// SetNewReplicaSetAnnotations sets new replica set's annotations appropriately by updating its revision and
// copying required deployment annotations to it; it returns true if replica set's annotation is changed.
func SetNewReplicaSetAnnotations(deployment *extensions.Deployment, newRS *extensions.ReplicaSet, newRevision string, exists bool) bool {
// First, copy deployment's annotations (except for apply and revision annotations)
annotationChanged := copyDeploymentAnnotationsToReplicaSet(deployment, newRS)
// Then, update replica set's revision annotation
if newRS.Annotations == nil {
newRS.Annotations = make(map[string]string)
}
oldRevision, ok := newRS.Annotations[RevisionAnnotation]
// The newRS's revision should be the greatest among all RSes. Usually, its revision number is newRevision (the max revision number
// of all old RSes + 1). However, it's possible that some of the old RSes are deleted after the newRS revision being updated, and
// newRevision becomes smaller than newRS's revision. We should only update newRS revision when it's smaller than newRevision.
if oldRevision < newRevision {
newRS.Annotations[RevisionAnnotation] = newRevision
annotationChanged = true
glog.V(4).Infof("Updating replica set %q revision to %s", newRS.Name, newRevision)
}
// If a revision annotation already existed and this replica set was updated with a new revision
// then that means we are rolling back to this replica set. We need to preserve the old revisions
// for historical information.
if ok && annotationChanged {
revisionHistoryAnnotation := newRS.Annotations[RevisionHistoryAnnotation]
oldRevisions := strings.Split(revisionHistoryAnnotation, ",")
if len(oldRevisions[0]) == 0 {
newRS.Annotations[RevisionHistoryAnnotation] = oldRevision
} else {
oldRevisions = append(oldRevisions, oldRevision)
newRS.Annotations[RevisionHistoryAnnotation] = strings.Join(oldRevisions, ",")
}
}
// If the new replica set is about to be created, we need to add replica annotations to it.
if !exists && SetReplicasAnnotations(newRS, *(deployment.Spec.Replicas), *(deployment.Spec.Replicas)+MaxSurge(*deployment)) {
annotationChanged = true
}
return annotationChanged
}
// updateReplicaSetStatus attempts to update the Status.Replicas of the given ReplicaSet, with a single GET/PUT retry.
func updateReplicaSetStatus(c unversionedextensions.ReplicaSetInterface, rs extensions.ReplicaSet, newStatus extensions.ReplicaSetStatus) (updateErr error) {
// This is the steady state. It happens when the ReplicaSet doesn't have any expectations, since
// we do a periodic relist every 30s. If the generations differ but the replicas are
// the same, a caller might've resized to the same replica count.
if rs.Status.Replicas == newStatus.Replicas &&
rs.Status.FullyLabeledReplicas == newStatus.FullyLabeledReplicas &&
rs.Status.ReadyReplicas == newStatus.ReadyReplicas &&
rs.Status.AvailableReplicas == newStatus.AvailableReplicas &&
rs.Generation == rs.Status.ObservedGeneration &&
reflect.DeepEqual(rs.Status.Conditions, newStatus.Conditions) {
return nil
}
// deep copy to avoid mutation now.
// TODO this method need some work. Retry on conflict probably, though I suspect this is stomping status to something it probably shouldn't
copyObj, err := api.Scheme.DeepCopy(rs)
if err != nil {
return err
}
rs = copyObj.(extensions.ReplicaSet)
// Save the generation number we acted on, otherwise we might wrongfully indicate
// that we've seen a spec update when we retry.
// TODO: This can clobber an update if we allow multiple agents to write to the
// same status.
newStatus.ObservedGeneration = rs.Generation
var getErr error
for i, rs := 0, &rs; ; i++ {
glog.V(4).Infof(fmt.Sprintf("Updating replica count for ReplicaSet: %s/%s, ", rs.Namespace, rs.Name) +
fmt.Sprintf("replicas %d->%d (need %d), ", rs.Status.Replicas, newStatus.Replicas, *(rs.Spec.Replicas)) +
fmt.Sprintf("fullyLabeledReplicas %d->%d, ", rs.Status.FullyLabeledReplicas, newStatus.FullyLabeledReplicas) +
fmt.Sprintf("readyReplicas %d->%d, ", rs.Status.ReadyReplicas, newStatus.ReadyReplicas) +
fmt.Sprintf("availableReplicas %d->%d, ", rs.Status.AvailableReplicas, newStatus.AvailableReplicas) +
fmt.Sprintf("sequence No: %v->%v", rs.Status.ObservedGeneration, newStatus.ObservedGeneration))
rs.Status = newStatus
_, updateErr = c.UpdateStatus(rs)
if updateErr == nil || i >= statusUpdateRetries {
return updateErr
}
// Update the ReplicaSet with the latest resource version for the next poll
if rs, getErr = c.Get(rs.Name); getErr != nil {
// If the GET fails we can't trust status.Replicas anymore. This error
// is bound to be more interesting than the update failure.
