mirrors/ceph-csi
1
0
Fork 0
mirror of https://github.com/ceph/ceph-csi.git synced 2025-06-13 10:33:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update to kube v1.17

Browse Source 
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
This commit is contained in:
Humble Chirammal
2020-01-14 16:08:55 +05:30
committed by mergify[bot]
parent 327fcd1b1b
commit 3af1e26d7c
1710 changed files with 289562 additions and 168638 deletions
Download Patch File Download Diff File Expand all files Collapse all files

120
vendor/k8s.io/kubernetes/pkg/controller/controller_ref_manager.go generated vendored
View File

@ -17,15 +17,17 @@ limitations under the License.
package controller
import (
"encoding/json"
"fmt"
"sync"
apps "k8s.io/api/apps/v1"
"k8s.io/api/core/v1"
v1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/api/errors"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/apimachinery/pkg/runtime/schema"
"k8s.io/apimachinery/pkg/types"
utilerrors "k8s.io/apimachinery/pkg/util/errors"
"k8s.io/klog"
)
@ -64,7 +66,7 @@ func (m *BaseControllerRefManager) CanAdopt() error {
//
// No reconciliation will be attempted if the controller is being deleted.
func (m *BaseControllerRefManager) ClaimObject(obj metav1.Object, match func(metav1.Object) bool, adopt, release func(metav1.Object) error) (bool, error) {
controllerRef := metav1.GetControllerOf(obj)
controllerRef := metav1.GetControllerOfNoCopy(obj)
if controllerRef != nil {
if controllerRef.UID != m.Controller.GetUID() {
// Owned by someone else. Ignore.
@ -213,11 +215,12 @@ func (m *PodControllerRefManager) AdoptPod(pod *v1.Pod) error {
}
// Note that ValidateOwnerReferences() will reject this patch if another
// OwnerReference exists with controller=true.
addControllerPatch := fmt.Sprintf(
`{"metadata":{"ownerReferences":[{"apiVersion":"%s","kind":"%s","name":"%s","uid":"%s","controller":true,"blockOwnerDeletion":true}],"uid":"%s"}}`,
m.controllerKind.GroupVersion(), m.controllerKind.Kind,
m.Controller.GetName(), m.Controller.GetUID(), pod.UID)
return m.podControl.PatchPod(pod.Namespace, pod.Name, []byte(addControllerPatch))
patchBytes, err := ownerRefControllerPatch(m.Controller, m.controllerKind, pod.UID)
if err != nil {
return err
}
return m.podControl.PatchPod(pod.Namespace, pod.Name, patchBytes)
}
// ReleasePod sends a patch to free the pod from the control of the controller.
@ -225,8 +228,11 @@ func (m *PodControllerRefManager) AdoptPod(pod *v1.Pod) error {
func (m *PodControllerRefManager) ReleasePod(pod *v1.Pod) error {
klog.V(2).Infof("patching pod %s_%s to remove its controllerRef to %s/%s:%s",
pod.Namespace, pod.Name, m.controllerKind.GroupVersion(), m.controllerKind.Kind, m.Controller.GetName())
deleteOwnerRefPatch := fmt.Sprintf(`{"metadata":{"ownerReferences":[{"$patch":"delete","uid":"%s"}],"uid":"%s"}}`, m.Controller.GetUID(), pod.UID)
err := m.podControl.PatchPod(pod.Namespace, pod.Name, []byte(deleteOwnerRefPatch))
patchBytes, err := deleteOwnerRefStrategicMergePatch(pod.UID, m.Controller.GetUID())
if err != nil {
return err
}
err = m.podControl.PatchPod(pod.Namespace, pod.Name, patchBytes)
if err != nil {
if errors.IsNotFound(err) {
// If the pod no longer exists, ignore it.
@ -335,11 +341,11 @@ func (m *ReplicaSetControllerRefManager) AdoptReplicaSet(rs *apps.ReplicaSet) er
}
// Note that ValidateOwnerReferences() will reject this patch if another
// OwnerReference exists with controller=true.
addControllerPatch := fmt.Sprintf(
`{"metadata":{"ownerReferences":[{"apiVersion":"%s","kind":"%s","name":"%s","uid":"%s","controller":true,"blockOwnerDeletion":true}],"uid":"%s"}}`,
m.controllerKind.GroupVersion(), m.controllerKind.Kind,
m.Controller.GetName(), m.Controller.GetUID(), rs.UID)
return m.rsControl.PatchReplicaSet(rs.Namespace, rs.Name, []byte(addControllerPatch))
patchBytes, err := ownerRefControllerPatch(m.Controller, m.controllerKind, rs.UID)
if err != nil {
return err
}
return m.rsControl.PatchReplicaSet(rs.Namespace, rs.Name, patchBytes)
}
// ReleaseReplicaSet sends a patch to free the ReplicaSet from the control of the Deployment controller.
@ -347,8 +353,11 @@ func (m *ReplicaSetControllerRefManager) AdoptReplicaSet(rs *apps.ReplicaSet) er
func (m *ReplicaSetControllerRefManager) ReleaseReplicaSet(replicaSet *apps.ReplicaSet) error {
klog.V(2).Infof("patching ReplicaSet %s_%s to remove its controllerRef to %s/%s:%s",
replicaSet.Namespace, replicaSet.Name, m.controllerKind.GroupVersion(), m.controllerKind.Kind, m.Controller.GetName())
deleteOwnerRefPatch := fmt.Sprintf(`{"metadata":{"ownerReferences":[{"$patch":"delete","uid":"%s"}],"uid":"%s"}}`, m.Controller.GetUID(), replicaSet.UID)
err := m.rsControl.PatchReplicaSet(replicaSet.Namespace, replicaSet.Name, []byte(deleteOwnerRefPatch))
patchBytes, err := deleteOwnerRefStrategicMergePatch(replicaSet.UID, m.Controller.GetUID())
if err != nil {
return err
}
err = m.rsControl.PatchReplicaSet(replicaSet.Namespace, replicaSet.Name, patchBytes)
if err != nil {
if errors.IsNotFound(err) {
// If the ReplicaSet no longer exists, ignore it.
@ -470,11 +479,11 @@ func (m *ControllerRevisionControllerRefManager) AdoptControllerRevision(history
}
// Note that ValidateOwnerReferences() will reject this patch if another
// OwnerReference exists with controller=true.
addControllerPatch := fmt.Sprintf(
`{"metadata":{"ownerReferences":[{"apiVersion":"%s","kind":"%s","name":"%s","uid":"%s","controller":true,"blockOwnerDeletion":true}],"uid":"%s"}}`,
m.controllerKind.GroupVersion(), m.controllerKind.Kind,
m.Controller.GetName(), m.Controller.GetUID(), history.UID)
return m.crControl.PatchControllerRevision(history.Namespace, history.Name, []byte(addControllerPatch))
patchBytes, err := ownerRefControllerPatch(m.Controller, m.controllerKind, history.UID)
if err != nil {
return err
}
return m.crControl.PatchControllerRevision(history.Namespace, history.Name, patchBytes)
}
// ReleaseControllerRevision sends a patch to free the ControllerRevision from the control of its controller.
@ -482,8 +491,12 @@ func (m *ControllerRevisionControllerRefManager) AdoptControllerRevision(history
func (m *ControllerRevisionControllerRefManager) ReleaseControllerRevision(history *apps.ControllerRevision) error {
klog.V(2).Infof("patching ControllerRevision %s_%s to remove its controllerRef to %s/%s:%s",
history.Namespace, history.Name, m.controllerKind.GroupVersion(), m.controllerKind.Kind, m.Controller.GetName())
deleteOwnerRefPatch := fmt.Sprintf(`{"metadata":{"ownerReferences":[{"$patch":"delete","uid":"%s"}],"uid":"%s"}}`, m.Controller.GetUID(), history.UID)
err := m.crControl.PatchControllerRevision(history.Namespace, history.Name, []byte(deleteOwnerRefPatch))
patchBytes, err := deleteOwnerRefStrategicMergePatch(history.UID, m.Controller.GetUID())
if err != nil {
return err
}
err = m.crControl.PatchControllerRevision(history.Namespace, history.Name, patchBytes)
if err != nil {
if errors.IsNotFound(err) {
// If the ControllerRevision no longer exists, ignore it.
@ -499,3 +512,64 @@ func (m *ControllerRevisionControllerRefManager) ReleaseControllerRevision(histo
}
return err
}
type objectForDeleteOwnerRefStrategicMergePatch struct {
Metadata objectMetaForMergePatch `json:"metadata"`
}
type objectMetaForMergePatch struct {
UID types.UID `json:"uid"`
OwnerReferences []map[string]string `json:"ownerReferences"`
}
func deleteOwnerRefStrategicMergePatch(dependentUID types.UID, ownerUIDs ...types.UID) ([]byte, error) {
var pieces []map[string]string
for _, ownerUID := range ownerUIDs {
pieces = append(pieces, map[string]string{"$patch": "delete", "uid": string(ownerUID)})
}
patch := objectForDeleteOwnerRefStrategicMergePatch{
Metadata: objectMetaForMergePatch{
UID: dependentUID,
OwnerReferences: pieces,
},
}
patchBytes, err := json.Marshal(&patch)
if err != nil {
return nil, err
}
return patchBytes, nil
}
type objectForAddOwnerRefPatch struct {
Metadata objectMetaForPatch `json:"metadata"`
}
type objectMetaForPatch struct {
OwnerReferences []metav1.OwnerReference `json:"ownerReferences"`
UID types.UID `json:"uid"`
}
func ownerRefControllerPatch(controller metav1.Object, controllerKind schema.GroupVersionKind, uid types.UID) ([]byte, error) {
blockOwnerDeletion := true
isController := true
addControllerPatch := objectForAddOwnerRefPatch{
Metadata: objectMetaForPatch{
UID: uid,
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: controllerKind.GroupVersion().String(),
Kind: controllerKind.Kind,
Name: controller.GetName(),
UID: controller.GetUID(),
Controller: &isController,
BlockOwnerDeletion: &blockOwnerDeletion,
},
},
},
}
patchBytes, err := json.Marshal(&addControllerPatch)
if err != nil {
return nil, err
}
return patchBytes, nil
}

153
vendor/k8s.io/kubernetes/pkg/controller/controller_utils.go generated vendored
View File

