/* Copyright 2023 The Kruise Authors. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. */ package sidecarterminator import ( "context" "encoding/json" "flag" "fmt" "strings" "time" corev1 "k8s.io/api/core/v1" "k8s.io/apimachinery/pkg/api/errors" metav1 "k8s.io/apimachinery/pkg/apis/meta/v1" "k8s.io/apimachinery/pkg/runtime" "k8s.io/apimachinery/pkg/types" "k8s.io/apimachinery/pkg/util/sets" "k8s.io/client-go/tools/record" "k8s.io/klog/v2" "sigs.k8s.io/controller-runtime/pkg/client" "sigs.k8s.io/controller-runtime/pkg/controller" "sigs.k8s.io/controller-runtime/pkg/manager" "sigs.k8s.io/controller-runtime/pkg/reconcile" "sigs.k8s.io/controller-runtime/pkg/source" appsv1alpha1 "github.com/openkruise/kruise/apis/apps/v1alpha1" "github.com/openkruise/kruise/pkg/features" "github.com/openkruise/kruise/pkg/util" utilclient "github.com/openkruise/kruise/pkg/util/client" utilfeature "github.com/openkruise/kruise/pkg/util/feature" "github.com/openkruise/kruise/pkg/util/ratelimiter" ) func init() { flag.IntVar(&concurrentReconciles, "sidecarterminator-workers", concurrentReconciles, "Max concurrent workers for SidecarTerminator controller.") } var ( concurrentReconciles = 3 SidecarTerminated corev1.PodConditionType = "SidecarTerminated" ) /** * USER ACTION REQUIRED: This is a scaffold file intended for the user to modify with their own Controller * business logic. Delete these comments after modifying this file.* */ // Add creates a new SidecarTerminator Controller and adds it to the Manager with default RBAC. The Manager will set fields on the Controller // and Start it when the Manager is Started. func Add(mgr manager.Manager) error { if !utilfeature.DefaultFeatureGate.Enabled(features.SidecarTerminator) { return nil } return add(mgr, newReconciler(mgr)) } // newReconciler returns a new reconcile.Reconciler func newReconciler(mgr manager.Manager) reconcile.Reconciler { cli := utilclient.NewClientFromManager(mgr, "sidecarterminator-controller") recorder := mgr.GetEventRecorderFor("sidecarterminator-controller") return &ReconcileSidecarTerminator{ Client: cli, recorder: recorder, scheme: mgr.GetScheme(), } } // add adds a new Controller to mgr with r as the reconcile.Reconciler func add(mgr manager.Manager, r reconcile.Reconciler) error { // Create a new controller c, err := controller.New("sidecarterminator-controller", mgr, controller.Options{ Reconciler: r, MaxConcurrentReconciles: concurrentReconciles, CacheSyncTimeout: util.GetControllerCacheSyncTimeout(), RateLimiter: ratelimiter.DefaultControllerRateLimiter()}) if err != nil { return err } // Watch for changes to Pod if err = c.Watch(source.Kind(mgr.GetCache(), &corev1.Pod{}), &enqueueRequestForPod{}); err != nil { return err } return nil } var _ reconcile.Reconciler = &ReconcileSidecarTerminator{} // ReconcileSidecarTerminator reconciles a SidecarTerminator object type ReconcileSidecarTerminator struct { client.Client recorder record.EventRecorder scheme *runtime.Scheme } // Reconcile get the pod whose sidecar containers should be stopped, and stop them. func (r *ReconcileSidecarTerminator) Reconcile(_ context.Context, request reconcile.Request) (reconcile.Result, error) { pod := &corev1.Pod{} err := r.Get(context.TODO(), request.NamespacedName, pod) if err != nil { if errors.IsNotFound(err) { // Object not found, return. Created objects are automatically garbage collected. // For additional cleanup logic use finalizers. return reconcile.Result{}, nil } // Error reading the object - requeue the request. return reconcile.Result{}, err } start := time.Now() klog.V(3).InfoS("SidecarTerminator -- begin to process pod for reconcile", "pod", request) defer func() { klog.V(3).InfoS("SidecarTerminator -- process pod done", "pod", request, "cost", time.Since(start)) }() return r.doReconcile(pod) } func (r *ReconcileSidecarTerminator) doReconcile(pod *corev1.Pod) (reconcile.Result, error) { if !isInterestingPod(pod) { return reconcile.Result{}, nil } sidecarNeedToExecuteKillContainer, sidecarNeedToExecuteInPlaceUpdate, err := r.groupSidecars(pod) if err != nil { return reconcile.Result{}, err } if err := r.executeInPlaceUpdateAction(pod, sidecarNeedToExecuteInPlaceUpdate); err != nil { return reconcile.Result{}, err } if err := r.markJobPodTerminated(pod); err != nil { return reconcile.Result{}, err } if err := r.executeKillContainerAction(pod, sidecarNeedToExecuteKillContainer); err != nil { return reconcile.Result{}, err } return reconcile.Result{}, nil } // markJobPodTerminated terminate the job pod and skip the state of the sidecar containers // This method should only be called before the executeKillContainerAction func (r *ReconcileSidecarTerminator) markJobPodTerminated(pod *corev1.Pod) error { if pod.Status.Phase == corev1.PodFailed || pod.Status.Phase == corev1.PodSucceeded { return nil } // after the pod is terminated by the sidecar terminator, kubelet will kill the containers that are not in the terminal phase // 1. sidecar container terminate with non-zero exit code // 2. sidecar container is not in a terminal phase (still running or waiting) klog.V(3).InfoS("All of the main containers are completed, will terminate the job pod", "pod", klog.KObj(pod)) // terminate the pod, ignore the status of the sidecar containers. // in kubelet,pods are not allowed to transition out of terminal phases. // patch pod condition status := corev1.PodStatus{ Conditions: []corev1.PodCondition{ { Type: SidecarTerminated, Status: corev1.ConditionTrue, LastTransitionTime: metav1.Now(), Message: "Terminated by Sidecar Terminator", }, }, } // patch pod phase if containersSucceeded(pod, getMain(pod)) { status.Phase = corev1.PodSucceeded } else { status.Phase = corev1.PodFailed } klog.V(3).InfoS("Terminated the job pod", "pod", klog.KObj(pod), "phase", status.Phase) by, _ := json.Marshal(status) patchCondition := fmt.Sprintf(`{"status":%s}`, string(by)) rcvObject := &corev1.Pod{ObjectMeta: metav1.ObjectMeta{Namespace: pod.Namespace, Name: pod.Name}} if err := r.Status().Patch(context.TODO(), rcvObject, client.RawPatch(types.StrategicMergePatchType, []byte(patchCondition))); err != nil { return fmt.Errorf("failed to patch pod status: %v", err) } return nil } func (r *ReconcileSidecarTerminator) groupSidecars(pod *corev1.Pod) (sets.String, sets.String, error) { runningOnVK, err := IsPodRunningOnVirtualKubelet(pod, r.Client) if err != nil { return nil, nil, client.IgnoreNotFound(err) } inPlaceUpdate := sets.NewString() killContainer := sets.NewString() for i := range pod.Spec.Containers { container := &pod.Spec.Containers[i] for j := range container.Env { if !runningOnVK && container.Env[j].Name == appsv1alpha1.KruiseTerminateSidecarEnv && strings.EqualFold(container.Env[j].Value, "true") { killContainer.Insert(container.Name) break } if container.Env[j].Name == appsv1alpha1.KruiseTerminateSidecarWithImageEnv && container.Env[j].Value != "" { inPlaceUpdate.Insert(container.Name) } } } return killContainer, inPlaceUpdate, nil } func containersCompleted(pod *corev1.Pod, containers sets.String) bool { if len(pod.Spec.Containers) != len(pod.Status.ContainerStatuses) { return false } for i := range pod.Status.ContainerStatuses { status := &pod.Status.ContainerStatuses[i] if containers.Has(status.Name) && status.State.Terminated == nil { return false } } return true } func containersSucceeded(pod *corev1.Pod, containers sets.String) bool { if len(pod.Spec.Containers) != len(pod.Status.ContainerStatuses) { return false } for i := range pod.Status.ContainerStatuses { status := &pod.Status.ContainerStatuses[i] if containers.Has(status.Name) && (status.State.Terminated == nil || status.State.Terminated.ExitCode != int32(0)) { return false } } return true }