// createPodSandbox creates a pod sandbox and returns (podSandBoxID, message, error).
func (m *kubeGenericRuntimeManager) createPodSandbox(pod *v1.Pod, attempt uint32) (string, string, error) {
podSandboxConfig, err := m.generatePodSandboxConfig(pod, attempt)
if err != nil {
message := fmt.Sprintf("GeneratePodSandboxConfig for pod %q failed: %v", format.Pod(pod), err)
glog.Error(message)
return "", message, err
}
// Create pod logs directory
err = m.osInterface.MkdirAll(podSandboxConfig.GetLogDirectory(), 0755)
if err != nil {
message := fmt.Sprintf("Create pod log directory for pod %q failed: %v", format.Pod(pod), err)
glog.Errorf(message)
return "", message, err
}
podSandBoxID, err := m.runtimeService.RunPodSandbox(podSandboxConfig)
if err != nil {
message := fmt.Sprintf("CreatePodSandbox for pod %q failed: %v", format.Pod(pod), err)
glog.Error(message)
return "", message, err
}
return podSandBoxID, "", nil
}
// runOnce runs a given set of pods and returns their status.
func (kl *Kubelet) runOnce(pods []*api.Pod, retryDelay time.Duration) (results []RunPodResult, err error) {
ch := make(chan RunPodResult)
admitted := []*api.Pod{}
for _, pod := range pods {
// Check if we can admit the pod.
if ok, reason, message := kl.canAdmitPod(append(admitted, pod), pod); !ok {
kl.rejectPod(pod, reason, message)
} else {
admitted = append(admitted, pod)
}
go func(pod *api.Pod) {
err := kl.runPod(pod, retryDelay)
ch <- RunPodResult{pod, err}
}(pod)
}
glog.Infof("waiting for %d pods", len(pods))
failedPods := []string{}
for i := 0; i < len(pods); i++ {
res := <-ch
results = append(results, res)
if res.Err != nil {
// TODO(proppy): report which containers failed the pod.
glog.Infof("failed to start pod %q: %v", format.Pod(res.Pod), res.Err)
failedPods = append(failedPods, format.Pod(res.Pod))
} else {
glog.Infof("started pod %q", format.Pod(res.Pod))
}
}
if len(failedPods) > 0 {
return results, fmt.Errorf("error running pods: %v", failedPods)
}
glog.Infof("%d pods started", len(pods))
return results, err
}
func (m *manager) AddPod(pod *v1.Pod) {
m.workerLock.Lock()
defer m.workerLock.Unlock()
key := probeKey{podUID: pod.UID}
for _, c := range pod.Spec.Containers {
key.containerName = c.Name
if c.ReadinessProbe != nil {
key.probeType = readiness
if _, ok := m.workers[key]; ok {
glog.Errorf("Readiness probe already exists! %v - %v",
format.Pod(pod), c.Name)
return
}
w := newWorker(m, readiness, pod, c)
m.workers[key] = w
go w.run()
}
if c.LivenessProbe != nil {
key.probeType = liveness
if _, ok := m.workers[key]; ok {
glog.Errorf("Liveness probe already exists! %v - %v",
format.Pod(pod), c.Name)
return
}
w := newWorker(m, liveness, pod, c)
m.workers[key] = w
go w.run()
}
}
}
// podSandboxChanged checks whether the spec of the pod is changed and returns
// (changed, new attempt, original sandboxID if exist).
func (m *kubeGenericRuntimeManager) podSandboxChanged(pod *api.Pod, podStatus *kubecontainer.PodStatus) (changed bool, attempt uint32, sandboxID string) {
if len(podStatus.SandboxStatuses) == 0 {
glog.V(2).Infof("No sandbox for pod %q can be found. Need to start a new one", format.Pod(pod))
return true, 0, ""
}
readySandboxCount := 0
for _, s := range podStatus.SandboxStatuses {
if s.GetState() == runtimeApi.PodSandBoxState_READY {
readySandboxCount++
}
}
// Needs to create a new sandbox when readySandboxCount > 1 or the ready sandbox is not the latest one.
sandboxStatus := podStatus.SandboxStatuses[0]
if readySandboxCount > 1 || sandboxStatus.GetState() != runtimeApi.PodSandBoxState_READY {
glog.V(2).Infof("No ready sandbox for pod %q can be found. Need to start a new one", format.Pod(pod))
return true, sandboxStatus.Metadata.GetAttempt() + 1, sandboxStatus.GetId()
}
// Needs to create a new sandbox when network namespace changed.
if sandboxStatus.Linux != nil && sandboxStatus.Linux.Namespaces.Options != nil &&
sandboxStatus.Linux.Namespaces.Options.GetHostNetwork() != kubecontainer.IsHostNetworkPod(pod) {
glog.V(2).Infof("Sandbox for pod %q has changed. Need to start a new one", format.Pod(pod))
return true, sandboxStatus.Metadata.GetAttempt() + 1, ""
}
return false, sandboxStatus.Metadata.GetAttempt(), sandboxStatus.GetId()
}
// updateStatusInternal updates the internal status cache, and queues an update to the api server if
// necessary. Returns whether an update was triggered.