@ -35,7 +35,6 @@ import (
"k8s.io/apimachinery/pkg/types"
"k8s.io/apimachinery/pkg/util/clock"
"k8s.io/apimachinery/pkg/util/rand"
utilruntime "k8s.io/apimachinery/pkg/util/runtime"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/apimachinery/pkg/util/strategicpatch"
"k8s.io/apimachinery/pkg/util/wait"
@ -96,7 +95,8 @@ var UpdateLabelBackoff = wait.Backoff{
}
var (
KeyFunc = cache.DeletionHandlingMetaNamespaceKeyFunc
KeyFunc = cache.DeletionHandlingMetaNamespaceKeyFunc
podPhaseToOrdinal = map[v1.PodPhase]int{v1.PodPending: 0, v1.PodUnknown: 1, v1.PodRunning: 2}
)
type ResyncPeriodFunc func() time.Duration
@ -337,17 +337,17 @@ func (u *UIDTrackingControllerExpectations) GetUIDs(controllerKey string) sets.S
// ExpectDeletions records expectations for the given deleteKeys, against the given controller.
func (u *UIDTrackingControllerExpectations) ExpectDeletions(rcKey string, deletedKeys []string) error {
expectedUIDs := sets.NewString()
for _, k := range deletedKeys {
expectedUIDs.Insert(k)
}
klog.V(4).Infof("Controller %v waiting on deletions for: %+v", rcKey, deletedKeys)
u.uidStoreLock.Lock()
defer u.uidStoreLock.Unlock()
if existing := u.GetUIDs(rcKey); existing != nil && existing.Len() != 0 {
klog.Errorf("Clobbering existing delete keys: %+v", existing)
}
expectedUIDs := sets.NewString()
for _, k := range deletedKeys {
expectedUIDs.Insert(k)
}
klog.V(4).Infof("Controller %v waiting on deletions for: %+v", rcKey, deletedKeys)
if err := u.uidStore.Add(&UIDSet{expectedUIDs, rcKey}); err != nil {
return err
}
@ -573,12 +573,15 @@ func (r RealPodControl) createPods(nodeName, namespace string, template *v1.PodT
if len(nodeName) != 0 {
pod.Spec.NodeName = nodeName
}
if labels.Set(pod.Labels).AsSelectorPreValidated().Empty() {
if len(labels.Set(pod.Labels)) == 0 {
return fmt.Errorf("unable to create pods, no labels")
}
newPod, err := r.KubeClient.CoreV1().Pods(namespace).Create(pod)
if err != nil {
r.Recorder.Eventf(object, v1.EventTypeWarning, FailedCreatePodReason, "Error creating: %v", err)
// only send an event if the namespace isn't terminating
if !apierrors.HasStatusCause(err, v1.NamespaceTerminatingCause) {
r.Recorder.Eventf(object, v1.EventTypeWarning, FailedCreatePodReason, "Error creating: %v", err)
}
return err
}
accessor, err := meta.Accessor(object)
@ -707,9 +710,8 @@ func (s ByLogging) Less(i, j int) bool {
return len(s[i].Spec.NodeName) > 0
}
// 2. PodRunning < PodUnknown < PodPending
m := map[v1.PodPhase]int{v1.PodRunning: 0, v1.PodUnknown: 1, v1.PodPending: 2}
if m[s[i].Status.Phase] != m[s[j].Status.Phase] {
return m[s[i].Status.Phase] < m[s[j].Status.Phase]
if s[i].Status.Phase != s[j].Status.Phase {
return podPhaseToOrdinal[s[i].Status.Phase] > podPhaseToOrdinal[s[j].Status.Phase]
}
// 3. ready < not ready
if podutil.IsPodReady(s[i]) != podutil.IsPodReady(s[j]) {
@ -718,8 +720,12 @@ func (s ByLogging) Less(i, j int) bool {
// TODO: take availability into account when we push minReadySeconds information from deployment into pods,
// see https://github.com/kubernetes/kubernetes/issues/22065
// 4. Been ready for more time < less time < empty time
if podutil.IsPodReady(s[i]) && podutil.IsPodReady(s[j]) && !podReadyTime(s[i]).Equal(podReadyTime(s[j])) {
return afterOrZero(podReadyTime(s[j]), podReadyTime(s[i]))
if podutil.IsPodReady(s[i]) && podutil.IsPodReady(s[j]) {
readyTime1 := podReadyTime(s[i])
readyTime2 := podReadyTime(s[j])
if !readyTime1.Equal(readyTime2) {
return afterOrZero(readyTime2, readyTime1)
}
}
// 5. Pods with containers with higher restart counts < lower restart counts
if maxContainerRestarts(s[i]) != maxContainerRestarts(s[j]) {
@ -745,9 +751,8 @@ func (s ActivePods) Less(i, j int) bool {
return len(s[i].Spec.NodeName) == 0
}
// 2. PodPending < PodUnknown < PodRunning
m := map[v1.PodPhase]int{v1.PodPending: 0, v1.PodUnknown: 1, v1.PodRunning: 2}
if m[s[i].Status.Phase] != m[s[j].Status.Phase] {
return m[s[i].Status.Phase] < m[s[j].Status.Phase]
if podPhaseToOrdinal[s[i].Status.Phase] != podPhaseToOrdinal[s[j].Status.Phase] {
return podPhaseToOrdinal[s[i].Status.Phase] < podPhaseToOrdinal[s[j].Status.Phase]
}
// 3. Not ready < ready
// If only one of the pods is not ready, the not ready one is smaller
@ -758,8 +763,12 @@ func (s ActivePods) Less(i, j int) bool {
// see https://github.com/kubernetes/kubernetes/issues/22065
// 4. Been ready for empty time < less time < more time
// If both pods are ready, the latest ready one is smaller
if podutil.IsPodReady(s[i]) && podutil.IsPodReady(s[j]) && !podReadyTime(s[i]).Equal(podReadyTime(s[j])) {
return afterOrZero(podReadyTime(s[i]), podReadyTime(s[j]))
if podutil.IsPodReady(s[i]) && podutil.IsPodReady(s[j]) {
readyTime1 := podReadyTime(s[i])
readyTime2 := podReadyTime(s[j])
if !readyTime1.Equal(readyTime2) {
return afterOrZero(readyTime1, readyTime2)
}
}
// 5. Pods with containers with higher restart counts < lower restart counts
if maxContainerRestarts(s[i]) != maxContainerRestarts(s[j]) {
@ -772,6 +781,97 @@ func (s ActivePods) Less(i, j int) bool {
return false
}
// ActivePodsWithRanks is a sortable list of pods and a list of corresponding
// ranks which will be considered during sorting. The two lists must have equal
// length. After sorting, the pods will be ordered as follows, applying each
// rule in turn until one matches:
//
// 1. If only one of the pods is assigned to a node, the pod that is not
// assigned comes before the pod that is.
// 2. If the pods' phases differ, a pending pod comes before a pod whose phase
// is unknown, and a pod whose phase is unknown comes before a running pod.
// 3. If exactly one of the pods is ready, the pod that is not ready comes
// before the ready pod.
// 4. If the pods' ranks differ, the pod with greater rank comes before the pod
// with lower rank.
// 5. If both pods are ready but have not been ready for the same amount of
// time, the pod that has been ready for a shorter amount of time comes
// before the pod that has been ready for longer.
// 6. If one pod has a container that has restarted more than any container in
// the other pod, the pod with the container with more restarts comes
// before the other pod.
// 7. If the pods' creation times differ, the pod that was created more recently
// comes before the older pod.
//
// If none of these rules matches, the second pod comes before the first pod.
//
// The intention of this ordering is to put pods that should be preferred for
// deletion first in the list.
type ActivePodsWithRanks struct {
// Pods is a list of pods.
Pods []*v1.Pod
// Rank is a ranking of pods. This ranking is used during sorting when
// comparing two pods that are both scheduled, in the same phase, and
// having the same ready status.
Rank []int
}
func (s ActivePodsWithRanks) Len() int {
return len(s.Pods)
}
func (s ActivePodsWithRanks) Swap(i, j int) {
s.Pods[i], s.Pods[j] = s.Pods[j], s.Pods[i]
s.Rank[i], s.Rank[j] = s.Rank[j], s.Rank[i]
}
// Less compares two pods with corresponding ranks and returns true if the first
// one should be preferred for deletion.
func (s ActivePodsWithRanks) Less(i, j int) bool {
// 1. Unassigned < assigned
// If only one of the pods is unassigned, the unassigned one is smaller
if s.Pods[i].Spec.NodeName != s.Pods[j].Spec.NodeName && (len(s.Pods[i].Spec.NodeName) == 0 || len(s.Pods[j].Spec.NodeName) == 0) {
return len(s.Pods[i].Spec.NodeName) == 0
}
// 2. PodPending < PodUnknown < PodRunning
if podPhaseToOrdinal[s.Pods[i].Status.Phase] != podPhaseToOrdinal[s.Pods[j].Status.Phase] {
return podPhaseToOrdinal[s.Pods[i].Status.Phase] < podPhaseToOrdinal[s.Pods[j].Status.Phase]
}
// 3. Not ready < ready
// If only one of the pods is not ready, the not ready one is smaller
if podutil.IsPodReady(s.Pods[i]) != podutil.IsPodReady(s.Pods[j]) {
return !podutil.IsPodReady(s.Pods[i])
}
// 4. Doubled up < not doubled up
// If one of the two pods is on the same node as one or more additional
// ready pods that belong to the same replicaset, whichever pod has more
// colocated ready pods is less
if s.Rank[i] != s.Rank[j] {
return s.Rank[i] > s.Rank[j]
}
// TODO: take availability into account when we push minReadySeconds information from deployment into pods,
// see https://github.com/kubernetes/kubernetes/issues/22065
// 5. Been ready for empty time < less time < more time
// If both pods are ready, the latest ready one is smaller
if podutil.IsPodReady(s.Pods[i]) && podutil.IsPodReady(s.Pods[j]) {
readyTime1 := podReadyTime(s.Pods[i])
readyTime2 := podReadyTime(s.Pods[j])
if !readyTime1.Equal(readyTime2) {
return afterOrZero(readyTime1, readyTime2)
}
}
// 6. Pods with containers with higher restart counts < lower restart counts
if maxContainerRestarts(s.Pods[i]) != maxContainerRestarts(s.Pods[j]) {
return maxContainerRestarts(s.Pods[i]) > maxContainerRestarts(s.Pods[j])
}
// 7. Empty creation time pods < newer pods < older pods
if !s.Pods[i].CreationTimestamp.Equal(&s.Pods[j].CreationTimestamp) {
return afterOrZero(&s.Pods[i].CreationTimestamp, &s.Pods[j].CreationTimestamp)
}
return false
}
// afterOrZero checks if time t1 is after time t2; if one of them
// is zero, the zero time is seen as after non-zero time.
func afterOrZero(t1, t2 *metav1.Time) bool {
@ -1022,21 +1122,6 @@ func PatchNodeTaints(c clientset.Interface, nodeName string, oldNode *v1.Node, n
return err
}
// WaitForCacheSync is a wrapper around cache.WaitForCacheSync that generates log messages
// indicating that the controller identified by controllerName is waiting for syncs, followed by
// either a successful or failed sync.
func WaitForCacheSync(controllerName string, stopCh <-chan struct{}, cacheSyncs ...cache.InformerSynced) bool {
klog.Infof("Waiting for caches to sync for %s controller", controllerName)
if !cache.WaitForCacheSync(stopCh, cacheSyncs...) {
utilruntime.HandleError(fmt.Errorf("unable to sync caches for %s controller", controllerName))
return false
}
klog.Infof("Caches are synced for %s controller", controllerName)
return true
}
// ComputeHash returns a hash value calculated from pod template and
// a collisionCount to avoid hash collision. The hash will be safe encoded to
// avoid bad words.

