletedPodKeys = append(deletedPodKeys, controller.PodKey(filteredPods[i]))
}
// We use pod namespace/name as a UID to wait for deletions, so if the
// labels on a pod/rc change in a way that the pod gets orphaned, the
// rc will only wake up after the expectation has expired.
rm.expectations.ExpectDeletions(rcKey, deletedPodKeys)
var wg sync.WaitGroup
wg.Add(diff)
for i := 0; i < diff; i++ {
go func(ix int) {
defer wg.Done()
if err := rm.podControl.DeletePod(rc.Namespace, filteredPods[ix].Name, rc); err != nil {
// Decrement the expected number of deletes because the informer won't observe this deletion
podKey := controller.PodKey(filteredPods[ix])
glog.V(2).Infof("Failed to delete %v due to %v, decrementing expectations for controller %q/%q", podKey, err, rc.Namespace, rc.Name)
rm.expectations.DeletionObserved(rcKey, podKey)
rm.enqueueController(rc)
utilruntime.HandleError(err)
}
}(i)
}
wg.Wait()
return nil
}
if diff > rm.burstReplicas {
diff = rm.burstReplicas
}
glog.V(2).Infof("Too many %q/%q replicas, need %d, deleting %d", rc.Namespace, rc.Name, rc.Spec.Replicas, diff)
// No need to sort pods if we are about to delete all of them
if rc.Spec.Replicas != 0 {
// Sort the pods in the order such that not-ready < ready, unscheduled
// < scheduled, and pending < running. This ensures that we delete pods
// in the earlier stages whenever possible.
sort.Sort(controller.ActivePods(filteredPods))
}
// Snapshot the UIDs (ns/name) of the pods we're expecting to see
// deleted, so we know to record their expectations exactly once either
// when we see it as an update of the deletion timestamp, or as a delete.
// Note that if the labels on a pod/rc change in a way that the pod gets
// orphaned, the rs will only wake up after the expectations have
// expired even if other pods are deleted.
deletedPodKeys := []string{}
for i := 0; i < diff; i++ {
deletedPodKeys = append(deletedPodKeys, controller.PodKey(filteredPods[i]))
}
// We use pod namespace/name as a UID to wait for deletions, so if the
// labels on a pod/rc change in a way that the pod gets orphaned, the
// rc will only wake up after the expectation has expired.
errCh := make(chan error, diff)
rm.expectations.ExpectDeletions(rcKey, deletedPodKeys)
var wg sync.WaitGroup
wg.Add(diff)
for i := 0; i < diff; i++ {
go func(ix int) {
defer wg.Done()
if err := rm.podControl.DeletePod(rc.Namespace, filteredPods[ix].Name, rc); err != nil {
// Decrement the expected number of deletes because the informer won't observe this deletion
podKey := controller.PodKey(filteredPods[ix])
glog.V(2).Infof("Failed to delete %v due to %v, decrementing expectations for controller %q/%q", podKey, err, rc.Namespace, rc.Name)
rm.expectations.DeletionObserved(rcKey, podKey)
errCh <- err
utilruntime.HandleError(err)
}
}(i)
}
wg.Wait()
select {
case err := <-errCh:
// all errors have been reported before and they're likely to be the same, so we'll only return the first one we hit.
if err != nil {
return err
}
default:
}
return nil
}
// syncReplicationController will sync the rc with the given key if it has had its expectations fulfilled, meaning
@ -600,8 +633,6 @@ func (rm *ReplicationManager) syncReplicationController(key string) error {
return nil
}
if err != nil {
glog.Infof("Unable to retrieve rc %v from store: %v", key, err)
rm.queue.Add(key)
return err
}
rc := *obj.(*api.ReplicationController)
@ -672,8 +703,9 @@ func (rm *ReplicationManager) syncReplicationController(key string) error {
filteredPods = controller.FilterActivePods(pods)
}
var manageReplicasErr error
if rcNeedsSync && rc.DeletionTimestamp == nil {
rm.manageReplicas(filteredPods, &rc)
manageReplicasErr = rm.manageReplicas(filteredPods, &rc)
}
trace.Step("manageReplicas done")
@ -692,10 +724,9 @@ func (rm *ReplicationManager) syncReplicationController(key string) error {
// Always updates status as pods come up or die.
if err := updateReplicaCount(rm.kubeClient.Core().ReplicationControllers(rc.Namespace), rc, len(filteredPods), fullyLabeledReplicasCount); err != nil {
// Multiple things could lead to this update failing. Requeuing the controller ensures
// we retry with some fairness.