// This method IS NOT THREAD SAFE and must be called from a locked function.
func (m *manager) updateStatusInternal(pod *api.Pod, status api.PodStatus, forceUpdate bool) bool {
var oldStatus api.PodStatus
cachedStatus, isCached := m.podStatuses[pod.UID]
if isCached {
oldStatus = cachedStatus.status
} else if mirrorPod, ok := m.podManager.GetMirrorPodByPod(pod); ok {
oldStatus = mirrorPod.Status
} else {
oldStatus = pod.Status
}
// Set ReadyCondition.LastTransitionTime.
if readyCondition := api.GetPodReadyCondition(status); readyCondition != nil {
// Need to set LastTransitionTime.
lastTransitionTime := unversioned.Now()
oldReadyCondition := api.GetPodReadyCondition(oldStatus)
if oldReadyCondition != nil && readyCondition.Status == oldReadyCondition.Status {
lastTransitionTime = oldReadyCondition.LastTransitionTime
}
readyCondition.LastTransitionTime = lastTransitionTime
}
// ensure that the start time does not change across updates.
if oldStatus.StartTime != nil && !oldStatus.StartTime.IsZero() {
status.StartTime = oldStatus.StartTime
} else if status.StartTime.IsZero() {
// if the status has no start time, we need to set an initial time
now := unversioned.Now()
status.StartTime = &now
}
normalizeStatus(&status)
// The intent here is to prevent concurrent updates to a pod's status from
// clobbering each other so the phase of a pod progresses monotonically.
if isCached && isStatusEqual(&cachedStatus.status, &status) && !forceUpdate {
glog.V(3).Infof("Ignoring same status for pod %q, status: %+v", format.Pod(pod), status)
return false // No new status.
}
newStatus := versionedPodStatus{
status: status,
version: cachedStatus.version + 1,
podName: pod.Name,
podNamespace: pod.Namespace,
}
m.podStatuses[pod.UID] = newStatus
select {
case m.podStatusChannel <- podStatusSyncRequest{pod.UID, newStatus}:
return true
default:
// Let the periodic syncBatch handle the update if the channel is full.
// We can't block, since we hold the mutex lock.
glog.V(4).Infof("Skpping the status update for pod %q for now because the channel is full; status: %+v",
format.Pod(pod), status)
return false
}
}
func (m *manager) SetContainerReadiness(pod *api.Pod, containerID kubecontainer.ContainerID, ready bool) {
m.podStatusesLock.Lock()
defer m.podStatusesLock.Unlock()
oldStatus, found := m.podStatuses[pod.UID]
if !found {
glog.Warningf("Container readiness changed before pod has synced: %q - %q",
format.Pod(pod), containerID.String())
return
}
// Find the container to update.
containerIndex := -1
for i, c := range oldStatus.status.ContainerStatuses {
if c.ContainerID == containerID.String() {
containerIndex = i
break
}
}
if containerIndex == -1 {
glog.Warningf("Container readiness changed for unknown container: %q - %q",
format.Pod(pod), containerID.String())
return
}
if oldStatus.status.ContainerStatuses[containerIndex].Ready == ready {
glog.V(4).Infof("Container readiness unchanged (%v): %q - %q", ready,
format.Pod(pod), containerID.String())
return
}
// Make sure we're not updating the cached version.
clone, err := api.Scheme.DeepCopy(&oldStatus.status)
if err != nil {
glog.Errorf("Failed to clone status %+v: %v", oldStatus.status, err)
return
}
status := *clone.(*api.PodStatus)
status.ContainerStatuses[containerIndex].Ready = ready
// Update pod condition.
readyConditionIndex := -1
for i, condition := range status.Conditions {
if condition.Type == api.PodReady {
readyConditionIndex = i
break
}
}
readyCondition := GeneratePodReadyCondition(&pod.Spec, status.ContainerStatuses, status.Phase)
if readyConditionIndex != -1 {
status.Conditions[readyConditionIndex] = readyCondition
} else {
glog.Warningf("PodStatus missing PodReady condition: %+v", status)
status.Conditions = append(status.Conditions, readyCondition)
}
m.updateStatusInternal(pod, status)
}
// Iterate through all pods in desired state of world, and remove if they no
// longer exist
func (dswp *desiredStateOfWorldPopulator) findAndRemoveDeletedPods() {
var runningPods []*kubecontainer.Pod
runningPodsFetched := false
for _, volumeToMount := range dswp.desiredStateOfWorld.GetVolumesToMount() {
if _, podExists :=
dswp.podManager.GetPodByUID(volumeToMount.Pod.UID); podExists {
continue
}
// Once a pod has been deleted from kubelet pod manager, do not delete
// it immediately from volume manager. Instead, check the kubelet
// containerRuntime to verify that all containers in the pod have been
// terminated.
if !runningPodsFetched {
var getPodsErr error
runningPods, getPodsErr = dswp.kubeContainerRuntime.GetPods(false)
if getPodsErr != nil {
glog.Errorf(
"kubeContainerRuntime.findAndRemoveDeletedPods returned error %v.",
getPodsErr)
continue
}
runningPodsFetched = true
dswp.timeOfLastGetPodStatus = time.Now()
}
runningContainers := false
for _, runningPod := range runningPods {
if runningPod.ID == volumeToMount.Pod.UID {
if len(runningPod.Containers) > 0 {
runningContainers = true
}
break
}
}
if runningContainers {
glog.V(5).Infof(
"Pod %q has been removed from pod manager. However, it still has one or more containers in the non-exited state. Therefore it will not be removed from volume manager.",
format.Pod(volumeToMount.Pod))
continue
}
glog.V(5).Infof(
"Removing volume %q (volSpec=%q) for pod %q from desired state.",
volumeToMount.VolumeName,
volumeToMount.VolumeSpec.Name(),
format.Pod(volumeToMount.Pod))
dswp.desiredStateOfWorld.DeletePodFromVolume(
volumeToMount.PodName, volumeToMount.VolumeName)
dswp.deleteProcessedPod(volumeToMount.PodName)
}
}
func (m *manager) SetContainerReadiness(podUID types.UID, containerID kubecontainer.ContainerID, ready bool) {
m.podStatusesLock.Lock()
defer m.podStatusesLock.Unlock()
pod, ok := m.podManager.GetPodByUID(podUID)
if !ok {
glog.V(4).Infof("Pod %q has been deleted, no need to update readiness", string(podUID))
return
}
oldStatus, found := m.podStatuses[pod.UID]
if !found {
glog.Warningf("Container readiness changed before pod has synced: %q - %q",
format.Pod(pod), containerID.String())
return
}
// Find the container to update.