18
vendor/k8s.io/kubernetes/pkg/controller/informer_factory.go generated vendored
View File

@ -18,8 +18,8 @@ package controller
import (
"k8s.io/apimachinery/pkg/runtime/schema"
"k8s.io/client-go/dynamic/dynamicinformer"
"k8s.io/client-go/informers"
"k8s.io/client-go/metadata/metadatainformer"
)
// InformerFactory creates informers for each group version resource.
@ -29,28 +29,28 @@ type InformerFactory interface {
}
type informerFactory struct {
typedInformerFactory informers.SharedInformerFactory
dynamicInformerFactory dynamicinformer.DynamicSharedInformerFactory
typedInformerFactory informers.SharedInformerFactory
metadataInformerFactory metadatainformer.SharedInformerFactory
}
func (i *informerFactory) ForResource(resource schema.GroupVersionResource) (informers.GenericInformer, error) {
informer, err := i.typedInformerFactory.ForResource(resource)
if err != nil {
return i.dynamicInformerFactory.ForResource(resource), nil
return i.metadataInformerFactory.ForResource(resource), nil
}
return informer, nil
}
func (i *informerFactory) Start(stopCh <-chan struct{}) {
i.typedInformerFactory.Start(stopCh)
i.dynamicInformerFactory.Start(stopCh)
i.metadataInformerFactory.Start(stopCh)
}
// NewInformerFactory creates a new InformerFactory which works with both typed
// resources and dynamic resources
func NewInformerFactory(typedInformerFactory informers.SharedInformerFactory, dynamicInformerFactory dynamicinformer.DynamicSharedInformerFactory) InformerFactory {
// resources and metadata-only resources
func NewInformerFactory(typedInformerFactory informers.SharedInformerFactory, metadataInformerFactory metadatainformer.SharedInformerFactory) InformerFactory {
return &informerFactory{
typedInformerFactory: typedInformerFactory,
dynamicInformerFactory: dynamicInformerFactory,
typedInformerFactory: typedInformerFactory,
metadataInformerFactory: metadataInformerFactory,
}
}

55
vendor/k8s.io/kubernetes/pkg/controller/nodelifecycle/metrics.go generated vendored
View File

@ -19,7 +19,8 @@ package nodelifecycle
import (
"sync"
"github.com/prometheus/client_golang/prometheus"
"k8s.io/component-base/metrics"
"k8s.io/component-base/metrics/legacyregistry"
)
const (
@ -31,35 +32,39 @@ const (
)
var (
zoneHealth = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Subsystem: nodeControllerSubsystem,
Name: zoneHealthStatisticKey,
Help: "Gauge measuring percentage of healthy nodes per zone.",
zoneHealth = metrics.NewGaugeVec(
&metrics.GaugeOpts{
Subsystem: nodeControllerSubsystem,
Name: zoneHealthStatisticKey,
Help: "Gauge measuring percentage of healthy nodes per zone.",
StabilityLevel: metrics.ALPHA,
},
[]string{"zone"},
)
zoneSize = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Subsystem: nodeControllerSubsystem,
Name: zoneSizeKey,
Help: "Gauge measuring number of registered Nodes per zones.",
zoneSize = metrics.NewGaugeVec(
&metrics.GaugeOpts{
Subsystem: nodeControllerSubsystem,
Name: zoneSizeKey,
Help: "Gauge measuring number of registered Nodes per zones.",
StabilityLevel: metrics.ALPHA,
},
[]string{"zone"},
)
unhealthyNodes = prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Subsystem: nodeControllerSubsystem,
Name: zoneNoUnhealthyNodesKey,
Help: "Gauge measuring number of not Ready Nodes per zones.",
unhealthyNodes = metrics.NewGaugeVec(
&metrics.GaugeOpts{
Subsystem: nodeControllerSubsystem,
Name: zoneNoUnhealthyNodesKey,
Help: "Gauge measuring number of not Ready Nodes per zones.",
StabilityLevel: metrics.ALPHA,
},
[]string{"zone"},
)
evictionsNumber = prometheus.NewCounterVec(
prometheus.CounterOpts{
Subsystem: nodeControllerSubsystem,
Name: evictionsNumberKey,
Help: "Number of Node evictions that happened since current instance of NodeController started.",
evictionsNumber = metrics.NewCounterVec(
&metrics.CounterOpts{
Subsystem: nodeControllerSubsystem,
Name: evictionsNumberKey,
Help: "Number of Node evictions that happened since current instance of NodeController started.",
StabilityLevel: metrics.ALPHA,
},
[]string{"zone"},
)
@ -70,9 +75,9 @@ var registerMetrics sync.Once
// Register the metrics that are to be monitored.
func Register() {
registerMetrics.Do(func() {
prometheus.MustRegister(zoneHealth)
prometheus.MustRegister(zoneSize)
prometheus.MustRegister(unhealthyNodes)
prometheus.MustRegister(evictionsNumber)
legacyregistry.MustRegister(zoneHealth)
legacyregistry.MustRegister(zoneSize)
legacyregistry.MustRegister(unhealthyNodes)
legacyregistry.MustRegister(evictionsNumber)
})
}

725
vendor/k8s.io/kubernetes/pkg/controller/nodelifecycle/node_lifecycle_controller.go generated vendored
View File

File diff suppressed because it is too large Load Diff

69
vendor/k8s.io/kubernetes/pkg/controller/nodelifecycle/scheduler/taint_manager.go generated vendored
View File

@ -20,14 +20,13 @@ import (
"fmt"
"hash/fnv"
"io"
"math"
"sync"
"time"
"k8s.io/api/core/v1"
apierrors "k8s.io/apimachinery/pkg/api/errors"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/fields"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/apimachinery/pkg/types"
utilruntime "k8s.io/apimachinery/pkg/util/runtime"
clientset "k8s.io/client-go/kubernetes"
@ -43,7 +42,7 @@ import (
const (
// TODO (k82cn): Figure out a reasonable number of workers/channels and propagate
// the number of workers up making it a paramater of Run() function.
// the number of workers up making it a parameter of Run() function.
// NodeUpdateChannelSize defines the size of channel for node update events.
NodeUpdateChannelSize = 10
@ -75,13 +74,17 @@ type GetPodFunc func(name, namespace string) (*v1.Pod, error)
// GetNodeFunc returns the node for the specified name, or a NotFound error if missing.
type GetNodeFunc func(name string) (*v1.Node, error)
// GetPodsByNodeNameFunc returns the list of pods assigned to the specified node.
type GetPodsByNodeNameFunc func(nodeName string) ([]*v1.Pod, error)
// NoExecuteTaintManager listens to Taint/Toleration changes and is responsible for removing Pods
// from Nodes tainted with NoExecute Taints.
type NoExecuteTaintManager struct {
client clientset.Interface
recorder record.EventRecorder
getPod GetPodFunc
getNode GetNodeFunc
client clientset.Interface
recorder record.EventRecorder
getPod GetPodFunc
getNode GetNodeFunc
getPodsAssignedToNode GetPodsByNodeNameFunc
taintEvictionQueue *TimedWorkerQueue
// keeps a map from nodeName to all noExecute taints on that Node
@ -125,28 +128,9 @@ func getNoExecuteTaints(taints []v1.Taint) []v1.Taint {
return result
}
func getPodsAssignedToNode(c clientset.Interface, nodeName string) ([]v1.Pod, error) {
selector := fields.SelectorFromSet(fields.Set{"spec.nodeName": nodeName})
pods, err := c.CoreV1().Pods(v1.NamespaceAll).List(metav1.ListOptions{
FieldSelector: selector.String(),
LabelSelector: labels.Everything().String(),
})
for i := 0; i < retries && err != nil; i++ {
pods, err = c.CoreV1().Pods(v1.NamespaceAll).List(metav1.ListOptions{
FieldSelector: selector.String(),
LabelSelector: labels.Everything().String(),
})
time.Sleep(100 * time.Millisecond)
}
if err != nil {
return []v1.Pod{}, fmt.Errorf("failed to get Pods assigned to node %v", nodeName)
}
return pods.Items, nil
}
// getMinTolerationTime returns minimal toleration time from the given slice, or -1 if it's infinite.
func getMinTolerationTime(tolerations []v1.Toleration) time.Duration {
minTolerationTime := int64(-1)
minTolerationTime := int64(math.MaxInt64)
if len(tolerations) == 0 {
return 0
}
@ -156,18 +140,21 @@ func getMinTolerationTime(tolerations []v1.Toleration) time.Duration {
tolerationSeconds := *(tolerations[i].TolerationSeconds)
if tolerationSeconds <= 0 {
return 0
} else if tolerationSeconds < minTolerationTime || minTolerationTime == -1 {
} else if tolerationSeconds < minTolerationTime {
minTolerationTime = tolerationSeconds
}
}
}
if minTolerationTime == int64(math.MaxInt64) {
return -1
}
return time.Duration(minTolerationTime) * time.Second
}
// NewNoExecuteTaintManager creates a new NoExecuteTaintManager that will use passed clientset to
// communicate with the API server.
func NewNoExecuteTaintManager(c clientset.Interface, getPod GetPodFunc, getNode GetNodeFunc) *NoExecuteTaintManager {
func NewNoExecuteTaintManager(c clientset.Interface, getPod GetPodFunc, getNode GetNodeFunc, getPodsAssignedToNode GetPodsByNodeNameFunc) *NoExecuteTaintManager {
eventBroadcaster := record.NewBroadcaster()
recorder := eventBroadcaster.NewRecorder(scheme.Scheme, v1.EventSource{Component: "taint-controller"})
eventBroadcaster.StartLogging(klog.Infof)
@ -179,11 +166,12 @@ func NewNoExecuteTaintManager(c clientset.Interface, getPod GetPodFunc, getNode
}
tm := &NoExecuteTaintManager{
client: c,
recorder: recorder,
getPod: getPod,
getNode: getNode,
taintedNodes: make(map[string][]v1.Taint),
client: c,
recorder: recorder,
getPod: getPod,
getNode: getNode,
getPodsAssignedToNode: getPodsAssignedToNode,
taintedNodes: make(map[string][]v1.Taint),
nodeUpdateQueue: workqueue.NewNamed("noexec_taint_node"),
podUpdateQueue: workqueue.NewNamed("noexec_taint_pod"),
@ -228,6 +216,10 @@ func (tc *NoExecuteTaintManager) Run(stopCh <-chan struct{}) {
if shutdown {
break
}
// The fact that pods are processed by the same worker as nodes is used to avoid races
// between node worker setting tc.taintedNodes and pod worker reading this to decide
// whether to delete pod.
// It's possible that even without this assumption this code is still correct.
podUpdate := item.(podUpdateItem)
hash := hash(podUpdate.nodeName, UpdateWorkerSize)
select {
@ -450,7 +442,11 @@ func (tc *NoExecuteTaintManager) handleNodeUpdate(nodeUpdate nodeUpdateItem) {
tc.taintedNodes[node.Name] = taints
}
}()
pods, err := getPodsAssignedToNode(tc.client, node.Name)
// This is critical that we update tc.taintedNodes before we call getPodsAssignedToNode:
// getPodsAssignedToNode can be delayed as long as all future updates to pods will call
// tc.PodUpdated which will use tc.taintedNodes to potentially delete delayed pods.
pods, err := tc.getPodsAssignedToNode(node.Name)
if err != nil {
klog.Errorf(err.Error())
return
@ -468,8 +464,7 @@ func (tc *NoExecuteTaintManager) handleNodeUpdate(nodeUpdate nodeUpdateItem) {
}
now := time.Now()
for i := range pods {
pod := &pods[i]
for _, pod := range pods {
podNamespacedName := types.NamespacedName{Namespace: pod.Namespace, Name: pod.Name}
tc.processPodOnNode(podNamespacedName, node.Name, pod.Spec.Tolerations, taints, now)
}

19
vendor/k8s.io/kubernetes/pkg/controller/service/doc.go generated vendored
View File

@ -1,19 +0,0 @@
/*
Copyright 2015 The Kubernetes 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 service contains code for syncing cloud load balancers
// with the service registry.
package service // import "k8s.io/kubernetes/pkg/controller/service"

59
vendor/k8s.io/kubernetes/pkg/controller/service/patch.go generated vendored
View File

@ -1,59 +0,0 @@
/*
Copyright 2019 The Kubernetes 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 service
import (
"encoding/json"
"fmt"
v1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/types"
"k8s.io/apimachinery/pkg/util/strategicpatch"
v1core "k8s.io/client-go/kubernetes/typed/core/v1"
)
// patch patches service's Status or ObjectMeta given the origin and
// updated ones. Change to spec will be ignored.
func patch(c v1core.CoreV1Interface, oldSvc *v1.Service, newSvc *v1.Service) (*v1.Service, error) {
// Reset spec to make sure only patch for Status or ObjectMeta.
newSvc.Spec = oldSvc.Spec
patchBytes, err := getPatchBytes(oldSvc, newSvc)
if err != nil {
return nil, err
}
return c.Services(oldSvc.Namespace).Patch(oldSvc.Name, types.StrategicMergePatchType, patchBytes, "status")
}
func getPatchBytes(oldSvc *v1.Service, newSvc *v1.Service) ([]byte, error) {
oldData, err := json.Marshal(oldSvc)
if err != nil {
return nil, fmt.Errorf("failed to Marshal oldData for svc %s/%s: %v", oldSvc.Namespace, oldSvc.Name, err)
}
newData, err := json.Marshal(newSvc)
if err != nil {
return nil, fmt.Errorf("failed to Marshal newData for svc %s/%s: %v", newSvc.Namespace, newSvc.Name, err)
}
patchBytes, err := strategicpatch.CreateTwoWayMergePatch(oldData, newData, v1.Service{})
if err != nil {
return nil, fmt.Errorf("failed to CreateTwoWayMergePatch for svc %s/%s: %v", oldSvc.Namespace, oldSvc.Name, err)
}
return patchBytes, nil
}

