Fix exclusive CPU allocations being deleted at container restart

The expectation is that exclusive CPU allocations happen at pod
creation time. When a container restarts, it should not have its
exclusive CPU allocations removed, and it should not need to
re-allocate CPUs.

There are a few places in the current code that look for containers
that have exited and call CpuManager.RemoveContainer() to clean up
the container.  This will end up deleting any exclusive CPU
allocations for that container, and if the container restarts within
the same pod it will end up using the default cpuset rather than
what should be exclusive CPUs.

Removing those calls and adding resource cleanup at allocation
time should get rid of the problem.

Signed-off-by: Chris Friesen <chris.friesen@windriver.com>

pkg/kubelet/cm/cpumanager/cpu_manager.go

@@ -212,6 +212,9 @@ func (m *manager) Start(activePods ActivePodsFunc, sourcesReady config.SourcesRe
 }
 
 func (m *manager) Allocate(p *v1.Pod, c *v1.Container) error {
+	// Garbage collect any stranded resources before allocating CPUs.
+	m.removeStaleState()
+
 	m.Lock()
 	defer m.Unlock()
 
@@ -384,18 +387,14 @@ func (m *manager) reconcileState() (success []reconciledContainer, failure []rec
 			}
 
 			if cstatus.State.Terminated != nil {
-				// Since the container is terminated, we know it is safe to
+				// The container is terminated but we can't call m.RemoveContainer()
-				// remove it without any reconciliation. Removing the container
+				// here because it could remove the allocated cpuset for the container
-				// will also remove it from the `containerMap` so that this
+				// which may be in the process of being restarted.  That would result
-				// container will be skipped next time around the loop.
+				// in the container losing any exclusively-allocated CPUs that it
+				// was allocated.
 				_, _, err := m.containerMap.GetContainerRef(containerID)
 				if err == nil {
-					klog.Warningf("[cpumanager] reconcileState: skipping container; already terminated (pod: %s, container id: %s)", pod.Name, containerID)
+					klog.Warningf("[cpumanager] reconcileState: ignoring terminated container (pod: %s, container id: %s)", pod.Name, containerID)
-					err := m.RemoveContainer(containerID)
-					if err != nil {
-						klog.Errorf("[cpumanager] reconcileState: failed to remove container (pod: %s, container id: %s, error: %v)", pod.Name, containerID, err)
-						failure = append(failure, reconciledContainer{pod.Name, container.Name, containerID})
-					}
 				}
 				continue
 			}

pkg/kubelet/cm/cpumanager/cpu_manager_test.go

@@ -269,12 +269,14 @@ func TestCPUManagerAdd(t *testing.T) {
 				err: testCase.updateErr,
 			},
 			containerMap:      containermap.NewContainerMap(),
-			activePods:        func() []*v1.Pod { return nil },
 			podStatusProvider: mockPodStatusProvider{},
+			sourcesReady:      &sourcesReadyStub{},
 		}
 
 		pod := makePod("fakePod", "fakeContainer", "2", "2")
 		container := &pod.Spec.Containers[0]
+		mgr.activePods = func() []*v1.Pod { return []*v1.Pod{pod} }
+
 		err := mgr.Allocate(pod, container)
 		if !reflect.DeepEqual(err, testCase.expAllocateErr) {
 			t.Errorf("CPU Manager Allocate() error (%v). expected error: %v but got: %v",
@@ -487,8 +489,11 @@ func TestCPUManagerAddWithInitContainers(t *testing.T) {
 			state:             state,
 			containerRuntime:  mockRuntimeService{},
 			containerMap:      containermap.NewContainerMap(),
-			activePods:        func() []*v1.Pod { return nil },
 			podStatusProvider: mockPodStatusProvider{},
+			sourcesReady:      &sourcesReadyStub{},
+			activePods: func() []*v1.Pod {
+				return []*v1.Pod{testCase.pod}
+			},
 		}
 
 		containers := append(
@@ -1021,12 +1026,14 @@ func TestCPUManagerAddWithResvList(t *testing.T) {
 				err: testCase.updateErr,
 			},
 			containerMap:      containermap.NewContainerMap(),
-			activePods:        func() []*v1.Pod { return nil },
 			podStatusProvider: mockPodStatusProvider{},
+			sourcesReady:      &sourcesReadyStub{},
 		}
 
 		pod := makePod("fakePod", "fakeContainer", "2", "2")
 		container := &pod.Spec.Containers[0]
+		mgr.activePods = func() []*v1.Pod { return []*v1.Pod{pod} }
+
 		err := mgr.Allocate(pod, container)
 		if !reflect.DeepEqual(err, testCase.expAllocateErr) {
 			t.Errorf("CPU Manager Allocate() error (%v). expected error: %v but got: %v",

pkg/kubelet/cm/internal_container_lifecycle.go

@@ -54,19 +54,10 @@ func (i *internalContainerLifecycleImpl) PreStartContainer(pod *v1.Pod, containe
 }
 
 func (i *internalContainerLifecycleImpl) PreStopContainer(containerID string) error {
-	if i.cpuManager != nil {
-		return i.cpuManager.RemoveContainer(containerID)
-	}
 	return nil
 }
 
 func (i *internalContainerLifecycleImpl) PostStopContainer(containerID string) error {
-	if i.cpuManager != nil {
-		err := i.cpuManager.RemoveContainer(containerID)
-		if err != nil {
-			return err
-		}
-	}
 	if utilfeature.DefaultFeatureGate.Enabled(kubefeatures.TopologyManager) {
 		err := i.topologyManager.RemoveContainer(containerID)
 		if err != nil {