containerStatus, _, ok := findContainerStatus(&oldStatus.status, containerID.String())
if !ok {
glog.Warningf("Container readiness changed for unknown container: %q - %q",
format.Pod(pod), containerID.String())
return
}
if containerStatus.Ready == ready {
glog.V(4).Infof("Container readiness unchanged (%v): %q - %q", ready,
format.Pod(pod), containerID.String())
return
}
// Make sure we're not updating the cached version.
status, err := copyStatus(&oldStatus.status)
if err != nil {
return
}
containerStatus, _, _ = findContainerStatus(&status, containerID.String())
containerStatus.Ready = ready
// Update pod condition.
readyConditionIndex := -1
for i, condition := range status.Conditions {
if condition.Type == api.PodReady {
readyConditionIndex = i
break
}
}
readyCondition := GeneratePodReadyCondition(&pod.Spec, status.ContainerStatuses, status.Phase)
if readyConditionIndex != -1 {
status.Conditions[readyConditionIndex] = readyCondition
} else {
glog.Warningf("PodStatus missing PodReady condition: %+v", status)
status.Conditions = append(status.Conditions, readyCondition)
}
m.updateStatusInternal(pod, status, false)
}
// syncPod syncs the given status with the API server. The caller must not hold the lock.
func (m *manager) syncPod(uid types.UID, status versionedPodStatus) {
if !m.needsUpdate(uid, status) {
glog.V(1).Infof("Status for pod %q is up-to-date; skipping", uid)
return
}
// TODO: make me easier to express from client code
pod, err := m.kubeClient.Core().Pods(status.podNamespace).Get(status.podName, metav1.GetOptions{})
if errors.IsNotFound(err) {
glog.V(3).Infof("Pod %q (%s) does not exist on the server", status.podName, uid)
// If the Pod is deleted the status will be cleared in
// RemoveOrphanedStatuses, so we just ignore the update here.
return
}
if err == nil {
translatedUID := m.podManager.TranslatePodUID(pod.UID)
if len(translatedUID) > 0 && translatedUID != uid {
glog.V(2).Infof("Pod %q was deleted and then recreated, skipping status update; old UID %q, new UID %q", format.Pod(pod), uid, translatedUID)
m.deletePodStatus(uid)
return
}
pod.Status = status.status
if err := podutil.SetInitContainersStatusesAnnotations(pod); err != nil {
glog.Error(err)
}
// TODO: handle conflict as a retry, make that easier too.
pod, err = m.kubeClient.Core().Pods(pod.Namespace).UpdateStatus(pod)
if err == nil {
glog.V(3).Infof("Status for pod %q updated successfully: %+v", format.Pod(pod), status)
m.apiStatusVersions[pod.UID] = status.version
if kubepod.IsMirrorPod(pod) {
// We don't handle graceful deletion of mirror pods.
return
}
if pod.DeletionTimestamp == nil {
return
}
if !notRunning(pod.Status.ContainerStatuses) {
glog.V(3).Infof("Pod %q is terminated, but some containers are still running", format.Pod(pod))
return
}
deleteOptions := v1.NewDeleteOptions(0)
// Use the pod UID as the precondition for deletion to prevent deleting a newly created pod with the same name and namespace.
deleteOptions.Preconditions = v1.NewUIDPreconditions(string(pod.UID))
if err = m.kubeClient.Core().Pods(pod.Namespace).Delete(pod.Name, deleteOptions); err == nil {
glog.V(3).Infof("Pod %q fully terminated and removed from etcd", format.Pod(pod))
m.deletePodStatus(uid)
return
}
}
}
// We failed to update status, wait for periodic sync to retry.
glog.Warningf("Failed to update status for pod %q: %v", format.Pod(pod), err)
}
// checkHostPortConflicts detects pods with conflicted host ports.
func hasHostPortConflicts(pods []*api.Pod) bool {
ports := sets.String{}
for _, pod := range pods {
if errs := validation.AccumulateUniqueHostPorts(pod.Spec.Containers, &ports, field.NewPath("spec", "containers")); len(errs) > 0 {
glog.Errorf("Pod %q: HostPort is already allocated, ignoring: %v", format.Pod(pod), errs)
return true
}
if errs := validation.AccumulateUniqueHostPorts(pod.Spec.InitContainers, &ports, field.NewPath("spec", "initContainers")); len(errs) > 0 {
glog.Errorf("Pod %q: HostPort is already allocated, ignoring: %v", format.Pod(pod), errs)
return true
}
}
return false
}
// RunPod first creates the unit file for a pod, and then
// starts the unit over d-bus.
func (r *Runtime) RunPod(pod *api.Pod, pullSecrets []api.Secret) error {
glog.V(4).Infof("Rkt starts to run pod: name %q.", format.Pod(pod))
name, runtimePod, prepareErr := r.preparePod(pod, pullSecrets)
// Set container references and generate events.
// If preparedPod fails, then send out 'failed' events for each container.