816
vendor/k8s.io/kubernetes/pkg/controller/service/service_controller.go generated vendored
View File

@ -1,816 +0,0 @@
/*
Copyright 2015 The Kubernetes 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 service
import (
"context"
"fmt"
"sync"
"time"
"reflect"
v1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/api/errors"
"k8s.io/apimachinery/pkg/util/runtime"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/apimachinery/pkg/util/wait"
utilfeature "k8s.io/apiserver/pkg/util/feature"
coreinformers "k8s.io/client-go/informers/core/v1"
clientset "k8s.io/client-go/kubernetes"
"k8s.io/client-go/kubernetes/scheme"
v1core "k8s.io/client-go/kubernetes/typed/core/v1"
corelisters "k8s.io/client-go/listers/core/v1"
"k8s.io/client-go/tools/cache"
"k8s.io/client-go/tools/record"
"k8s.io/client-go/util/workqueue"
cloudprovider "k8s.io/cloud-provider"
servicehelper "k8s.io/cloud-provider/service/helpers"
"k8s.io/klog"
v1helper "k8s.io/kubernetes/pkg/apis/core/v1/helper"
"k8s.io/kubernetes/pkg/controller"
kubefeatures "k8s.io/kubernetes/pkg/features"
"k8s.io/kubernetes/pkg/util/metrics"
"k8s.io/kubernetes/pkg/util/slice"
)
const (
// Interval of synchronizing service status from apiserver
serviceSyncPeriod = 30 * time.Second
// Interval of synchronizing node status from apiserver
nodeSyncPeriod = 100 * time.Second
// How long to wait before retrying the processing of a service change.
// If this changes, the sleep in hack/jenkins/e2e.sh before downing a cluster
// should be changed appropriately.
minRetryDelay = 5 * time.Second
maxRetryDelay = 300 * time.Second
clientRetryCount = 5
clientRetryInterval = 5 * time.Second
// LabelNodeRoleMaster specifies that a node is a master
// It's copied over to kubeadm until it's merged in core: https://github.com/kubernetes/kubernetes/pull/39112
LabelNodeRoleMaster = "node-role.kubernetes.io/master"
// LabelNodeRoleExcludeBalancer specifies that the node should be
// exclude from load balancers created by a cloud provider.
LabelNodeRoleExcludeBalancer = "alpha.service-controller.kubernetes.io/exclude-balancer"
)
type cachedService struct {
// The cached state of the service
state *v1.Service
}
type serviceCache struct {
mu sync.Mutex // protects serviceMap
serviceMap map[string]*cachedService
}
// ServiceController keeps cloud provider service resources
// (like load balancers) in sync with the registry.
type ServiceController struct {
cloud cloudprovider.Interface
knownHosts []*v1.Node
servicesToUpdate []*v1.Service
kubeClient clientset.Interface
clusterName string
balancer cloudprovider.LoadBalancer
cache *serviceCache
serviceLister corelisters.ServiceLister
serviceListerSynced cache.InformerSynced
eventBroadcaster record.EventBroadcaster
eventRecorder record.EventRecorder
nodeLister corelisters.NodeLister
nodeListerSynced cache.InformerSynced
// services that need to be synced
queue workqueue.RateLimitingInterface
}
// New returns a new service controller to keep cloud provider service resources
// (like load balancers) in sync with the registry.
func New(
cloud cloudprovider.Interface,
kubeClient clientset.Interface,
serviceInformer coreinformers.ServiceInformer,
nodeInformer coreinformers.NodeInformer,
clusterName string,
) (*ServiceController, error) {
broadcaster := record.NewBroadcaster()
broadcaster.StartLogging(klog.Infof)
broadcaster.StartRecordingToSink(&v1core.EventSinkImpl{Interface: kubeClient.CoreV1().Events("")})
recorder := broadcaster.NewRecorder(scheme.Scheme, v1.EventSource{Component: "service-controller"})
if kubeClient != nil && kubeClient.CoreV1().RESTClient().GetRateLimiter() != nil {
if err := metrics.RegisterMetricAndTrackRateLimiterUsage("service_controller", kubeClient.CoreV1().RESTClient().GetRateLimiter()); err != nil {
return nil, err
}
}
s := &ServiceController{
cloud: cloud,
knownHosts: []*v1.Node{},
kubeClient: kubeClient,
clusterName: clusterName,
cache: &serviceCache{serviceMap: make(map[string]*cachedService)},
eventBroadcaster: broadcaster,
eventRecorder: recorder,
nodeLister: nodeInformer.Lister(),
nodeListerSynced: nodeInformer.Informer().HasSynced,
queue: workqueue.NewNamedRateLimitingQueue(workqueue.NewItemExponentialFailureRateLimiter(minRetryDelay, maxRetryDelay), "service"),
}
serviceInformer.Informer().AddEventHandlerWithResyncPeriod(
cache.ResourceEventHandlerFuncs{
AddFunc: func(cur interface{}) {
svc, ok := cur.(*v1.Service)
if ok && (wantsLoadBalancer(svc) || needsCleanup(svc)) {
s.enqueueService(cur)
}
},
UpdateFunc: func(old, cur interface{}) {
oldSvc, ok1 := old.(*v1.Service)
curSvc, ok2 := cur.(*v1.Service)
if ok1 && ok2 && (s.needsUpdate(oldSvc, curSvc) || needsCleanup(curSvc)) {
s.enqueueService(cur)
}
},
DeleteFunc: func(old interface{}) {
if utilfeature.DefaultFeatureGate.Enabled(kubefeatures.ServiceLoadBalancerFinalizer) {
// No need to handle deletion event if finalizer feature gate is
// enabled. Because the deletion would be handled by the update
// path when the deletion timestamp is added.
return
}
s.enqueueService(old)
},
},
serviceSyncPeriod,
)
s.serviceLister = serviceInformer.Lister()
s.serviceListerSynced = serviceInformer.Informer().HasSynced
if err := s.init(); err != nil {
return nil, err
}
return s, nil
}
// obj could be an *v1.Service, or a DeletionFinalStateUnknown marker item.
func (s *ServiceController) enqueueService(obj interface{}) {
key, err := controller.KeyFunc(obj)
if err != nil {
runtime.HandleError(fmt.Errorf("couldn't get key for object %#v: %v", obj, err))
return
}
s.queue.Add(key)
}
// Run starts a background goroutine that watches for changes to services that
// have (or had) LoadBalancers=true and ensures that they have
// load balancers created and deleted appropriately.
// serviceSyncPeriod controls how often we check the cluster's services to
// ensure that the correct load balancers exist.
// nodeSyncPeriod controls how often we check the cluster's nodes to determine
// if load balancers need to be updated to point to a new set.
//
// It's an error to call Run() more than once for a given ServiceController
// object.
func (s *ServiceController) Run(stopCh <-chan struct{}, workers int) {
defer runtime.HandleCrash()
defer s.queue.ShutDown()
klog.Info("Starting service controller")
defer klog.Info("Shutting down service controller")
if !controller.WaitForCacheSync("service", stopCh, s.serviceListerSynced, s.nodeListerSynced) {
return
}
for i := 0; i < workers; i++ {
go wait.Until(s.worker, time.Second, stopCh)
}
go wait.Until(s.nodeSyncLoop, nodeSyncPeriod, stopCh)
<-stopCh
}
// worker runs a worker thread that just dequeues items, processes them, and marks them done.
// It enforces that the syncHandler is never invoked concurrently with the same key.
func (s *ServiceController) worker() {
for s.processNextWorkItem() {
}
}
func (s *ServiceController) processNextWorkItem() bool {
key, quit := s.queue.Get()
if quit {
return false
}
defer s.queue.Done(key)
err := s.syncService(key.(string))
if err == nil {
s.queue.Forget(key)
return true
}
runtime.HandleError(fmt.Errorf("error processing service %v (will retry): %v", key, err))
s.queue.AddRateLimited(key)
return true
}
func (s *ServiceController) init() error {
if s.cloud == nil {
return fmt.Errorf("WARNING: no cloud provider provided, services of type LoadBalancer will fail")
}
balancer, ok := s.cloud.LoadBalancer()
if !ok {
return fmt.Errorf("the cloud provider does not support external load balancers")
}
s.balancer = balancer
return nil
}
// processServiceCreateOrUpdate operates loadbalancers for the incoming service accordingly.
// Returns an error if processing the service update failed.
func (s *ServiceController) processServiceCreateOrUpdate(service *v1.Service, key string) error {
// TODO(@MrHohn): Remove the cache once we get rid of the non-finalizer deletion
// path. Ref https://github.com/kubernetes/enhancements/issues/980.
cachedService := s.cache.getOrCreate(key)
if cachedService.state != nil && cachedService.state.UID != service.UID {
// This happens only when a service is deleted and re-created
// in a short period, which is only possible when it doesn't
// contain finalizer.
if err := s.processLoadBalancerDelete(cachedService.state, key); err != nil {
return err
}
}
// Always cache the service, we need the info for service deletion in case
// when load balancer cleanup is not handled via finalizer.
cachedService.state = service
op, err := s.syncLoadBalancerIfNeeded(service, key)
if err != nil {
s.eventRecorder.Eventf(service, v1.EventTypeWarning, "SyncLoadBalancerFailed", "Error syncing load balancer: %v", err)
return err
}
if op == deleteLoadBalancer {
// Only delete the cache upon successful load balancer deletion.
s.cache.delete(key)
}
return nil
}
type loadBalancerOperation int
const (
deleteLoadBalancer loadBalancerOperation = iota
ensureLoadBalancer
)
// syncLoadBalancerIfNeeded ensures that service's status is synced up with loadbalancer
// i.e. creates loadbalancer for service if requested and deletes loadbalancer if the service
// doesn't want a loadbalancer no more. Returns whatever error occurred.
func (s *ServiceController) syncLoadBalancerIfNeeded(service *v1.Service, key string) (loadBalancerOperation, error) {
// Note: It is safe to just call EnsureLoadBalancer. But, on some clouds that requires a delete & create,
// which may involve service interruption. Also, we would like user-friendly events.
// Save the state so we can avoid a write if it doesn't change
previousStatus := v1helper.LoadBalancerStatusDeepCopy(&service.Status.LoadBalancer)
var newStatus *v1.LoadBalancerStatus
var op loadBalancerOperation
var err error
if !wantsLoadBalancer(service) || needsCleanup(service) {
// Delete the load balancer if service no longer wants one, or if service needs cleanup.
op = deleteLoadBalancer
newStatus = &v1.LoadBalancerStatus{}
_, exists, err := s.balancer.GetLoadBalancer(context.TODO(), s.clusterName, service)
if err != nil {
return op, fmt.Errorf("failed to check if load balancer exists before cleanup: %v", err)
}
if exists {
klog.V(2).Infof("Deleting existing load balancer for service %s", key)
s.eventRecorder.Event(service, v1.EventTypeNormal, "DeletingLoadBalancer", "Deleting load balancer")
if err := s.balancer.EnsureLoadBalancerDeleted(context.TODO(), s.clusterName, service); err != nil {
return op, fmt.Errorf("failed to delete load balancer: %v", err)
}
}
// Always try to remove finalizer when load balancer is deleted.
// It will be a no-op if finalizer does not exist.
// Note this also clears up finalizer if the cluster is downgraded
// from a version that attaches finalizer to a version that doesn't.
if err := s.removeFinalizer(service); err != nil {
return op, fmt.Errorf("failed to remove load balancer cleanup finalizer: %v", err)
}
s.eventRecorder.Event(service, v1.EventTypeNormal, "DeletedLoadBalancer", "Deleted load balancer")
} else {
// Create or update the load balancer if service wants one.
op = ensureLoadBalancer
klog.V(2).Infof("Ensuring load balancer for service %s", key)
s.eventRecorder.Event(service, v1.EventTypeNormal, "EnsuringLoadBalancer", "Ensuring load balancer")
if utilfeature.DefaultFeatureGate.Enabled(kubefeatures.ServiceLoadBalancerFinalizer) {
// Always try to add finalizer prior to load balancer creation.
// It will be a no-op if finalizer already exists.
// Note this also retrospectively puts on finalizer if the cluster
// is upgraded from a version that doesn't attach finalizer to a
// version that does.
if err := s.addFinalizer(service); err != nil {
return op, fmt.Errorf("failed to add load balancer cleanup finalizer: %v", err)
}
}
newStatus, err = s.ensureLoadBalancer(service)
if err != nil {
return op, fmt.Errorf("failed to ensure load balancer: %v", err)
}
s.eventRecorder.Event(service, v1.EventTypeNormal, "EnsuredLoadBalancer", "Ensured load balancer")
}
if err := s.patchStatus(service, previousStatus, newStatus); err != nil {
// Only retry error that isn't not found:
// - Not found error mostly happens when service disappears right after
// we remove the finalizer.
// - We can't patch status on non-exist service anyway.
if !errors.IsNotFound(err) {
return op, fmt.Errorf("failed to update load balancer status: %v", err)
}
}
return op, nil
}
func (s *ServiceController) ensureLoadBalancer(service *v1.Service) (*v1.LoadBalancerStatus, error) {
nodes, err := s.nodeLister.ListWithPredicate(getNodeConditionPredicate())
if err != nil {
return nil, err
}
// If there are no available nodes for LoadBalancer service, make a EventTypeWarning event for it.
if len(nodes) == 0 {
s.eventRecorder.Event(service, v1.EventTypeWarning, "UnAvailableLoadBalancer", "There are no available nodes for LoadBalancer")
}
// - Only one protocol supported per service
// - Not all cloud providers support all protocols and the next step is expected to return
// an error for unsupported protocols
return s.balancer.EnsureLoadBalancer(context.TODO(), s.clusterName, service, nodes)
}
// ListKeys implements the interface required by DeltaFIFO to list the keys we
// already know about.
func (s *serviceCache) ListKeys() []string {
s.mu.Lock()
defer s.mu.Unlock()
keys := make([]string, 0, len(s.serviceMap))
for k := range s.serviceMap {
keys = append(keys, k)
}
return keys
}
// GetByKey returns the value stored in the serviceMap under the given key
func (s *serviceCache) GetByKey(key string) (interface{}, bool, error) {
s.mu.Lock()
defer s.mu.Unlock()
if v, ok := s.serviceMap[key]; ok {
return v, true, nil
}
return nil, false, nil
}
// ListKeys implements the interface required by DeltaFIFO to list the keys we
// already know about.
func (s *serviceCache) allServices() []*v1.Service {
s.mu.Lock()
defer s.mu.Unlock()
services := make([]*v1.Service, 0, len(s.serviceMap))
