func NewPodGC(kubeClient clientset.Interface, podInformer cache.SharedIndexInformer, terminatedPodThreshold int) *PodGCController {
if kubeClient != nil && kubeClient.Core().RESTClient().GetRateLimiter() != nil {
metrics.RegisterMetricAndTrackRateLimiterUsage("gc_controller", kubeClient.Core().RESTClient().GetRateLimiter())
}
gcc := &PodGCController{
kubeClient: kubeClient,
terminatedPodThreshold: terminatedPodThreshold,
deletePod: func(namespace, name string) error {
glog.Infof("PodGC is force deleting Pod: %v:%v", namespace, name)
return kubeClient.Core().Pods(namespace).Delete(name, v1.NewDeleteOptions(0))
},
}
gcc.podStore.Indexer = podInformer.GetIndexer()
gcc.podController = podInformer.GetController()
gcc.nodeStore.Store, gcc.nodeController = cache.NewInformer(
&cache.ListWatch{
ListFunc: func(options v1.ListOptions) (runtime.Object, error) {
return gcc.kubeClient.Core().Nodes().List(options)
},
WatchFunc: func(options v1.ListOptions) (watch.Interface, error) {
return gcc.kubeClient.Core().Nodes().Watch(options)
},
},
&v1.Node{},
controller.NoResyncPeriodFunc(),
cache.ResourceEventHandlerFuncs{},
)
return gcc
}
// newReplicationManager configures a replication manager with the specified event recorder
func newReplicationManager(eventRecorder record.EventRecorder, podInformer cache.SharedIndexInformer, kubeClient clientset.Interface, resyncPeriod controller.ResyncPeriodFunc, burstReplicas int, lookupCacheSize int, garbageCollectorEnabled bool) *ReplicationManager {
if kubeClient != nil && kubeClient.Core().RESTClient().GetRateLimiter() != nil {
metrics.RegisterMetricAndTrackRateLimiterUsage("replication_controller", kubeClient.Core().RESTClient().GetRateLimiter())
}
rm := &ReplicationManager{
kubeClient: kubeClient,
podControl: controller.RealPodControl{
KubeClient: kubeClient,
Recorder: eventRecorder,
},
burstReplicas: burstReplicas,
expectations: controller.NewUIDTrackingControllerExpectations(controller.NewControllerExpectations()),
queue: workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "replicationmanager"),
garbageCollectorEnabled: garbageCollectorEnabled,
}
rm.rcStore.Indexer, rm.rcController = cache.NewIndexerInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return rm.kubeClient.Core().ReplicationControllers(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return rm.kubeClient.Core().ReplicationControllers(api.NamespaceAll).Watch(options)
},
},
&api.ReplicationController{},
// TODO: Can we have much longer period here?
FullControllerResyncPeriod,
cache.ResourceEventHandlerFuncs{
AddFunc: rm.enqueueController,
UpdateFunc: rm.updateRC,
// This will enter the sync loop and no-op, because the controller has been deleted from the store.
// Note that deleting a controller immediately after scaling it to 0 will not work. The recommended
// way of achieving this is by performing a `stop` operation on the controller.
DeleteFunc: rm.enqueueController,
},
cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc},
)
podInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
AddFunc: rm.addPod,
// This invokes the rc for every pod change, eg: host assignment. Though this might seem like overkill
// the most frequent pod update is status, and the associated rc will only list from local storage, so
// it should be ok.
UpdateFunc: rm.updatePod,
DeleteFunc: rm.deletePod,
})
rm.podStore.Indexer = podInformer.GetIndexer()
rm.podController = podInformer.GetController()
rm.syncHandler = rm.syncReplicationController
rm.podStoreSynced = rm.podController.HasSynced
rm.lookupCache = controller.NewMatchingCache(lookupCacheSize)
return rm
}
// NewPetSetController creates a new petset controller.
func NewPetSetController(podInformer cache.SharedIndexInformer, kubeClient internalclientset.Interface, resyncPeriod time.Duration) *PetSetController {
eventBroadcaster := record.NewBroadcaster()
eventBroadcaster.StartLogging(glog.Infof)
eventBroadcaster.StartRecordingToSink(&unversionedcore.EventSinkImpl{Interface: kubeClient.Core().Events("")})
recorder := eventBroadcaster.NewRecorder(api.EventSource{Component: "petset"})
pc := &apiServerPetClient{kubeClient, recorder, &defaultPetHealthChecker{}}
psc := &PetSetController{
kubeClient: kubeClient,
blockingPetStore: newUnHealthyPetTracker(pc),
newSyncer: func(blockingPet *pcb) *petSyncer {
return &petSyncer{pc, blockingPet}
},
queue: workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "petset"),
}
podInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
// lookup the petset and enqueue
AddFunc: psc.addPod,
// lookup current and old petset if labels changed
UpdateFunc: psc.updatePod,
// lookup petset accounting for deletion tombstones
DeleteFunc: psc.deletePod,
})
psc.podStore.Indexer = podInformer.GetIndexer()
psc.podController = podInformer.GetController()
psc.psStore.Store, psc.psController = cache.NewInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return psc.kubeClient.Apps().PetSets(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return psc.kubeClient.Apps().PetSets(api.NamespaceAll).Watch(options)
},
},
&apps.PetSet{},
petSetResyncPeriod,
cache.ResourceEventHandlerFuncs{
AddFunc: psc.enqueuePetSet,
UpdateFunc: func(old, cur interface{}) {
oldPS := old.(*apps.PetSet)
curPS := cur.(*apps.PetSet)
if oldPS.Status.Replicas != curPS.Status.Replicas {
glog.V(4).Infof("Observed updated replica count for PetSet: %v, %d->%d", curPS.Name, oldPS.Status.Replicas, curPS.Status.Replicas)
}
psc.enqueuePetSet(cur)
},
DeleteFunc: psc.enqueuePetSet,
},
)
// TODO: Watch volumes
psc.podStoreSynced = psc.podController.HasSynced
psc.syncHandler = psc.Sync
return psc
}
// NewEndpointController returns a new *EndpointController.