// Otherwise, store the container references so we can use them later to send events.
for i, c := range pod.Spec.Containers {
ref, err := kubecontainer.GenerateContainerRef(pod, &c)
if err != nil {
glog.Errorf("Couldn't make a ref to pod %q, container %v: '%v'", format.Pod(pod), c.Name, err)
continue
}
if prepareErr != nil {
r.recorder.Eventf(ref, api.EventTypeWarning, kubecontainer.FailedToCreateContainer, "Failed to create rkt container with error: %v", prepareErr)
continue
}
containerID := runtimePod.Containers[i].ID
r.containerRefManager.SetRef(containerID, ref)
}
if prepareErr != nil {
return prepareErr
}
r.generateEvents(runtimePod, "Created", nil)
// RestartUnit has the same effect as StartUnit if the unit is not running, besides it can restart
// a unit if the unit file is changed and reloaded.
reschan := make(chan string)
_, err := r.systemd.RestartUnit(name, "replace", reschan)
if err != nil {
r.generateEvents(runtimePod, "Failed", err)
return err
}
res := <-reschan
if res != "done" {
err := fmt.Errorf("Failed to restart unit %q: %s", name, res)
r.generateEvents(runtimePod, "Failed", err)
return err
}
r.generateEvents(runtimePod, "Started", nil)
return nil
}
// Admit rejects a pod if its not safe to admit for node stability.
func (m *managerImpl) Admit(attrs *lifecycle.PodAdmitAttributes) lifecycle.PodAdmitResult {
m.RLock()
defer m.RUnlock()
if len(m.nodeConditions) == 0 {
return lifecycle.PodAdmitResult{Admit: true}
}
// Check the node conditions to identify the resource under pressure.
// The resource can only be either disk or memory; set the default to disk.
resource := api.ResourceStorage
if hasNodeCondition(m.nodeConditions, api.NodeMemoryPressure) {
resource = api.ResourceMemory
// the node has memory pressure, admit if not best-effort
notBestEffort := qos.BestEffort != qos.GetPodQOS(attrs.Pod)
if notBestEffort {
return lifecycle.PodAdmitResult{Admit: true}
}
}
// reject pods when under memory pressure (if pod is best effort), or if under disk pressure.
glog.Warningf("Failed to admit pod %q - node has conditions: %v", format.Pod(attrs.Pod), m.nodeConditions)
return lifecycle.PodAdmitResult{
Admit: false,
Reason: reason,
Message: getMessage(resource),
}
}
// TODO(random-liu): This should be removed soon after rkt implements GetPodStatus.
func ShouldContainerBeRestartedOldVersion(container *api.Container, pod *api.Pod, podStatus *api.PodStatus) bool {
// Get all dead container status.
var resultStatus []*api.ContainerStatus
for i, containerStatus := range podStatus.ContainerStatuses {
if containerStatus.Name == container.Name && containerStatus.State.Terminated != nil {
resultStatus = append(resultStatus, &podStatus.ContainerStatuses[i])
}
}
// Check RestartPolicy for dead container.
if len(resultStatus) > 0 {
if pod.Spec.RestartPolicy == api.RestartPolicyNever {
glog.V(4).Infof("Already ran container %q of pod %q, do nothing", container.Name, format.Pod(pod))
return false
}
if pod.Spec.RestartPolicy == api.RestartPolicyOnFailure {
// Check the exit code of last run. Note: This assumes the result is sorted
// by the created time in reverse order.
if resultStatus[0].State.Terminated.ExitCode == 0 {
glog.V(4).Infof("Already successfully ran container %q of pod %q, do nothing", container.Name, format.Pod(pod))
return false
}
}
}
return true
}
// updateStatusInternal updates the internal status cache, and queues an update to the api server if
// necessary. Returns whether an update was triggered.
// This method IS NOT THREAD SAFE and must be called from a locked function.
func (m *manager) updateStatusInternal(pod *api.Pod, status api.PodStatus) bool {
// The intent here is to prevent concurrent updates to a pod's status from
// clobbering each other so the phase of a pod progresses monotonically.
oldStatus, found := m.podStatuses[pod.UID]
if found && isStatusEqual(&oldStatus.status, &status) && pod.DeletionTimestamp == nil {
glog.V(3).Infof("Ignoring same status for pod %q, status: %+v", format.Pod(pod), status)
return false // No new status.
}
newStatus := versionedPodStatus{
status: status,
version: oldStatus.version + 1,
podName: pod.Name,
podNamespace: pod.Namespace,
}
m.podStatuses[pod.UID] = newStatus
select {
case m.podStatusChannel <- podStatusSyncRequest{pod.UID, newStatus}:
return true
default:
// Let the periodic syncBatch handle the update if the channel is full.
// We can't block, since we hold the mutex lock.
return false
}
}
// needsReconcile compares the given status with the status in the pod manager (which
// in fact comes from apiserver), returns whether the status needs to be reconciled with
// the apiserver. Now when pod status is inconsistent between apiserver and kubelet,
// kubelet should forcibly send an update to reconclie the inconsistence, because kubelet
// should be the source of truth of pod status.
// NOTE(random-liu): It's simpler to pass in mirror pod uid and get mirror pod by uid, but
// now the pod manager only supports getting mirror pod by static pod, so we have to pass
// static pod uid here.