for _, v := range s.serviceMap {
services = append(services, v.state)
}
return services
}
func (s *serviceCache) get(serviceName string) (*cachedService, bool) {
s.mu.Lock()
defer s.mu.Unlock()
service, ok := s.serviceMap[serviceName]
return service, ok
}
func (s *serviceCache) getOrCreate(serviceName string) *cachedService {
s.mu.Lock()
defer s.mu.Unlock()
service, ok := s.serviceMap[serviceName]
if !ok {
service = &cachedService{}
s.serviceMap[serviceName] = service
}
return service
}
func (s *serviceCache) set(serviceName string, service *cachedService) {
s.mu.Lock()
defer s.mu.Unlock()
s.serviceMap[serviceName] = service
}
func (s *serviceCache) delete(serviceName string) {
s.mu.Lock()
defer s.mu.Unlock()
delete(s.serviceMap, serviceName)
}
// needsCleanup checks if load balancer needs to be cleaned up as indicated by finalizer.
func needsCleanup(service *v1.Service) bool {
return service.ObjectMeta.DeletionTimestamp != nil && servicehelper.HasLBFinalizer(service)
}
// needsUpdate checks if load balancer needs to be updated due to change in attributes.
func (s *ServiceController) needsUpdate(oldService *v1.Service, newService *v1.Service) bool {
if !wantsLoadBalancer(oldService) && !wantsLoadBalancer(newService) {
return false
}
if wantsLoadBalancer(oldService) != wantsLoadBalancer(newService) {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "Type", "%v -> %v",
oldService.Spec.Type, newService.Spec.Type)
return true
}
if wantsLoadBalancer(newService) && !reflect.DeepEqual(oldService.Spec.LoadBalancerSourceRanges, newService.Spec.LoadBalancerSourceRanges) {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "LoadBalancerSourceRanges", "%v -> %v",
oldService.Spec.LoadBalancerSourceRanges, newService.Spec.LoadBalancerSourceRanges)
return true
}
if !portsEqualForLB(oldService, newService) || oldService.Spec.SessionAffinity != newService.Spec.SessionAffinity {
return true
}
if !loadBalancerIPsAreEqual(oldService, newService) {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "LoadbalancerIP", "%v -> %v",
oldService.Spec.LoadBalancerIP, newService.Spec.LoadBalancerIP)
return true
}
if len(oldService.Spec.ExternalIPs) != len(newService.Spec.ExternalIPs) {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "ExternalIP", "Count: %v -> %v",
len(oldService.Spec.ExternalIPs), len(newService.Spec.ExternalIPs))
return true
}
for i := range oldService.Spec.ExternalIPs {
if oldService.Spec.ExternalIPs[i] != newService.Spec.ExternalIPs[i] {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "ExternalIP", "Added: %v",
newService.Spec.ExternalIPs[i])
return true
}
}
if !reflect.DeepEqual(oldService.Annotations, newService.Annotations) {
return true
}
if oldService.UID != newService.UID {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "UID", "%v -> %v",
oldService.UID, newService.UID)
return true
}
if oldService.Spec.ExternalTrafficPolicy != newService.Spec.ExternalTrafficPolicy {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "ExternalTrafficPolicy", "%v -> %v",
oldService.Spec.ExternalTrafficPolicy, newService.Spec.ExternalTrafficPolicy)
return true
}
if oldService.Spec.HealthCheckNodePort != newService.Spec.HealthCheckNodePort {
s.eventRecorder.Eventf(newService, v1.EventTypeNormal, "HealthCheckNodePort", "%v -> %v",
oldService.Spec.HealthCheckNodePort, newService.Spec.HealthCheckNodePort)
return true
}
return false
}
func (s *ServiceController) loadBalancerName(service *v1.Service) string {
return s.balancer.GetLoadBalancerName(context.TODO(), "", service)
}
func getPortsForLB(service *v1.Service) ([]*v1.ServicePort, error) {
var protocol v1.Protocol
ports := []*v1.ServicePort{}
for i := range service.Spec.Ports {
sp := &service.Spec.Ports[i]
// The check on protocol was removed here. The cloud provider itself is now responsible for all protocol validation
ports = append(ports, sp)
if protocol == "" {
protocol = sp.Protocol
} else if protocol != sp.Protocol && wantsLoadBalancer(service) {
// TODO: Convert error messages to use event recorder
return nil, fmt.Errorf("mixed protocol external load balancers are not supported")
}
}
return ports, nil
}
func portsEqualForLB(x, y *v1.Service) bool {
xPorts, err := getPortsForLB(x)
if err != nil {
return false
}
yPorts, err := getPortsForLB(y)
if err != nil {
return false
}
return portSlicesEqualForLB(xPorts, yPorts)
}
func portSlicesEqualForLB(x, y []*v1.ServicePort) bool {
if len(x) != len(y) {
return false
}
for i := range x {
if !portEqualForLB(x[i], y[i]) {
return false
}
}
return true
}
func portEqualForLB(x, y *v1.ServicePort) bool {
// TODO: Should we check name? (In theory, an LB could expose it)
if x.Name != y.Name {
return false
}
if x.Protocol != y.Protocol {
return false
}
if x.Port != y.Port {
return false
}
if x.NodePort != y.NodePort {
return false
}
// We don't check TargetPort; that is not relevant for load balancing
// TODO: Should we blank it out? Or just check it anyway?
return true
}
func nodeNames(nodes []*v1.Node) sets.String {
ret := sets.NewString()
for _, node := range nodes {
ret.Insert(node.Name)
}
return ret
}
func nodeSlicesEqualForLB(x, y []*v1.Node) bool {
if len(x) != len(y) {
return false
}
return nodeNames(x).Equal(nodeNames(y))
}
func getNodeConditionPredicate() corelisters.NodeConditionPredicate {
return func(node *v1.Node) bool {
// We add the master to the node list, but its unschedulable. So we use this to filter
// the master.
if node.Spec.Unschedulable {
return false
}
// As of 1.6, we will taint the master, but not necessarily mark it unschedulable.
// Recognize nodes labeled as master, and filter them also, as we were doing previously.
if _, hasMasterRoleLabel := node.Labels[LabelNodeRoleMaster]; hasMasterRoleLabel {
return false
}
if utilfeature.DefaultFeatureGate.Enabled(kubefeatures.ServiceNodeExclusion) {
if _, hasExcludeBalancerLabel := node.Labels[LabelNodeRoleExcludeBalancer]; hasExcludeBalancerLabel {
return false
}
}
// If we have no info, don't accept
if len(node.Status.Conditions) == 0 {
return false
}
for _, cond := range node.Status.Conditions {
// We consider the node for load balancing only when its NodeReady condition status
// is ConditionTrue
if cond.Type == v1.NodeReady && cond.Status != v1.ConditionTrue {
klog.V(4).Infof("Ignoring node %v with %v condition status %v", node.Name, cond.Type, cond.Status)
return false
}
}
return true
}
}
// nodeSyncLoop handles updating the hosts pointed to by all load
// balancers whenever the set of nodes in the cluster changes.
func (s *ServiceController) nodeSyncLoop() {
newHosts, err := s.nodeLister.ListWithPredicate(getNodeConditionPredicate())
if err != nil {
runtime.HandleError(fmt.Errorf("Failed to retrieve current set of nodes from node lister: %v", err))
return
}
if nodeSlicesEqualForLB(newHosts, s.knownHosts) {
// The set of nodes in the cluster hasn't changed, but we can retry
// updating any services that we failed to update last time around.
s.servicesToUpdate = s.updateLoadBalancerHosts(s.servicesToUpdate, newHosts)
return
}
klog.V(2).Infof("Detected change in list of current cluster nodes. New node set: %v",
nodeNames(newHosts))
// Try updating all services, and save the ones that fail to try again next
// round.
s.servicesToUpdate = s.cache.allServices()
numServices := len(s.servicesToUpdate)
s.servicesToUpdate = s.updateLoadBalancerHosts(s.servicesToUpdate, newHosts)
klog.V(2).Infof("Successfully updated %d out of %d load balancers to direct traffic to the updated set of nodes",
numServices-len(s.servicesToUpdate), numServices)
s.knownHosts = newHosts
}
// updateLoadBalancerHosts updates all existing load balancers so that
// they will match the list of hosts provided.
// Returns the list of services that couldn't be updated.
func (s *ServiceController) updateLoadBalancerHosts(services []*v1.Service, hosts []*v1.Node) (servicesToRetry []*v1.Service) {
for _, service := range services {
func() {
if service == nil {
return
}
if err := s.lockedUpdateLoadBalancerHosts(service, hosts); err != nil {
runtime.HandleError(fmt.Errorf("failed to update load balancer hosts for service %s/%s: %v", service.Namespace, service.Name, err))
servicesToRetry = append(servicesToRetry, service)
}
}()
}
return servicesToRetry
}
// Updates the load balancer of a service, assuming we hold the mutex
// associated with the service.
func (s *ServiceController) lockedUpdateLoadBalancerHosts(service *v1.Service, hosts []*v1.Node) error {
if !wantsLoadBalancer(service) {
return nil
}
// This operation doesn't normally take very long (and happens pretty often), so we only record the final event
err := s.balancer.UpdateLoadBalancer(context.TODO(), s.clusterName, service, hosts)
if err == nil {
// If there are no available nodes for LoadBalancer service, make a EventTypeWarning event for it.
if len(hosts) == 0 {
s.eventRecorder.Event(service, v1.EventTypeWarning, "UnAvailableLoadBalancer", "There are no available nodes for LoadBalancer")
} else {
s.eventRecorder.Event(service, v1.EventTypeNormal, "UpdatedLoadBalancer", "Updated load balancer with new hosts")
}
return nil
}
// It's only an actual error if the load balancer still exists.
if _, exists, err := s.balancer.GetLoadBalancer(context.TODO(), s.clusterName, service); err != nil {
runtime.HandleError(fmt.Errorf("failed to check if load balancer exists for service %s/%s: %v", service.Namespace, service.Name, err))
} else if !exists {
return nil
}
s.eventRecorder.Eventf(service, v1.EventTypeWarning, "UpdateLoadBalancerFailed", "Error updating load balancer with new hosts %v: %v", nodeNames(hosts), err)
return err
}
func wantsLoadBalancer(service *v1.Service) bool {
return service.Spec.Type == v1.ServiceTypeLoadBalancer
}
func loadBalancerIPsAreEqual(oldService, newService *v1.Service) bool {
return oldService.Spec.LoadBalancerIP == newService.Spec.LoadBalancerIP
}
// syncService will sync the Service with the given key if it has had its expectations fulfilled,
// meaning it did not expect to see any more of its pods created or deleted. This function is not meant to be
// invoked concurrently with the same key.
func (s *ServiceController) syncService(key string) error {
startTime := time.Now()
defer func() {
klog.V(4).Infof("Finished syncing service %q (%v)", key, time.Since(startTime))
}()
namespace, name, err := cache.SplitMetaNamespaceKey(key)
if err != nil {
return err
}
// service holds the latest service info from apiserver
service, err := s.serviceLister.Services(namespace).Get(name)
switch {
case errors.IsNotFound(err):
// service absence in store means watcher caught the deletion, ensure LB info is cleaned
err = s.processServiceDeletion(key)
case err != nil:
runtime.HandleError(fmt.Errorf("Unable to retrieve service %v from store: %v", key, err))
default:
err = s.processServiceCreateOrUpdate(service, key)
}
return err
}
func (s *ServiceController) processServiceDeletion(key string) error {
cachedService, ok := s.cache.get(key)
if !ok {
// Cache does not contains the key means:
// - We didn't create a Load Balancer for the deleted service at all.
// - We already deleted the Load Balancer that was created for the service.
// In both cases we have nothing left to do.
return nil
}
klog.V(2).Infof("Service %v has been deleted. Attempting to cleanup load balancer resources", key)
if err := s.processLoadBalancerDelete(cachedService.state, key); err != nil {
return err
}
s.cache.delete(key)
return nil
}
func (s *ServiceController) processLoadBalancerDelete(service *v1.Service, key string) error {
// delete load balancer info only if the service type is LoadBalancer
if !wantsLoadBalancer(service) {
return nil
}
s.eventRecorder.Event(service, v1.EventTypeNormal, "DeletingLoadBalancer", "Deleting load balancer")
if err := s.balancer.EnsureLoadBalancerDeleted(context.TODO(), s.clusterName, service); err != nil {
s.eventRecorder.Eventf(service, v1.EventTypeWarning, "DeleteLoadBalancerFailed", "Error deleting load balancer: %v", err)
return err
}
s.eventRecorder.Event(service, v1.EventTypeNormal, "DeletedLoadBalancer", "Deleted load balancer")
return nil
}
// addFinalizer patches the service to add finalizer.
func (s *ServiceController) addFinalizer(service *v1.Service) error {
if servicehelper.HasLBFinalizer(service) {
return nil
}
// Make a copy so we don't mutate the shared informer cache.
updated := service.DeepCopy()
updated.ObjectMeta.Finalizers = append(updated.ObjectMeta.Finalizers, servicehelper.LoadBalancerCleanupFinalizer)
klog.V(2).Infof("Adding finalizer to service %s/%s", updated.Namespace, updated.Name)
_, err := patch(s.kubeClient.CoreV1(), service, updated)
return err
}
// removeFinalizer patches the service to remove finalizer.
func (s *ServiceController) removeFinalizer(service *v1.Service) error {
if !servicehelper.HasLBFinalizer(service) {
return nil
}
// Make a copy so we don't mutate the shared informer cache.
updated := service.DeepCopy()
updated.ObjectMeta.Finalizers = slice.RemoveString(updated.ObjectMeta.Finalizers, servicehelper.LoadBalancerCleanupFinalizer, nil)
klog.V(2).Infof("Removing finalizer from service %s/%s", updated.Namespace, updated.Name)
_, err := patch(s.kubeClient.CoreV1(), service, updated)
return err
}
// patchStatus patches the service with the given LoadBalancerStatus.
func (s *ServiceController) patchStatus(service *v1.Service, previousStatus, newStatus *v1.LoadBalancerStatus) error {
if v1helper.LoadBalancerStatusEqual(previousStatus, newStatus) {
return nil
}
// Make a copy so we don't mutate the shared informer cache.
updated := service.DeepCopy()
updated.Status.LoadBalancer = *newStatus
klog.V(2).Infof("Patching status for service %s/%s", updated.Namespace, updated.Name)
_, err := patch(s.kubeClient.CoreV1(), service, updated)
return err
}

