support fix containerrecreaterequest

apis/apps/v1alpha1/containerrecreaterequest_types.go

@@ -37,6 +37,9 @@ const (
 	// ContainerRecreateRequestUnreadyAcquiredKey indicates the Pod has been forced to not-ready.
 	// It is required if the unreadyGracePeriodSeconds is set in ContainerRecreateRequests.
 	ContainerRecreateRequestUnreadyAcquiredKey = "crr.apps.kruise.io/unready-acquired"
+
+	// It is a status key for containers which are be killed by crr
+	ContainerRecreateRequestForceKilledContainerStatusesKey = "crr.apps.kruise.io/force-killed-container-statuses"
 )
 
 // ContainerRecreateRequestSpec defines the desired state of ContainerRecreateRequest
@@ -106,6 +109,8 @@ type ContainerRecreateRequestStrategy struct {
 	FailurePolicy ContainerRecreateRequestFailurePolicyType `json:"failurePolicy,omitempty"`
 	// OrderedRecreate indicates whether to recreate the next container only if the previous one has recreated completely.
 	OrderedRecreate bool `json:"orderedRecreate,omitempty"`
+	// ForceRecreate indicates whether to force kill the container even if the previous container is not running.
+	ForceRecreate bool `json:"forceRecreate,omitempty"`
 	// TerminationGracePeriodSeconds is the optional duration in seconds to wait the container terminating gracefully.
 	// Value must be non-negative integer. The value zero indicates delete immediately.
 	// If this value is nil, we will use pod.Spec.TerminationGracePeriodSeconds as default value.
@@ -158,6 +163,12 @@ type ContainerRecreateRequestContainerRecreateState struct {
 	Phase ContainerRecreateRequestPhase `json:"phase"`
 	// A human readable message indicating details about this state.
 	Message string `json:"message,omitempty"`
+	// Container should be killed by force
+	ShouldBeKilledByForce bool `json:"shouldBeKilledByForce"`
+	// Container has been killed by force
+	IsForceKilled bool `json:"isForceKilled"`
+	// Container's ID in the format 'docker://<container_id>'.
+	ContainerID string `json:"containerID,omitempty"`
 }
 
 // ContainerRecreateRequestSyncContainerStatus only uses in the annotation `crr.apps.kruise.io/sync-container-statuses`.
@@ -172,6 +183,16 @@ type ContainerRecreateRequestSyncContainerStatus struct {
 	ContainerID string `json:"containerID,omitempty"`
 }
 
+// ContainerRecreateRequestKilledContainerStatus contains the state of the last killed container.
+type ContainerRecreateRequestForceKilledContainerStatus struct {
+	// Name of the container.
+	Name string `json:"name"`
+	// Container is killed by force
+	IsForceKilled bool `json:"restartCount"`
+	// Container's ID in the format 'docker://<container_id>'.
+	ContainerID string `json:"containerID,omitempty"`
+}
+
 // +genclient
 // +k8s:openapi-gen=true
 // +kubebuilder:object:root=true

pkg/daemon/containerrecreate/crr_daemon_controller.go

@@ -362,6 +362,11 @@ func (c *Controller) manage(crr *appsv1alpha1.ContainerRecreateRequest) error {
 			}
 			return c.patchCRRContainerRecreateStates(crr, newCRRContainerRecreateStates)
 		}
+
+		if crr.Spec.Strategy.ForceRecreate && state.ShouldBeKilledByForce {
+			state.IsForceKilled = true
+			state.ContainerID = kubeContainerStatus.ID.String()
+		}
 		state.Phase = appsv1alpha1.ContainerRecreateRequestRecreating
 		break
 	}

pkg/daemon/containerrecreate/crr_daemon_util.go

@@ -69,6 +69,8 @@ func getCurrentCRRContainersRecreateStates(
 	}
 
 	syncContainerStatuses := getCRRSyncContainerStatuses(crr)
+	// the statuses of force killed containers for this crr resource
+	forceKillContainerStatuses := getCRRForceKillContainersStatuses(crr)
 	var statuses []appsv1alpha1.ContainerRecreateRequestContainerRecreateState
 
 	for i := range crr.Spec.Containers {
@@ -84,6 +86,8 @@ func getCurrentCRRContainersRecreateStates(
 		syncContainerStatus := syncContainerStatuses[c.Name]
 		kubeContainerStatus := podStatus.FindContainerStatusByName(c.Name)
 
+		killedContainerStatus := forceKillContainerStatuses[c.Name]
+
 		var currentState appsv1alpha1.ContainerRecreateRequestContainerRecreateState
 		if kubeContainerStatus == nil {
 			// not found the real container
@@ -114,7 +118,11 @@ func getCurrentCRRContainersRecreateStates(
 				syncContainerStatus.Ready {
 				currentState.Phase = appsv1alpha1.ContainerRecreateRequestSucceeded
 			}
-
+			if crr.Spec.Strategy.ForceRecreate && (len(forceKillContainerStatuses) == 0 ||
+				containerIsNotForceKilled(killedContainerStatus)) {
+				currentState.ShouldBeKilledByForce = true
+				currentState.Phase = appsv1alpha1.ContainerRecreateRequestPending
+			}
 		} else {
 			currentState = appsv1alpha1.ContainerRecreateRequestContainerRecreateState{
 				Name:  c.Name,
@@ -128,6 +136,16 @@ func getCurrentCRRContainersRecreateStates(
 	return statuses
 }
 
+func containerIsNotForceKilled(recreateState *appsv1alpha1.ContainerRecreateRequestContainerRecreateState) bool {
+	if recreateState == nil {
+		return true
+	}
+	if !recreateState.IsForceKilled {
+		return true
+	}
+	return false
+}
+
 func getCRRContainerRecreateState(crr *appsv1alpha1.ContainerRecreateRequest, name string) *appsv1alpha1.ContainerRecreateRequestContainerRecreateState {
 	for i := range crr.Status.ContainerRecreateStates {
 		c := &crr.Status.ContainerRecreateStates[i]
@@ -157,6 +175,19 @@ func getCRRSyncContainerStatuses(crr *appsv1alpha1.ContainerRecreateRequest) map
 	return statuses
 }
 
+func getCRRForceKillContainersStatuses(crr *appsv1alpha1.ContainerRecreateRequest) map[string]*appsv1alpha1.ContainerRecreateRequestContainerRecreateState {
+	crrContainerRecreateStatus := crr.Status.ContainerRecreateStates
+	forceKillStatuses := make(map[string]*appsv1alpha1.ContainerRecreateRequestContainerRecreateState, len(crrContainerRecreateStatus))
+	for i := range crrContainerRecreateStatus {
+		c := &crrContainerRecreateStatus[i]
+		if !c.IsForceKilled {
+			continue
+		}
+		forceKillStatuses[c.Name] = c
+	}
+	return forceKillStatuses
+}
+
 func convertCRRToPod(crr *appsv1alpha1.ContainerRecreateRequest) *v1.Pod {
 	podName := crr.Spec.PodName
 	podUID := types.UID(crr.Labels[appsv1alpha1.ContainerRecreateRequestPodUIDKey])