// TODO(random-liu): Simplify the logic when mirror pod manager is added.
func (m *manager) needsReconcile(uid types.UID, status api.PodStatus) bool {
// The pod could be a static pod, so we should translate first.
pod, ok := m.podManager.GetPodByUID(uid)
if !ok {
// Although we get uid from pod manager in syncBatch, it still could be deleted before here.
glog.V(4).Infof("Pod %q has been deleted, no need to reconcile", format.Pod(pod))
return false
}
// If the pod is a static pod, we should check its mirror pod, because only status in mirror pod is meaningful to us.
if kubepod.IsStaticPod(pod) {
mirrorPod, ok := m.podManager.GetMirrorPodByPod(pod)
if !ok {
glog.V(4).Infof("Static pod %q has no corresponding mirror pod, no need to reconcile", format.Pod(pod))
return false
}
pod = mirrorPod
}
if isStatusEqual(&pod.Status, &status) {
// If the status from the source is the same with the cached status,
// reconcile is not needed. Just return.
return false
}
glog.V(3).Infof("Pod status is inconsistent with cached status, a reconciliation should be triggered:\n %+v", util.ObjectDiff(pod.Status, status))
return true
}
func (p *podWorkers) managePodLoop(podUpdates <-chan UpdatePodOptions) {
var lastSyncTime time.Time
for update := range podUpdates {
err := func() error {
podUID := update.Pod.UID
// This is a blocking call that would return only if the cache
// has an entry for the pod that is newer than minRuntimeCache
// Time. This ensures the worker doesn't start syncing until
// after the cache is at least newer than the finished time of
// the previous sync.
status, err := p.podCache.GetNewerThan(podUID, lastSyncTime)
if err != nil {
return err
}
err = p.syncPodFn(syncPodOptions{
mirrorPod: update.MirrorPod,
pod: update.Pod,
podStatus: status,
killPodOptions: update.KillPodOptions,
updateType: update.UpdateType,
})
lastSyncTime = time.Now()
return err
}()
// notify the call-back function if the operation succeeded or not
if update.OnCompleteFunc != nil {
update.OnCompleteFunc(err)
}
if err != nil {
glog.Errorf("Error syncing pod %s (%q), skipping: %v", update.Pod.UID, format.Pod(update.Pod), err)
p.recorder.Eventf(update.Pod, v1.EventTypeWarning, events.FailedSync, "Error syncing pod, skipping: %v", err)
}
p.wrapUp(update.Pod.UID, err)
}
}
// update ready status of all pods running on given node from master
// return true if success
func (nc *NodeController) markAllPodsNotReady(nodeName string) error {
glog.V(2).Infof("Update ready status of pods on node [%v]", nodeName)
opts := api.ListOptions{FieldSelector: fields.OneTermEqualSelector(client.PodHost, nodeName)}
pods, err := nc.kubeClient.Core().Pods(api.NamespaceAll).List(opts)
if err != nil {
return err
}
errMsg := []string{}
for _, pod := range pods.Items {
// Defensive check, also needed for tests.
if pod.Spec.NodeName != nodeName {
continue
}
for i, cond := range pod.Status.Conditions {
if cond.Type == api.PodReady {
pod.Status.Conditions[i].Status = api.ConditionFalse
glog.V(2).Infof("Updating ready status of pod %v to false", pod.Name)
pod, err := nc.kubeClient.Core().Pods(pod.Namespace).UpdateStatus(&pod)
if err != nil {
glog.Warningf("Failed to updated status for pod %q: %v", format.Pod(pod), err)
errMsg = append(errMsg, fmt.Sprintf("%v", err))
}
break
}
}
}
if len(errMsg) == 0 {
return nil
}
return fmt.Errorf("%v", strings.Join(errMsg, "; "))
}
// ShouldContainerBeRestarted checks whether a container needs to be restarted.
// TODO(yifan): Think about how to refactor this.
func ShouldContainerBeRestarted(container *api.Container, pod *api.Pod, podStatus *PodStatus) bool {
// Get latest container status.
status := podStatus.FindContainerStatusByName(container.Name)
// If the container was never started before, we should start it.
// NOTE(random-liu): If all historical containers were GC'd, we'll also return true here.
if status == nil {
return true
}
// Check whether container is running
if status.State == ContainerStateRunning {
return false
}
// Always restart container in unknown state now
if status.State == ContainerStateUnknown {
return true
}
// Check RestartPolicy for dead container
if pod.Spec.RestartPolicy == api.RestartPolicyNever {
glog.V(4).Infof("Already ran container %q of pod %q, do nothing", container.Name, format.Pod(pod))
return false
}
if pod.Spec.RestartPolicy == api.RestartPolicyOnFailure {
// Check the exit code.
if status.ExitCode == 0 {
glog.V(4).Infof("Already successfully ran container %q of pod %q, do nothing", container.Name, format.Pod(pod))
return false
}
}
return true
}
// GetPodStatus retrieves the status of the pod, including the
// information of all containers in the pod that are visble in Runtime.
func (m *kubeGenericRuntimeManager) GetPodStatus(uid kubetypes.UID, name, namespace string) (*kubecontainer.PodStatus, error) {
// Now we retain restart count of container as a container label. Each time a container
// restarts, pod will read the restart count from the registered dead container, increment
// it to get the new restart count, and then add a label with the new restart count on
// the newly started container.
// However, there are some limitations of this method:
// 1. When all dead containers were garbage collected, the container status could
// not get the historical value and would be *inaccurate*. Fortunately, the chance
// is really slim.
// 2. When working with old version containers which have no restart count label,
// we can only assume their restart count is 0.
// Anyhow, we only promised "best-effort" restart count reporting, we can just ignore
// these limitations now.