59
vendor/k8s.io/kubernetes/pkg/controller/util/node/controller_utils.go generated vendored
View File

@ -22,7 +22,6 @@ import (
apierrors "k8s.io/apimachinery/pkg/api/errors"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/fields"
"k8s.io/apimachinery/pkg/types"
utilerrors "k8s.io/apimachinery/pkg/util/errors"
utilruntime "k8s.io/apimachinery/pkg/util/runtime"
@ -33,7 +32,6 @@ import (
clientset "k8s.io/client-go/kubernetes"
appsv1listers "k8s.io/client-go/listers/apps/v1"
utilpod "k8s.io/kubernetes/pkg/api/v1/pod"
api "k8s.io/kubernetes/pkg/apis/core"
"k8s.io/kubernetes/pkg/controller"
"k8s.io/kubernetes/pkg/kubelet/util/format"
nodepkg "k8s.io/kubernetes/pkg/util/node"
@ -44,32 +42,27 @@ import (
// 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 appsv1listers.DaemonSetLister) (bool, error) {
func DeletePods(kubeClient clientset.Interface, pods []*v1.Pod, recorder record.EventRecorder, nodeName, nodeUID string, daemonStore appsv1listers.DaemonSetLister) (bool, error) {
remaining := false
selector := fields.OneTermEqualSelector(api.PodHostField, nodeName).String()
options := metav1.ListOptions{FieldSelector: selector}
pods, err := kubeClient.CoreV1().Pods(metav1.NamespaceAll).List(options)
var updateErrList []error
if err != nil {
return remaining, err
}
if len(pods.Items) > 0 {
if len(pods) > 0 {
RecordNodeEvent(recorder, nodeName, nodeUID, v1.EventTypeNormal, "DeletingAllPods", fmt.Sprintf("Deleting all Pods from Node %v.", nodeName))
}
for _, pod := range pods.Items {
for i := range pods {
// Defensive check, also needed for tests.
if pod.Spec.NodeName != nodeName {
if pods[i].Spec.NodeName != nodeName {
continue
}
// Pod will be modified, so making copy is required.
pod := pods[i].DeepCopy()
// Set reason and message in the pod object.
if _, err = SetPodTerminationReason(kubeClient, &pod, nodeName); err != nil {
if _, err := SetPodTerminationReason(kubeClient, pod, nodeName); err != nil {
if apierrors.IsConflict(err) {
updateErrList = append(updateErrList,
fmt.Errorf("update status failed for pod %q: %v", format.Pod(&pod), err))
fmt.Errorf("update status failed for pod %q: %v", format.Pod(pod), err))
continue
}
}
@ -79,14 +72,19 @@ func DeletePods(kubeClient clientset.Interface, recorder record.EventRecorder, n
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
if _, err := daemonStore.GetPodDaemonSets(pod); err == nil {
// No error means at least one daemonset was found
continue
}
klog.V(2).Infof("Starting deletion of pod %v/%v", pod.Namespace, pod.Name)
recorder.Eventf(&pod, v1.EventTypeNormal, "NodeControllerEviction", "Marking for deletion Pod %s from Node %s", pod.Name, nodeName)
recorder.Eventf(pod, v1.EventTypeNormal, "NodeControllerEviction", "Marking for deletion Pod %s from Node %s", pod.Name, nodeName)
if err := kubeClient.CoreV1().Pods(pod.Namespace).Delete(pod.Name, nil); err != nil {
if apierrors.IsNotFound(err) {
// NotFound error means that pod was already deleted.
// There is nothing left to do with this pod.
continue
}
return false, err
}
remaining = true
@ -117,24 +115,20 @@ func SetPodTerminationReason(kubeClient clientset.Interface, pod *v1.Pod, nodeNa
return updatedPod, nil
}
// MarkAllPodsNotReady updates ready status of all pods running on
// MarkPodsNotReady updates ready status of given pods running on
// given node from master return true if success
func MarkAllPodsNotReady(kubeClient clientset.Interface, node *v1.Node) error {
nodeName := node.Name
func MarkPodsNotReady(kubeClient clientset.Interface, pods []*v1.Pod, nodeName string) error {
klog.V(2).Infof("Update ready status of pods on node [%v]", nodeName)
opts := metav1.ListOptions{FieldSelector: fields.OneTermEqualSelector(api.PodHostField, nodeName).String()}
pods, err := kubeClient.CoreV1().Pods(metav1.NamespaceAll).List(opts)
if err != nil {
return err
}
errMsg := []string{}
for _, pod := range pods.Items {
for i := range pods {
// Defensive check, also needed for tests.
if pod.Spec.NodeName != nodeName {
if pods[i].Spec.NodeName != nodeName {
continue
}
// Pod will be modified, so making copy is required.
pod := pods[i].DeepCopy()
for _, cond := range pod.Status.Conditions {
if cond.Type == v1.PodReady {
cond.Status = v1.ConditionFalse
@ -142,9 +136,14 @@ func MarkAllPodsNotReady(kubeClient clientset.Interface, node *v1.Node) error {
break
}
klog.V(2).Infof("Updating ready status of pod %v to false", pod.Name)
_, err := kubeClient.CoreV1().Pods(pod.Namespace).UpdateStatus(&pod)
_, err := kubeClient.CoreV1().Pods(pod.Namespace).UpdateStatus(pod)
if err != nil {
klog.Warningf("Failed to update status for pod %q: %v", format.Pod(&pod), err)
if apierrors.IsNotFound(err) {
// NotFound error means that pod was already deleted.
// There is nothing left to do with this pod.
continue
}
klog.Warningf("Failed to update status for pod %q: %v", format.Pod(pod), err)
errMsg = append(errMsg, fmt.Sprintf("%v", err))
}
break

