Ensure deletion of pods in queues and cache

When the client misses a delete event from the watcher, it will use the last state of the pod in the informer cache to produce a delete event. At that point, it's not clear if the pod was in the queues or the cache, so we should issue a deletion in both.

The pod could be assumed, so deletion of assumed pods from the cache should work.

Change-Id: I11ce9785de603924fc121fe2fa6ed5cb1e16922f

pkg/scheduler/eventhandlers.go

@@ -280,8 +280,10 @@ func addAllEventHandlers(
 				case *v1.Pod:
 					return assignedPod(t)
 				case cache.DeletedFinalStateUnknown:
-					if pod, ok := t.Obj.(*v1.Pod); ok {
-						return assignedPod(pod)
+					if _, ok := t.Obj.(*v1.Pod); ok {
+						// The carried object may be stale, so we don't use it to check if
+						// it's assigned or not. Attempting to cleanup anyways.
+						return true
 					}
 					utilruntime.HandleError(fmt.Errorf("unable to convert object %T to *v1.Pod in %T", obj, sched))
 					return false
@@ -306,7 +308,9 @@ func addAllEventHandlers(
 					return !assignedPod(t) && responsibleForPod(t, sched.Profiles)
 				case cache.DeletedFinalStateUnknown:
 					if pod, ok := t.Obj.(*v1.Pod); ok {
-						return !assignedPod(pod) && responsibleForPod(pod, sched.Profiles)
+						// The carried object may be stale, so we don't use it to check if
+						// it's assigned or not.
+						return responsibleForPod(pod, sched.Profiles)
 					}
 					utilruntime.HandleError(fmt.Errorf("unable to convert object %T to *v1.Pod in %T", obj, sched))
 					return false

pkg/scheduler/internal/cache/cache.go

@@ -537,22 +537,19 @@ func (cache *schedulerCache) RemovePod(pod *v1.Pod) error {
 
 	currState, ok := cache.podStates[key]
 	switch {
-	// An assumed pod won't have Delete/Remove event. It needs to have Add event
-	// before Remove event, in which case the state would change from Assumed to Added.
-	case ok && !cache.assumedPods.Has(key):
+	case ok:
 		if currState.pod.Spec.NodeName != pod.Spec.NodeName {
 			klog.Errorf("Pod %v was assumed to be on %v but got added to %v", key, pod.Spec.NodeName, currState.pod.Spec.NodeName)
-			klog.Fatalf("Schedulercache is corrupted and can badly affect scheduling decisions")
-		}
-		err := cache.removePod(currState.pod)
-		if err != nil {
-			return err
+			if pod.Spec.NodeName != "" {
+				// An empty NodeName is possible when the scheduler misses a Delete
+				// event and it gets the last known state from the informer cache.
+				klog.Fatalf("Schedulercache is corrupted and can badly affect scheduling decisions")
+			}
 		}
-		delete(cache.podStates, key)
+		return cache.expirePod(key, currState)
 	default:
 		return fmt.Errorf("pod %v is not found in scheduler cache, so cannot be removed from it", key)
 	}
-	return nil
 }
 
 func (cache *schedulerCache) IsAssumedPod(pod *v1.Pod) (bool, error) {

pkg/scheduler/internal/cache/cache_test.go

@@ -807,57 +807,62 @@ func TestEphemeralStorageResource(t *testing.T) {
 
 // TestRemovePod tests after added pod is removed, its information should also be subtracted.
 func TestRemovePod(t *testing.T) {
-	basePod := makeBasePod(t, "node-1", "test", "100m", "500", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 80, Protocol: "TCP"}})
-	tests := []struct {
-		nodes     []*v1.Node
-		pod       *v1.Pod
-		wNodeInfo *framework.NodeInfo
-	}{{
-		nodes: []*v1.Node{
-			{
-				ObjectMeta: metav1.ObjectMeta{Name: "node-1"},
-			},
-			{
-				ObjectMeta: metav1.ObjectMeta{Name: "node-2"},
-			},
+	pod := makeBasePod(t, "node-1", "test", "100m", "500", "", []v1.ContainerPort{{HostIP: "127.0.0.1", HostPort: 80, Protocol: "TCP"}})
+	nodes := []*v1.Node{
+		{
+			ObjectMeta: metav1.ObjectMeta{Name: "node-1"},
 		},
-		pod: basePod,
-		wNodeInfo: newNodeInfo(
-			&framework.Resource{
-				MilliCPU: 100,
-				Memory:   500,
-			},
-			&framework.Resource{
-				MilliCPU: 100,
-				Memory:   500,
-			},
-			[]*v1.Pod{basePod},
-			newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
-			make(map[string]*framework.ImageStateSummary),
-		),
-	}}
+		{
+			ObjectMeta: metav1.ObjectMeta{Name: "node-2"},
+		},
+	}
+	wNodeInfo := newNodeInfo(
+		&framework.Resource{
+			MilliCPU: 100,
+			Memory:   500,
+		},
+		&framework.Resource{
+			MilliCPU: 100,
+			Memory:   500,
+		},
+		[]*v1.Pod{pod},
+		newHostPortInfoBuilder().add("TCP", "127.0.0.1", 80).build(),
+		make(map[string]*framework.ImageStateSummary),
+	)
+	tests := map[string]struct {
+		assume bool
+	}{
+		"bound":   {},
+		"assumed": {assume: true},
+	}
 
-	for i, tt := range tests {
-		t.Run(fmt.Sprintf("case_%d", i), func(t *testing.T) {
-			nodeName := tt.pod.Spec.NodeName
+	for name, tt := range tests {
+		t.Run(name, func(t *testing.T) {
+			nodeName := pod.Spec.NodeName
 			cache := newSchedulerCache(time.Second, time.Second, nil)
-			// Add pod succeeds even before adding the nodes.
-			if err := cache.AddPod(tt.pod); err != nil {
-				t.Fatalf("AddPod failed: %v", err)
+			// Add/Assume pod succeeds even before adding the nodes.
+			if tt.assume {
+				if err := cache.AddPod(pod); err != nil {
+					t.Fatalf("AddPod failed: %v", err)
+				}
+			} else {
+				if err := cache.AssumePod(pod); err != nil {
+					t.Fatalf("AssumePod failed: %v", err)
+				}
 			}
 			n := cache.nodes[nodeName]
-			if err := deepEqualWithoutGeneration(n, tt.wNodeInfo); err != nil {
+			if err := deepEqualWithoutGeneration(n, wNodeInfo); err != nil {
 				t.Error(err)
 			}
-			for _, n := range tt.nodes {
+			for _, n := range nodes {
 				cache.AddNode(n)
 			}
 
-			if err := cache.RemovePod(tt.pod); err != nil {
+			if err := cache.RemovePod(pod); err != nil {
 				t.Fatalf("RemovePod failed: %v", err)
 			}
 
-			if _, err := cache.GetPod(tt.pod); err == nil {
+			if _, err := cache.GetPod(pod); err == nil {
 				t.Errorf("pod was not deleted")
 			}