// TODO: move this comment to SyncPod.
podSandboxIDs, err := m.getSandboxIDByPodUID(string(uid), nil)
if err != nil {
return nil, err
}
podFullName := format.Pod(&api.Pod{
ObjectMeta: api.ObjectMeta{
Name: name,
Namespace: namespace,
UID: uid,
},
})
glog.V(4).Infof("getSandboxIDByPodUID got sandbox IDs %q for pod %q", podSandboxIDs, podFullName)
sandboxStatuses := make([]*runtimeApi.PodSandboxStatus, len(podSandboxIDs))
podIP := ""
for idx, podSandboxID := range podSandboxIDs {
podSandboxStatus, err := m.runtimeService.PodSandboxStatus(podSandboxID)
if err != nil {
glog.Errorf("PodSandboxStatus of sandbox %q for pod %q error: %v", podSandboxID, podFullName, err)
return nil, err
}
sandboxStatuses[idx] = podSandboxStatus
// Only get pod IP from latest sandbox
if idx == 0 && podSandboxStatus.GetState() == runtimeApi.PodSandBoxState_READY {
podIP = m.determinePodSandboxIP(namespace, name, podSandboxStatus)
}
}
// Get statuses of all containers visible in the pod.
containerStatuses, err := m.getPodContainerStatuses(uid, name, namespace)
if err != nil {
glog.Errorf("getPodContainerStatuses for pod %q failed: %v", podFullName, err)
return nil, err
}
return &kubecontainer.PodStatus{
ID: uid,
Name: name,
Namespace: namespace,
IP: podIP,
SandboxStatuses: sandboxStatuses,
ContainerStatuses: containerStatuses,
}, nil
}
func (pb *prober) runProbe(p *api.Probe, pod *api.Pod, status api.PodStatus, container api.Container, containerID kubecontainer.ContainerID) (probe.Result, string, error) {
timeout := time.Duration(p.TimeoutSeconds) * time.Second
if p.Exec != nil {
glog.V(4).Infof("Exec-Probe Pod: %v, Container: %v, Command: %v", pod, container, p.Exec.Command)
return pb.exec.Probe(pb.newExecInContainer(container, containerID, p.Exec.Command))
}
if p.HTTPGet != nil {
scheme := strings.ToLower(string(p.HTTPGet.Scheme))
host := p.HTTPGet.Host
if host == "" {
host = status.PodIP
}
port, err := extractPort(p.HTTPGet.Port, container)
if err != nil {
return probe.Unknown, "", err
}
path := p.HTTPGet.Path
glog.V(4).Infof("HTTP-Probe Host: %v://%v, Port: %v, Path: %v", scheme, host, port, path)
url := formatURL(scheme, host, port, path)
headers := buildHeader(p.HTTPGet.HTTPHeaders)
glog.V(4).Infof("HTTP-Probe Headers: %v", headers)
return pb.http.Probe(url, headers, timeout)
}
if p.TCPSocket != nil {
port, err := extractPort(p.TCPSocket.Port, container)
if err != nil {
return probe.Unknown, "", err
}
glog.V(4).Infof("TCP-Probe PodIP: %v, Port: %v, Timeout: %v", status.PodIP, port, timeout)
return pb.tcp.Probe(status.PodIP, port, timeout)
}
glog.Warningf("Failed to find probe builder for container: %v", container)
return probe.Unknown, "", fmt.Errorf("Missing probe handler for %s:%s", format.Pod(pod), container.Name)
}
// If a container is still in backoff, the function will return a brief backoff error and
// a detailed error message.
func (m *kubeGenericRuntimeManager) doBackOff(pod *api.Pod, container *api.Container, podStatus *kubecontainer.PodStatus, backOff *flowcontrol.Backoff) (bool, string, error) {
var cStatus *kubecontainer.ContainerStatus
for _, c := range podStatus.ContainerStatuses {
if c.Name == container.Name && c.State == kubecontainer.ContainerStateExited {
cStatus = c
break
}
}
if cStatus == nil {
return false, "", nil
}
glog.Infof("checking backoff for container %q in pod %q", container.Name, format.Pod(pod))
// Use the finished time of the latest exited container as the start point to calculate whether to do back-off.
ts := cStatus.FinishedAt
// backOff requires a unique key to identify the container.
key := getStableKey(pod, container)
if backOff.IsInBackOffSince(key, ts) {
if ref, err := kubecontainer.GenerateContainerRef(pod, container); err == nil {
m.recorder.Eventf(ref, api.EventTypeWarning, events.BackOffStartContainer, "Back-off restarting failed container")
}
err := fmt.Errorf("Back-off %s restarting failed container=%s pod=%s", backOff.Get(key), container.Name, format.Pod(pod))
glog.Infof("%s", err.Error())
return true, err.Error(), kubecontainer.ErrCrashLoopBackOff
}
backOff.Next(key, ts)
return false, "", nil
}
// createPodSandbox creates a pod sandbox and returns (podSandBoxID, message, error).
func (m *kubeGenericRuntimeManager) createPodSandbox(pod *api.Pod, attempt uint32) (string, string, error) {
podSandboxConfig, err := m.generatePodSandboxConfig(pod, attempt)
if err != nil {
message := fmt.Sprintf("GeneratePodSandboxConfig for pod %q failed: %v", format.Pod(pod), err)
glog.Error(message)
return "", message, err
}
podSandBoxID, err := m.runtimeService.RunPodSandbox(podSandboxConfig)
if err != nil {
message := fmt.Sprintf("CreatePodSandbox for pod %q failed: %v", format.Pod(pod), err)
glog.Error(message)
return "", message, err
}
return podSandBoxID, "", nil
}
// GetContainerLogs uses rkt's GetLogs API to get the logs of the container.
// By default, it returns a snapshot of the container log. Set |follow| to true to
// stream the log. Set |follow| to false and specify the number of lines (e.g.