glog.V(2).Infof("Failed to update replica count for controller %v/%v; requeuing; error: %v", rc.Namespace, rc.Name, err)
rm.enqueueController(&rc)
// Multiple things could lead to this update failing. Returning an error causes a requeue without forcing a hotloop
return err
}
return nil
return manageReplicasErr
}

View File

@ -32,6 +32,7 @@ import (
"k8s.io/kubernetes/pkg/client/cache"
clientset "k8s.io/kubernetes/pkg/client/clientset_generated/internalclientset"
"k8s.io/kubernetes/pkg/client/clientset_generated/internalclientset/fake"
fakeclientset "k8s.io/kubernetes/pkg/client/clientset_generated/internalclientset/fake"
"k8s.io/kubernetes/pkg/client/restclient"
"k8s.io/kubernetes/pkg/client/testing/core"
"k8s.io/kubernetes/pkg/controller"
@ -606,23 +607,11 @@ func TestControllerUpdateRequeue(t *testing.T) {
fakePodControl := controller.FakePodControl{}
manager.podControl = &fakePodControl
manager.syncReplicationController(getKey(rc, t))
ch := make(chan interface{})
go func() {
item, _ := manager.queue.Get()
ch <- item
}()
select {
case key := <-ch:
expectedKey := getKey(rc, t)
if key != expectedKey {
t.Errorf("Expected requeue of controller with key %s got %s", expectedKey, key)
}
case <-time.After(wait.ForeverTestTimeout):
manager.queue.ShutDown()
t.Errorf("Expected to find an rc in the queue, found none.")
// an error from the sync function will be requeued, check to make sure we returned an error
if err := manager.syncReplicationController(getKey(rc, t)); err == nil {
t.Errorf("missing error for requeue")
}
// 1 Update and 1 GET, both of which fail
fakeHandler.ValidateRequestCount(t, 2)
}
@ -1136,8 +1125,8 @@ func BenchmarkGetPodControllerSingleNS(b *testing.B) {
}
// setupManagerWithGCEnabled creates a RC manager with a fakePodControl and with garbageCollectorEnabled set to true
func setupManagerWithGCEnabled() (manager *ReplicationManager, fakePodControl *controller.FakePodControl) {
c := clientset.NewForConfigOrDie(&restclient.Config{Host: "", ContentConfig: restclient.ContentConfig{GroupVersion: testapi.Default.GroupVersion()}})
func setupManagerWithGCEnabled(objs ...runtime.Object) (manager *ReplicationManager, fakePodControl *controller.FakePodControl) {
c := fakeclientset.NewSimpleClientset(objs...)
fakePodControl = &controller.FakePodControl{}
manager = NewReplicationManagerFromClient(c, controller.NoResyncPeriodFunc, BurstReplicas, 0)
manager.garbageCollectorEnabled = true
@ -1165,8 +1154,8 @@ func TestDoNotPatchPodWithOtherControlRef(t *testing.T) {
}
func TestPatchPodWithOtherOwnerRef(t *testing.T) {
manager, fakePodControl := setupManagerWithGCEnabled()
rc := newReplicationController(2)
manager, fakePodControl := setupManagerWithGCEnabled(rc)
manager.rcStore.Indexer.Add(rc)
// add to podStore one more matching pod that doesn't have a controller
// ref, but has an owner ref pointing to other object. Expect a patch to
@ -1185,8 +1174,8 @@ func TestPatchPodWithOtherOwnerRef(t *testing.T) {
}
func TestPatchPodWithCorrectOwnerRef(t *testing.T) {
manager, fakePodControl := setupManagerWithGCEnabled()
rc := newReplicationController(2)
manager, fakePodControl := setupManagerWithGCEnabled(rc)
manager.rcStore.Indexer.Add(rc)
// add to podStore a matching pod that has an ownerRef pointing to the rc,
// but ownerRef.Controller is false. Expect a patch to take control it.
@ -1204,8 +1193,8 @@ func TestPatchPodWithCorrectOwnerRef(t *testing.T) {
}
func TestPatchPodFails(t *testing.T) {
manager, fakePodControl := setupManagerWithGCEnabled()
rc := newReplicationController(2)
manager, fakePodControl := setupManagerWithGCEnabled(rc)
manager.rcStore.Indexer.Add(rc)
// add to podStore two matching pods. Expect two patches to take control
// them.
@ -1215,16 +1204,16 @@ func TestPatchPodFails(t *testing.T) {
// control of the pods and create new ones.
fakePodControl.Err = fmt.Errorf("Fake Error")
err := manager.syncReplicationController(getKey(rc, t))
if err != nil {
t.Fatal(err)
if err == nil || err.Error() != "Fake Error" {
t.Fatalf("expected Fake Error, got %v", err)
}
// 2 patches to take control of pod1 and pod2 (both fail), 2 creates.
validateSyncReplication(t, fakePodControl, 2, 0, 2)
}
func TestPatchExtraPodsThenDelete(t *testing.T) {
manager, fakePodControl := setupManagerWithGCEnabled()
rc := newReplicationController(2)
manager, fakePodControl := setupManagerWithGCEnabled(rc)
manager.rcStore.Indexer.Add(rc)
// add to podStore three matching pods. Expect three patches to take control
// them, and later delete one of them.