44
vendor/k8s.io/kubernetes/pkg/controller/volume/persistentvolume/util/util.go generated vendored
View File

@ -19,8 +19,9 @@ package persistentvolume
import (
"fmt"
"k8s.io/api/core/v1"
v1 "k8s.io/api/core/v1"
storage "k8s.io/api/storage/v1"
apierrors "k8s.io/apimachinery/pkg/api/errors"
"k8s.io/apimachinery/pkg/api/resource"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/labels"
@ -69,6 +70,16 @@ const (
AnnStorageProvisioner = "volume.beta.kubernetes.io/storage-provisioner"
)
// IsDelayBindingProvisioning checks if claim provisioning with selected-node annotation
func IsDelayBindingProvisioning(claim *v1.PersistentVolumeClaim) bool {
// When feature VolumeScheduling enabled,
// Scheduler signal to the PV controller to start dynamic
// provisioning by setting the "AnnSelectedNode" annotation
// in the PVC
_, ok := claim.Annotations[AnnSelectedNode]
return ok
}
// IsDelayBindingMode checks if claim is in delay binding mode.
func IsDelayBindingMode(claim *v1.PersistentVolumeClaim, classLister storagelisters.StorageClassLister) (bool, error) {
className := v1helper.GetPersistentVolumeClaimClass(claim)
@ -78,7 +89,10 @@ func IsDelayBindingMode(claim *v1.PersistentVolumeClaim, classLister storagelist
class, err := classLister.Get(className)
if err != nil {
return false, nil
if apierrors.IsNotFound(err) {
return false, nil
}
return false, err
}
if class.VolumeBindingMode == nil {
@ -193,13 +207,8 @@ func FindMatchingVolume(
volumeQty := volume.Spec.Capacity[v1.ResourceStorage]
// check if volumeModes do not match (feature gate protected)
isMismatch, err := CheckVolumeModeMismatches(&claim.Spec, &volume.Spec)
if err != nil {
return nil, fmt.Errorf("error checking if volumeMode was a mismatch: %v", err)
}
// filter out mismatching volumeModes
if isMismatch {
if CheckVolumeModeMismatches(&claim.Spec, &volume.Spec) {
continue
}
@ -295,9 +304,22 @@ func FindMatchingVolume(
// CheckVolumeModeMismatches is a convenience method that checks volumeMode for PersistentVolume
// and PersistentVolumeClaims
func CheckVolumeModeMismatches(pvcSpec *v1.PersistentVolumeClaimSpec, pvSpec *v1.PersistentVolumeSpec) (bool, error) {
func CheckVolumeModeMismatches(pvcSpec *v1.PersistentVolumeClaimSpec, pvSpec *v1.PersistentVolumeSpec) bool {
if !utilfeature.DefaultFeatureGate.Enabled(features.BlockVolume) {
return false, nil
if pvcSpec.VolumeMode != nil && *pvcSpec.VolumeMode == v1.PersistentVolumeBlock {
// Block PVC does not match anything when the feature is off. We explicitly want
// to prevent binding block PVC to filesystem PV.
// The PVC should be ignored by PV controller.
return true
}
if pvSpec.VolumeMode != nil && *pvSpec.VolumeMode == v1.PersistentVolumeBlock {
// Block PV does not match anything when the feature is off. We explicitly want
// to prevent binding block PV to filesystem PVC.
// The PV should be ignored by PV controller.
return true
}
// Both PV + PVC are not block.
return false
}
// In HA upgrades, we cannot guarantee that the apiserver is on a version >= controller-manager.
@ -310,7 +332,7 @@ func CheckVolumeModeMismatches(pvcSpec *v1.PersistentVolumeClaimSpec, pvSpec *v1
if pvSpec.VolumeMode != nil {
pvVolumeMode = *pvSpec.VolumeMode
}
return requestedVolumeMode != pvVolumeMode, nil
return requestedVolumeMode != pvVolumeMode
}
// CheckAccessModes returns true if PV satisfies all the PVC's requested AccessModes

46
vendor/k8s.io/kubernetes/pkg/controller/volume/scheduling/scheduler_bind_cache_metrics.go → vendor/k8s.io/kubernetes/pkg/controller/volume/scheduling/metrics/metrics.go generated vendored
View File

@ -14,10 +14,11 @@ See the License for the specific language governing permissions and
limitations under the License.
*/
package scheduling
package metrics
import (
"github.com/prometheus/client_golang/prometheus"
"k8s.io/component-base/metrics"
"k8s.io/component-base/metrics/legacyregistry"
)
// VolumeSchedulerSubsystem - subsystem name used by scheduler
@ -25,30 +26,33 @@ const VolumeSchedulerSubsystem = "scheduler_volume"
var (
// VolumeBindingRequestSchedulerBinderCache tracks the number of volume binder cache operations.
VolumeBindingRequestSchedulerBinderCache = prometheus.NewCounterVec(
prometheus.CounterOpts{
Subsystem: VolumeSchedulerSubsystem,
Name: "binder_cache_requests_total",
Help: "Total number for request volume binding cache",
VolumeBindingRequestSchedulerBinderCache = metrics.NewCounterVec(
&metrics.CounterOpts{
Subsystem: VolumeSchedulerSubsystem,
Name: "binder_cache_requests_total",
Help: "Total number for request volume binding cache",
StabilityLevel: metrics.ALPHA,
},
[]string{"operation"},
)
// VolumeSchedulingStageLatency tracks the latency of volume scheduling operations.
VolumeSchedulingStageLatency = prometheus.NewHistogramVec(
prometheus.HistogramOpts{
Subsystem: VolumeSchedulerSubsystem,
Name: "scheduling_duration_seconds",
Help: "Volume scheduling stage latency",
Buckets: prometheus.ExponentialBuckets(1000, 2, 15),
VolumeSchedulingStageLatency = metrics.NewHistogramVec(
&metrics.HistogramOpts{
Subsystem: VolumeSchedulerSubsystem,
Name: "scheduling_duration_seconds",
Help: "Volume scheduling stage latency",
Buckets: metrics.ExponentialBuckets(1000, 2, 15),
StabilityLevel: metrics.ALPHA,
},
[]string{"operation"},
)
// VolumeSchedulingStageFailed tracks the number of failed volume scheduling operations.
VolumeSchedulingStageFailed = prometheus.NewCounterVec(
prometheus.CounterOpts{
Subsystem: VolumeSchedulerSubsystem,
Name: "scheduling_stage_error_total",
Help: "Volume scheduling stage error count",
VolumeSchedulingStageFailed = metrics.NewCounterVec(
&metrics.CounterOpts{
Subsystem: VolumeSchedulerSubsystem,
Name: "scheduling_stage_error_total",
Help: "Volume scheduling stage error count",
StabilityLevel: metrics.ALPHA,
},
[]string{"operation"},
)
@ -57,7 +61,7 @@ var (
// RegisterVolumeSchedulingMetrics is used for scheduler, because the volume binding cache is a library
// used by scheduler process.
func RegisterVolumeSchedulingMetrics() {
prometheus.MustRegister(VolumeBindingRequestSchedulerBinderCache)
prometheus.MustRegister(VolumeSchedulingStageLatency)
prometheus.MustRegister(VolumeSchedulingStageFailed)
legacyregistry.MustRegister(VolumeBindingRequestSchedulerBinderCache)
legacyregistry.MustRegister(VolumeSchedulingStageLatency)
legacyregistry.MustRegister(VolumeSchedulingStageFailed)
}

8
vendor/k8s.io/kubernetes/pkg/controller/volume/scheduling/scheduler_assume_cache.go generated vendored
View File

@ -185,8 +185,11 @@ func (c *assumeCache) add(obj interface{}) {
}
objInfo := &objInfo{name: name, latestObj: obj, apiObj: obj}
c.store.Update(objInfo)
klog.V(10).Infof("Adding %v %v to assume cache: %+v ", c.description, name, obj)
if err = c.store.Update(objInfo); err != nil {
klog.Warningf("got error when updating stored object : %v", err)
} else {
klog.V(10).Infof("Adding %v %v to assume cache: %+v ", c.description, name, obj)
}
}
func (c *assumeCache) update(oldObj interface{}, newObj interface{}) {
@ -396,6 +399,7 @@ func (c *pvAssumeCache) ListPVs(storageClassName string) []*v1.PersistentVolume
pv, ok := obj.(*v1.PersistentVolume)
if !ok {
klog.Errorf("ListPVs: %v", &errWrongType{"v1.PersistentVolume", obj})
continue
}
pvs = append(pvs, pv)
}

158
vendor/k8s.io/kubernetes/pkg/controller/volume/scheduling/scheduler_binder.go generated vendored
View File