// "100" or "all") to tail the log.
//
// TODO(yifan): This doesn't work with lkvm stage1 yet.
func (r *Runtime) GetContainerLogs(pod *v1.Pod, containerID kubecontainer.ContainerID, logOptions *v1.PodLogOptions, stdout, stderr io.Writer) error {
id, err := parseContainerID(containerID)
if err != nil {
return err
}
var since int64
if logOptions.SinceSeconds != nil {
t := metav1.Now().Add(-time.Duration(*logOptions.SinceSeconds) * time.Second)
since = t.Unix()
}
if logOptions.SinceTime != nil {
since = logOptions.SinceTime.Unix()
}
getLogsRequest := &rktapi.GetLogsRequest{
PodId: id.uuid,
AppName: id.appName,
Follow: logOptions.Follow,
SinceTime: since,
}
if logOptions.TailLines != nil {
getLogsRequest.Lines = int32(*logOptions.TailLines)
}
stream, err := r.apisvc.GetLogs(context.Background(), getLogsRequest)
if err != nil {
glog.Errorf("rkt: Failed to create log stream for pod %q: %v", format.Pod(pod), err)
return err
}
for {
log, err := stream.Recv()
if err == io.EOF {
break
}
if err != nil {
glog.Errorf("rkt: Failed to receive log for pod %q: %v", format.Pod(pod), err)
return err
}
processLines(log.Lines, logOptions, stdout, stderr)
}
return nil
}
// runPod runs a single pod and wait until all containers are running.
func (kl *Kubelet) runPod(pod *api.Pod, retryDelay time.Duration) error {
delay := retryDelay
retry := 0
for {
status, err := kl.containerRuntime.GetPodStatus(pod.UID, pod.Name, pod.Namespace)
if err != nil {
return fmt.Errorf("Unable to get status for pod %q: %v", format.Pod(pod), err)
}
if kl.isPodRunning(pod, status) {
glog.Infof("pod %q containers running", format.Pod(pod))
return nil
}
glog.Infof("pod %q containers not running: syncing", format.Pod(pod))
glog.Infof("Creating a mirror pod for static pod %q", format.Pod(pod))
if err := kl.podManager.CreateMirrorPod(pod); err != nil {
glog.Errorf("Failed creating a mirror pod %q: %v", format.Pod(pod), err)
}
mirrorPod, _ := kl.podManager.GetMirrorPodByPod(pod)
if err = kl.syncPod(pod, mirrorPod, status, kubetypes.SyncPodUpdate); err != nil {
return fmt.Errorf("error syncing pod %q: %v", format.Pod(pod), err)
}
if retry >= runOnceMaxRetries {
return fmt.Errorf("timeout error: pod %q containers not running after %d retries", format.Pod(pod), runOnceMaxRetries)
}
// TODO(proppy): health checking would be better than waiting + checking the state at the next iteration.
glog.Infof("pod %q containers synced, waiting for %v", format.Pod(pod), delay)
time.Sleep(delay)
retry++
delay *= runOnceRetryDelayBackoff
}
}
// isPodRunning returns true if all containers of a manifest are running.
func (kl *Kubelet) isPodRunning(pod *api.Pod, status *kubecontainer.PodStatus) bool {
for _, c := range pod.Spec.Containers {
cs := status.FindContainerStatusByName(c.Name)
if cs == nil || cs.State != kubecontainer.ContainerStateRunning {
glog.Infof("Container %q for pod %q not running", c.Name, format.Pod(pod))
return false
}
}
return true
}
// deletePods will delete all pods from master running on given node, and return true
// if any pods were deleted, or were found pending deletion.
func deletePods(kubeClient clientset.Interface, recorder record.EventRecorder, nodeName, nodeUID string, daemonStore cache.StoreToDaemonSetLister) (bool, error) {
remaining := false
selector := fields.OneTermEqualSelector(api.PodHostField, nodeName)
options := api.ListOptions{FieldSelector: selector}
pods, err := kubeClient.Core().Pods(api.NamespaceAll).List(options)
var updateErrList []error
if err != nil {
return remaining, err
}
if len(pods.Items) > 0 {
recordNodeEvent(recorder, nodeName, nodeUID, api.EventTypeNormal, "DeletingAllPods", fmt.Sprintf("Deleting all Pods from Node %v.", nodeName))
}
for _, pod := range pods.Items {
// Defensive check, also needed for tests.
if pod.Spec.NodeName != nodeName {
continue
}
// Set reason and message in the pod object.
if _, err = setPodTerminationReason(kubeClient, &pod, nodeName); err != nil {
if errors.IsConflict(err) {
updateErrList = append(updateErrList,
fmt.Errorf("update status failed for pod %q: %v", format.Pod(&pod), err))
continue
}
}
// if the pod has already been marked for deletion, we still return true that there are remaining pods.
if pod.DeletionGracePeriodSeconds != nil {
remaining = true
continue
}
// if the pod is managed by a daemonset, ignore it
_, err := daemonStore.GetPodDaemonSets(&pod)
if err == nil { // No error means at least one daemonset was found
continue
}
glog.V(2).Infof("Starting deletion of pod %v", pod.Name)
recorder.Eventf(&pod, api.EventTypeNormal, "NodeControllerEviction", "Marking for deletion Pod %s from Node %s", pod.Name, nodeName)
if err := kubeClient.Core().Pods(pod.Namespace).Delete(pod.Name, nil); err != nil {
return false, err
}
remaining = true
}
if len(updateErrList) > 0 {
return false, utilerrors.NewAggregate(updateErrList)
}
return remaining, nil
}
func (vm *volumeManager) WaitForAttachAndMount(pod *v1.Pod) error {
expectedVolumes := getExpectedVolumes(pod)
if len(expectedVolumes) == 0 {
// No volumes to verify
return nil
}
glog.V(3).Infof("Waiting for volumes to attach and mount for pod %q", format.Pod(pod))
uniquePodName := volumehelper.GetUniquePodName(pod)
// Some pods expect to have Setup called over and over again to update.