return getErr
}
}
}
// SetReplicasAnnotations sets the desiredReplicas and maxReplicas into the annotations
func SetReplicasAnnotations(rs *extensions.ReplicaSet, desiredReplicas, maxReplicas int32) bool {
updated := false
if rs.Annotations == nil {
rs.Annotations = make(map[string]string)
}
desiredString := fmt.Sprintf("%d", desiredReplicas)
if hasString := rs.Annotations[DesiredReplicasAnnotation]; hasString != desiredString {
rs.Annotations[DesiredReplicasAnnotation] = desiredString
updated = true
}
maxString := fmt.Sprintf("%d", maxReplicas)
if hasString := rs.Annotations[MaxReplicasAnnotation]; hasString != maxString {
rs.Annotations[MaxReplicasAnnotation] = maxString
updated = true
}
return updated
}
// copyDeploymentAnnotationsToReplicaSet copies deployment's annotations to replica set's annotations,
// and returns true if replica set's annotation is changed.
// Note that apply and revision annotations are not copied.
func copyDeploymentAnnotationsToReplicaSet(deployment *extensions.Deployment, rs *extensions.ReplicaSet) bool {
rsAnnotationsChanged := false
if rs.Annotations == nil {
rs.Annotations = make(map[string]string)
}
for k, v := range deployment.Annotations {
// newRS revision is updated automatically in getNewReplicaSet, and the deployment's revision number is then updated
// by copying its newRS revision number. We should not copy deployment's revision to its newRS, since the update of
// deployment revision number may fail (revision becomes stale) and the revision number in newRS is more reliable.
if skipCopyAnnotation(k) || rs.Annotations[k] == v {
continue
}
rs.Annotations[k] = v
rsAnnotationsChanged = true
}
return rsAnnotationsChanged
}
// Returns a replica set that matches the intent of the given deployment. Returns nil if the new replica set doesn't exist yet.
// 1. Get existing new RS (the RS that the given deployment targets, whose pod template is the same as deployment's).
// 2. If there's existing new RS, update its revision number if it's smaller than (maxOldRevision + 1), where maxOldRevision is the max revision number among all old RSes.
// 3. If there's no existing new RS and createIfNotExisted is true, create one with appropriate revision number (maxOldRevision + 1) and replicas.
// Note that the pod-template-hash will be added to adopted RSes and pods.
func (dc *DeploymentController) getNewReplicaSet(deployment *extensions.Deployment, rsList, oldRSs []*extensions.ReplicaSet, createIfNotExisted bool) (*extensions.ReplicaSet, error) {
existingNewRS, err := deploymentutil.FindNewReplicaSet(deployment, rsList)
if err != nil {
return nil, err
}
// Calculate the max revision number among all old RSes
maxOldRevision := deploymentutil.MaxRevision(oldRSs)
// Calculate revision number for this new replica set
newRevision := strconv.FormatInt(maxOldRevision+1, 10)
// Latest replica set exists. We need to sync its annotations (includes copying all but
// annotationsToSkip from the parent deployment, and update revision, desiredReplicas,
// and maxReplicas) and also update the revision annotation in the deployment with the
// latest revision.
if existingNewRS != nil {
objCopy, err := api.Scheme.Copy(existingNewRS)
if err != nil {
return nil, err
}
rsCopy := objCopy.(*extensions.ReplicaSet)
// Set existing new replica set's annotation
annotationsUpdated := deploymentutil.SetNewReplicaSetAnnotations(deployment, rsCopy, newRevision, true)
minReadySecondsNeedsUpdate := rsCopy.Spec.MinReadySeconds != deployment.Spec.MinReadySeconds
if annotationsUpdated || minReadySecondsNeedsUpdate {
rsCopy.Spec.MinReadySeconds = deployment.Spec.MinReadySeconds
return dc.client.Extensions().ReplicaSets(rsCopy.ObjectMeta.Namespace).Update(rsCopy)
}
updateConditions := deploymentutil.SetDeploymentRevision(deployment, newRevision)
// If no other Progressing condition has been recorded and we need to estimate the progress
// of this deployment then it is likely that old users started caring about progress. In that
// case we need to take into account the first time we noticed their new replica set.