@ -19,23 +19,39 @@ package scheduling
import (
"fmt"
"sort"
"strings"
"time"
v1 "k8s.io/api/core/v1"
storagev1 "k8s.io/api/storage/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/apimachinery/pkg/util/wait"
"k8s.io/apiserver/pkg/storage/etcd"
"k8s.io/apiserver/pkg/storage/etcd3"
utilfeature "k8s.io/apiserver/pkg/util/feature"
coreinformers "k8s.io/client-go/informers/core/v1"
storageinformers "k8s.io/client-go/informers/storage/v1"
clientset "k8s.io/client-go/kubernetes"
storagelisters "k8s.io/client-go/listers/storage/v1"
csitrans "k8s.io/csi-translation-lib"
csiplugins "k8s.io/csi-translation-lib/plugins"
"k8s.io/klog"
v1helper "k8s.io/kubernetes/pkg/apis/core/v1/helper"
pvutil "k8s.io/kubernetes/pkg/controller/volume/persistentvolume/util"
"k8s.io/kubernetes/pkg/controller/volume/scheduling/metrics"
"k8s.io/kubernetes/pkg/features"
volumeutil "k8s.io/kubernetes/pkg/volume/util"
)
// InTreeToCSITranslator contains methods required to check migratable status
// and perform translations from InTree PV's to CSI
type InTreeToCSITranslator interface {
IsPVMigratable(pv *v1.PersistentVolume) bool
GetInTreePluginNameFromSpec(pv *v1.PersistentVolume, vol *v1.Volume) (string, error)
TranslateInTreePVToCSI(pv *v1.PersistentVolume) (*v1.PersistentVolume, error)
}
// SchedulerVolumeBinder is used by the scheduler to handle PVC/PV binding
// and dynamic provisioning. The binding decisions are integrated into the pod scheduling
// workflow so that the PV NodeAffinity is also considered along with the pod's other
@ -102,9 +118,10 @@ type volumeBinder struct {
kubeClient clientset.Interface
classLister storagelisters.StorageClassLister
nodeInformer coreinformers.NodeInformer
pvcCache PVCAssumeCache
pvCache PVAssumeCache
nodeInformer coreinformers.NodeInformer
csiNodeInformer storageinformers.CSINodeInformer
pvcCache PVCAssumeCache
pvCache PVAssumeCache
// Stores binding decisions that were made in FindPodVolumes for use in AssumePodVolumes.
// AssumePodVolumes modifies the bindings again for use in BindPodVolumes.
@ -112,12 +129,15 @@ type volumeBinder struct {
// Amount of time to wait for the bind operation to succeed
bindTimeout time.Duration
translator InTreeToCSITranslator
}
// NewVolumeBinder sets up all the caches needed for the scheduler to make volume binding decisions.
func NewVolumeBinder(
kubeClient clientset.Interface,
nodeInformer coreinformers.NodeInformer,
csiNodeInformer storageinformers.CSINodeInformer,
pvcInformer coreinformers.PersistentVolumeClaimInformer,
pvInformer coreinformers.PersistentVolumeInformer,
storageClassInformer storageinformers.StorageClassInformer,
@ -127,10 +147,12 @@ func NewVolumeBinder(
kubeClient: kubeClient,
classLister: storageClassInformer.Lister(),
nodeInformer: nodeInformer,
csiNodeInformer: csiNodeInformer,
pvcCache: NewPVCAssumeCache(pvcInformer.Informer()),
pvCache: NewPVAssumeCache(pvInformer.Informer()),
podBindingCache: NewPodBindingCache(),
bindTimeout: bindTimeout,
translator: csitrans.New(),
}
return b
@ -154,9 +176,9 @@ func (b *volumeBinder) FindPodVolumes(pod *v1.Pod, node *v1.Node) (unboundVolume
boundVolumesSatisfied = true
start := time.Now()
defer func() {
VolumeSchedulingStageLatency.WithLabelValues("predicate").Observe(time.Since(start).Seconds())
metrics.VolumeSchedulingStageLatency.WithLabelValues("predicate").Observe(time.Since(start).Seconds())
if err != nil {
VolumeSchedulingStageFailed.WithLabelValues("predicate").Inc()
metrics.VolumeSchedulingStageFailed.WithLabelValues("predicate").Inc()
}
}()
@ -256,9 +278,9 @@ func (b *volumeBinder) AssumePodVolumes(assumedPod *v1.Pod, nodeName string) (al
klog.V(4).Infof("AssumePodVolumes for pod %q, node %q", podName, nodeName)
start := time.Now()
defer func() {
VolumeSchedulingStageLatency.WithLabelValues("assume").Observe(time.Since(start).Seconds())
metrics.VolumeSchedulingStageLatency.WithLabelValues("assume").Observe(time.Since(start).Seconds())
if err != nil {
VolumeSchedulingStageFailed.WithLabelValues("assume").Inc()
metrics.VolumeSchedulingStageFailed.WithLabelValues("assume").Inc()
}
}()
@ -332,9 +354,9 @@ func (b *volumeBinder) BindPodVolumes(assumedPod *v1.Pod) (err error) {
start := time.Now()
defer func() {
VolumeSchedulingStageLatency.WithLabelValues("bind").Observe(time.Since(start).Seconds())
metrics.VolumeSchedulingStageLatency.WithLabelValues("bind").Observe(time.Since(start).Seconds())
if err != nil {
VolumeSchedulingStageFailed.WithLabelValues("bind").Inc()
metrics.VolumeSchedulingStageFailed.WithLabelValues("bind").Inc()
}
}()
@ -347,10 +369,14 @@ func (b *volumeBinder) BindPodVolumes(assumedPod *v1.Pod) (err error) {
return err
}
return wait.Poll(time.Second, b.bindTimeout, func() (bool, error) {
err = wait.Poll(time.Second, b.bindTimeout, func() (bool, error) {
b, err := b.checkBindings(assumedPod, bindings, claimsToProvision)
return b, err
})
if err != nil {
return fmt.Errorf("Failed to bind volumes: %v", err)
}
return nil
}
func getPodName(pod *v1.Pod) string {
@ -425,7 +451,7 @@ func (b *volumeBinder) bindAPIUpdate(podName string, bindings []*bindingInfo, cl
}
var (
versioner = etcd.APIObjectVersioner{}
versioner = etcd3.APIObjectVersioner{}
)
// checkBindings runs through all the PVCs in the Pod and checks:
@ -452,6 +478,12 @@ func (b *volumeBinder) checkBindings(pod *v1.Pod, bindings []*bindingInfo, claim
return false, fmt.Errorf("failed to get node %q: %v", pod.Spec.NodeName, err)
}
csiNode, err := b.csiNodeInformer.Lister().Get(node.Name)
if err != nil {
// TODO: return the error once CSINode is created by default
klog.V(4).Infof("Could not get a CSINode object for the node %q: %v", node.Name, err)
}
// Check for any conditions that might require scheduling retry
// When pod is removed from scheduling queue because of deletion or any
@ -480,6 +512,11 @@ func (b *volumeBinder) checkBindings(pod *v1.Pod, bindings []*bindingInfo, claim
return false, nil
}
pv, err = b.tryTranslatePVToCSI(pv, csiNode)
if err != nil {
return false, fmt.Errorf("failed to translate pv to csi: %v", err)
}
// Check PV's node affinity (the node might not have the proper label)
if err := volumeutil.CheckNodeAffinity(pv, node.Labels); err != nil {
return false, fmt.Errorf("pv %q node affinity doesn't match node %q: %v", pv.Name, node.Name, err)
@ -514,7 +551,10 @@ func (b *volumeBinder) checkBindings(pod *v1.Pod, bindings []*bindingInfo, claim
}
selectedNode := pvc.Annotations[pvutil.AnnSelectedNode]
if selectedNode != pod.Spec.NodeName {
return false, fmt.Errorf("selectedNode annotation value %q not set to scheduled node %q", selectedNode, pod.Spec.NodeName)
// If provisioner fails to provision a volume, selectedNode
// annotation will be removed to signal back to the scheduler to
// retry.
return false, fmt.Errorf("provisioning failed for PVC %q", pvc.Name)
}
// If the PVC is bound to a PV, check its node affinity
@ -530,6 +570,12 @@ func (b *volumeBinder) checkBindings(pod *v1.Pod, bindings []*bindingInfo, claim
}
return false, fmt.Errorf("failed to get pv %q from cache: %v", pvc.Spec.VolumeName, err)
}
pv, err = b.tryTranslatePVToCSI(pv, csiNode)
if err != nil {
return false, err
}
if err := volumeutil.CheckNodeAffinity(pv, node.Labels); err != nil {
return false, fmt.Errorf("pv %q node affinity doesn't match node %q: %v", pv.Name, node.Name, err)
}
@ -633,6 +679,12 @@ func (b *volumeBinder) getPodVolumes(pod *v1.Pod) (boundClaims []*v1.PersistentV
}
func (b *volumeBinder) checkBoundClaims(claims []*v1.PersistentVolumeClaim, node *v1.Node, podName string) (bool, error) {
csiNode, err := b.csiNodeInformer.Lister().Get(node.Name)
if err != nil {
// TODO: return the error once CSINode is created by default
klog.V(4).Infof("Could not get a CSINode object for the node %q: %v", node.Name, err)
}
for _, pvc := range claims {
pvName := pvc.Spec.VolumeName
pv, err := b.pvCache.GetPV(pvName)
@ -640,6 +692,11 @@ func (b *volumeBinder) checkBoundClaims(claims []*v1.PersistentVolumeClaim, node
return false, err
}
pv, err = b.tryTranslatePVToCSI(pv, csiNode)
if err != nil {
return false, err
}
err = volumeutil.CheckNodeAffinity(pv, node.Labels)
if err != nil {
klog.V(4).Infof("PersistentVolume %q, Node %q mismatch for Pod %q: %v", pvName, node.Name, podName, err)
@ -665,11 +722,7 @@ func (b *volumeBinder) findMatchingVolumes(pod *v1.Pod, claimsToBind []*v1.Persi
for _, pvc := range claimsToBind {
// Get storage class name from each PVC
storageClassName := ""
storageClass := pvc.Spec.StorageClassName
if storageClass != nil {
storageClassName = *storageClass
}
storageClassName := v1helper.GetPersistentVolumeClaimClass(pvc)
allPVs := b.pvCache.ListPVs(storageClassName)
pvcName := getPVCName(pvc)
@ -779,3 +832,72 @@ func (a byPVCSize) Less(i, j int) bool {
func claimToClaimKey(claim *v1.PersistentVolumeClaim) string {
return fmt.Sprintf("%s/%s", claim.Namespace, claim.Name)
}
// isCSIMigrationOnForPlugin checks if CSI migrartion is enabled for a given plugin.
func isCSIMigrationOnForPlugin(pluginName string) bool {
switch pluginName {
case csiplugins.AWSEBSInTreePluginName:
return utilfeature.DefaultFeatureGate.Enabled(features.CSIMigrationAWS)
case csiplugins.GCEPDInTreePluginName:
return utilfeature.DefaultFeatureGate.Enabled(features.CSIMigrationGCE)
case csiplugins.AzureDiskInTreePluginName:
return utilfeature.DefaultFeatureGate.Enabled(features.CSIMigrationAzureDisk)
case csiplugins.CinderInTreePluginName:
return utilfeature.DefaultFeatureGate.Enabled(features.CSIMigrationOpenStack)
}
return false
}
// isPluginMigratedToCSIOnNode checks if an in-tree plugin has been migrated to a CSI driver on the node.
func isPluginMigratedToCSIOnNode(pluginName string, csiNode *storagev1.CSINode) bool {
if csiNode == nil {
return false
}
csiNodeAnn := csiNode.GetAnnotations()
if csiNodeAnn == nil {
return false
}
var mpaSet sets.String
mpa := csiNodeAnn[v1.MigratedPluginsAnnotationKey]
if len(mpa) == 0 {
mpaSet = sets.NewString()
} else {
tok := strings.Split(mpa, ",")
mpaSet = sets.NewString(tok...)
}
return mpaSet.Has(pluginName)
}
// tryTranslatePVToCSI will translate the in-tree PV to CSI if it meets the criteria. If not, it returns the unmodified in-tree PV.
func (b *volumeBinder) tryTranslatePVToCSI(pv *v1.PersistentVolume, csiNode *storagev1.CSINode) (*v1.PersistentVolume, error) {
if !b.translator.IsPVMigratable(pv) {
return pv, nil
}
if !utilfeature.DefaultFeatureGate.Enabled(features.CSIMigration) {
return pv, nil
}
pluginName, err := b.translator.GetInTreePluginNameFromSpec(pv, nil)
if err != nil {
return nil, fmt.Errorf("could not get plugin name from pv: %v", err)
}
if !isCSIMigrationOnForPlugin(pluginName) {
return pv, nil
}
if !isPluginMigratedToCSIOnNode(pluginName, csiNode) {
return pv, nil
}
transPV, err := b.translator.TranslateInTreePVToCSI(pv)
if err != nil {
return nil, fmt.Errorf("could not translate pv: %v", err)
}
return transPV, nil
}

7
vendor/k8s.io/kubernetes/pkg/controller/volume/scheduling/scheduler_binder_cache.go generated vendored
View File

@ -19,7 +19,8 @@ package scheduling
import (
"sync"
"k8s.io/api/core/v1"
v1 "k8s.io/api/core/v1"
"k8s.io/kubernetes/pkg/controller/volume/scheduling/metrics"
)
// PodBindingCache stores PV binding decisions per pod per node.
@ -93,7 +94,7 @@ func (c *podBindingCache) DeleteBindings(pod *v1.Pod) {
if _, ok := c.bindingDecisions[podName]; ok {
delete(c.bindingDecisions, podName)
VolumeBindingRequestSchedulerBinderCache.WithLabelValues("delete").Inc()
metrics.VolumeBindingRequestSchedulerBinderCache.WithLabelValues("delete").Inc()
}
}
@ -113,7 +114,7 @@ func (c *podBindingCache) UpdateBindings(pod *v1.Pod, node string, bindings []*b
bindings: bindings,
provisionings: pvcs,
}
VolumeBindingRequestSchedulerBinderCache.WithLabelValues("add").Inc()
metrics.VolumeBindingRequestSchedulerBinderCache.WithLabelValues("add").Inc()
} else {
decision.bindings = bindings
decision.provisionings = pvcs