// Remount plugins for which this is true. (Atomically updating volumes,
// like Downward API, depend on this to update the contents of the volume).
vm.desiredStateOfWorldPopulator.ReprocessPod(uniquePodName)
vm.actualStateOfWorld.MarkRemountRequired(uniquePodName)
err := wait.Poll(
podAttachAndMountRetryInterval,
podAttachAndMountTimeout,
vm.verifyVolumesMountedFunc(uniquePodName, expectedVolumes))
if err != nil {
// Timeout expired
unmountedVolumes :=
vm.getUnmountedVolumes(uniquePodName, expectedVolumes)
if len(unmountedVolumes) == 0 {
return nil
}
return fmt.Errorf(
"timeout expired waiting for volumes to attach/mount for pod %q/%q. list of unattached/unmounted volumes=%v",
pod.Namespace,
pod.Name,
unmountedVolumes)
}
glog.V(3).Infof("All volumes are attached and mounted for pod %q", format.Pod(pod))
return nil
}
func (m *manager) SetContainerReadiness(pod *api.Pod, containerID kubecontainer.ContainerID, ready bool) {
m.podStatusesLock.Lock()
defer m.podStatusesLock.Unlock()
oldStatus, found := m.podStatuses[pod.UID]
if !found {
glog.Warningf("Container readiness changed before pod has synced: %q - %q",
format.Pod(pod), containerID.String())
return
}
status := oldStatus.status
// Find the container to update.
containerIndex := -1
for i, c := range status.ContainerStatuses {
if c.ContainerID == containerID.String() {
containerIndex = i
break
}
}
if containerIndex == -1 {
glog.Warningf("Container readiness changed for unknown container: %q - %q",
format.Pod(pod), containerID.String())
return
}
if status.ContainerStatuses[containerIndex].Ready == ready {
glog.V(4).Infof("Container readiness unchanged (%v): %q - %q", ready,
format.Pod(pod), containerID.String())
return
}
// Make sure we're not updating the cached version.
status.ContainerStatuses = make([]api.ContainerStatus, len(status.ContainerStatuses))
copy(status.ContainerStatuses, oldStatus.status.ContainerStatuses)
status.ContainerStatuses[containerIndex].Ready = ready
m.updateStatusInternal(pod, status)
}
// isPodRunning returns true if all containers of a manifest are running.
func (kl *Kubelet) isPodRunning(pod *api.Pod, runningPod container.Pod) (bool, error) {
// TODO(random-liu): Change this to new pod status
status, err := kl.containerRuntime.GetAPIPodStatus(pod)
if err != nil {
glog.Infof("Failed to get the status of pod %q: %v", format.Pod(pod), err)
return false, err
}
for _, st := range status.ContainerStatuses {
if st.State.Running == nil {
glog.Infof("Container %q not running: %#v", st.Name, st.State)
return false, nil
}
}
return true, nil
}
// generateAPIPodStatus creates the final API pod status for a pod, given the
// internal pod status.
func (kl *Kubelet) generateAPIPodStatus(pod *api.Pod, podStatus *kubecontainer.PodStatus) api.PodStatus {
glog.V(3).Infof("Generating status for %q", format.Pod(pod))
// check if an internal module has requested the pod is evicted.
for _, podSyncHandler := range kl.PodSyncHandlers {
if result := podSyncHandler.ShouldEvict(pod); result.Evict {
return api.PodStatus{
Phase: api.PodFailed,
Reason: result.Reason,
Message: result.Message,
}
}
}
s := kl.convertStatusToAPIStatus(pod, podStatus)
// Assume info is ready to process
spec := &pod.Spec
allStatus := append(append([]api.ContainerStatus{}, s.ContainerStatuses...), s.InitContainerStatuses...)
s.Phase = GetPhase(spec, allStatus)
kl.probeManager.UpdatePodStatus(pod.UID, s)
s.Conditions = append(s.Conditions, status.GeneratePodInitializedCondition(spec, s.InitContainerStatuses, s.Phase))
s.Conditions = append(s.Conditions, status.GeneratePodReadyCondition(spec, s.ContainerStatuses, s.Phase))
// s (the PodStatus we are creating) will not have a PodScheduled condition yet, because converStatusToAPIStatus()
// does not create one. If the existing PodStatus has a PodScheduled condition, then copy it into s and make sure
// it is set to true. If the existing PodStatus does not have a PodScheduled condition, then create one that is set to true.
if _, oldPodScheduled := api.GetPodCondition(&pod.Status, api.PodScheduled); oldPodScheduled != nil {
s.Conditions = append(s.Conditions, *oldPodScheduled)
}
api.UpdatePodCondition(&pod.Status, &api.PodCondition{
Type: api.PodScheduled,
Status: api.ConditionTrue,
})
if !kl.standaloneMode {
hostIP, err := kl.getHostIPAnyWay()
if err != nil {
glog.V(4).Infof("Cannot get host IP: %v", err)
} else {
s.HostIP = hostIP.String()
if podUsesHostNetwork(pod) && s.PodIP == "" {
s.PodIP = hostIP.String()
}
}
}
return *s
}