cond := deploymentutil.GetDeploymentCondition(deployment.Status, extensions.DeploymentProgressing)
if deployment.Spec.ProgressDeadlineSeconds != nil && cond == nil {
msg := fmt.Sprintf("Found new replica set %q", rsCopy.Name)
condition := deploymentutil.NewDeploymentCondition(extensions.DeploymentProgressing, v1.ConditionTrue, deploymentutil.FoundNewRSReason, msg)
deploymentutil.SetDeploymentCondition(&deployment.Status, *condition)
updateConditions = true
}
if updateConditions {
if deployment, err = dc.client.Extensions().Deployments(deployment.Namespace).UpdateStatus(deployment); err != nil {
return nil, err
}
}
return rsCopy, nil
}
if !createIfNotExisted {
return nil, nil
}
// new ReplicaSet does not exist, create one.
namespace := deployment.Namespace
podTemplateSpecHash := fmt.Sprintf("%d", deploymentutil.GetPodTemplateSpecHash(deployment.Spec.Template))
newRSTemplate := deploymentutil.GetNewReplicaSetTemplate(deployment)
newRSTemplate.Labels = labelsutil.CloneAndAddLabel(deployment.Spec.Template.Labels, extensions.DefaultDeploymentUniqueLabelKey, podTemplateSpecHash)
// Add podTemplateHash label to selector.
newRSSelector := labelsutil.CloneSelectorAndAddLabel(deployment.Spec.Selector, extensions.DefaultDeploymentUniqueLabelKey, podTemplateSpecHash)
// Create new ReplicaSet
newRS := extensions.ReplicaSet{
ObjectMeta: metav1.ObjectMeta{
// Make the name deterministic, to ensure idempotence
Name: deployment.Name + "-" + podTemplateSpecHash,
Namespace: namespace,
},
Spec: extensions.ReplicaSetSpec{
Replicas: func(i int32) *int32 { return &i }(0),
MinReadySeconds: deployment.Spec.MinReadySeconds,
Selector: newRSSelector,
Template: newRSTemplate,
},
}
var trueVar = true
controllerRef := &metav1.OwnerReference{
APIVersion: getDeploymentKind().GroupVersion().String(),
Kind: getDeploymentKind().Kind,
Name: deployment.Name,
UID: deployment.UID,
Controller: &trueVar,
}
newRS.OwnerReferences = append(newRS.OwnerReferences, *controllerRef)
allRSs := append(oldRSs, &newRS)
newReplicasCount, err := deploymentutil.NewRSNewReplicas(deployment, allRSs, &newRS)
if err != nil {
return nil, err
}
*(newRS.Spec.Replicas) = newReplicasCount
// Set new replica set's annotation
deploymentutil.SetNewReplicaSetAnnotations(deployment, &newRS, newRevision, false)
createdRS, err := dc.client.Extensions().ReplicaSets(namespace).Create(&newRS)
switch {
// We may end up hitting this due to a slow cache or a fast resync of the deployment.
// TODO: Restore once https://github.com/kubernetes/kubernetes/issues/29735 is fixed
// ie. we start using a new hashing algorithm.
case errors.IsAlreadyExists(err):
return nil, err
// return dc.rsLister.ReplicaSets(namespace).Get(newRS.Name)
case err != nil:
msg := fmt.Sprintf("Failed to create new replica set %q: %v", newRS.Name, err)
if deployment.Spec.ProgressDeadlineSeconds != nil {
cond := deploymentutil.NewDeploymentCondition(extensions.DeploymentProgressing, v1.ConditionFalse, deploymentutil.FailedRSCreateReason, msg)
deploymentutil.SetDeploymentCondition(&deployment.Status, *cond)
// We don't really care about this error at this point, since we have a bigger issue to report.
// TODO: Update the rest of the Deployment status, too. We may need to do this every time we
// error out in all other places in the controller so that we let users know that their deployments
// have been noticed by the controller, albeit with errors.
// TODO: Identify which errors are permanent and switch DeploymentIsFailed to take into account
// these reasons as well. Related issue: https://github.com/kubernetes/kubernetes/issues/18568
_, _ = dc.client.Extensions().Deployments(deployment.ObjectMeta.Namespace).UpdateStatus(deployment)
}
dc.eventRecorder.Eventf(deployment, v1.EventTypeWarning, deploymentutil.FailedRSCreateReason, msg)
return nil, err
}
if newReplicasCount > 0 {
dc.eventRecorder.Eventf(deployment, v1.EventTypeNormal, "ScalingReplicaSet", "Scaled up replica set %s to %d", createdRS.Name, newReplicasCount)
}
deploymentutil.SetDeploymentRevision(deployment, newRevision)
if deployment.Spec.ProgressDeadlineSeconds != nil {
msg := fmt.Sprintf("Created new replica set %q", createdRS.Name)
condition := deploymentutil.NewDeploymentCondition(extensions.DeploymentProgressing, v1.ConditionTrue, deploymentutil.NewReplicaSetReason, msg)
deploymentutil.SetDeploymentCondition(&deployment.Status, *condition)
}
_, err = dc.client.Extensions().Deployments(deployment.Namespace).UpdateStatus(deployment)
return createdRS, err
}