func NewEndpointController(podInformer cache.SharedIndexInformer, client clientset.Interface) *EndpointController {
if client != nil && client.Core().GetRESTClient().GetRateLimiter() != nil {
metrics.RegisterMetricAndTrackRateLimiterUsage("endpoint_controller", client.Core().GetRESTClient().GetRateLimiter())
}
e := &EndpointController{
client: client,
queue: workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "endpoint"),
}
e.serviceStore.Indexer, e.serviceController = cache.NewIndexerInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return e.client.Core().Services(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return e.client.Core().Services(api.NamespaceAll).Watch(options)
},
},
&api.Service{},
// TODO: Can we have much longer period here?
FullServiceResyncPeriod,
cache.ResourceEventHandlerFuncs{
AddFunc: e.enqueueService,
UpdateFunc: func(old, cur interface{}) {
e.enqueueService(cur)
},
DeleteFunc: e.enqueueService,
},
cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc},
)
podInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
AddFunc: e.addPod,
UpdateFunc: e.updatePod,
DeleteFunc: e.deletePod,
})
e.podStore.Indexer = podInformer.GetIndexer()
e.podController = podInformer.GetController()
e.podStoreSynced = podInformer.HasSynced
return e
}
func NewDisruptionController(podInformer cache.SharedIndexInformer, kubeClient internalclientset.Interface) *DisruptionController {
dc := &DisruptionController{
kubeClient: kubeClient,
podController: podInformer.GetController(),
queue: workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "disruption"),
recheckQueue: workqueue.NewNamedDelayingQueue("disruption-recheck"),
broadcaster: record.NewBroadcaster(),
}
dc.recorder = dc.broadcaster.NewRecorder(api.EventSource{Component: "controllermanager"})
dc.getUpdater = func() updater { return dc.writePdbStatus }
dc.podLister.Indexer = podInformer.GetIndexer()
podInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
AddFunc: dc.addPod,
UpdateFunc: dc.updatePod,
DeleteFunc: dc.deletePod,
})
dc.pdbStore, dc.pdbController = cache.NewInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return dc.kubeClient.Policy().PodDisruptionBudgets(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return dc.kubeClient.Policy().PodDisruptionBudgets(api.NamespaceAll).Watch(options)
},
},
&policy.PodDisruptionBudget{},
30*time.Second,
cache.ResourceEventHandlerFuncs{
AddFunc: dc.addDb,
UpdateFunc: dc.updateDb,
DeleteFunc: dc.removeDb,
},
)
dc.pdbLister.Store = dc.pdbStore
dc.rcIndexer, dc.rcController = cache.NewIndexerInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return dc.kubeClient.Core().ReplicationControllers(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return dc.kubeClient.Core().ReplicationControllers(api.NamespaceAll).Watch(options)
},
},
&api.ReplicationController{},
30*time.Second,
cache.ResourceEventHandlerFuncs{},
cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc},
)
dc.rcLister.Indexer = dc.rcIndexer
dc.rsLister.Indexer, dc.rsController = cache.NewIndexerInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return dc.kubeClient.Extensions().ReplicaSets(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return dc.kubeClient.Extensions().ReplicaSets(api.NamespaceAll).Watch(options)
},
},
&extensions.ReplicaSet{},
30*time.Second,
cache.ResourceEventHandlerFuncs{},
cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc},
)
dc.rsStore = dc.rsLister.Indexer
dc.dIndexer, dc.dController = cache.NewIndexerInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return dc.kubeClient.Extensions().Deployments(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return dc.kubeClient.Extensions().Deployments(api.NamespaceAll).Watch(options)
},
},
&extensions.Deployment{},
30*time.Second,
cache.ResourceEventHandlerFuncs{},
cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc},
)
dc.dLister.Indexer = dc.dIndexer
return dc
}
// NewNodeController returns a new node controller to sync instances from cloudprovider.
// This method returns an error if it is unable to initialize the CIDR bitmap with
// podCIDRs it has already allocated to nodes. Since we don't allow podCIDR changes
// currently, this should be handled as a fatal error.
func NewNodeController(
podInformer cache.SharedIndexInformer,
cloud cloudprovider.Interface,
kubeClient clientset.Interface,
podEvictionTimeout time.Duration,
evictionLimiterQPS float32,
secondaryEvictionLimiterQPS float32,
largeClusterThreshold int32,
unhealthyZoneThreshold float32,
nodeMonitorGracePeriod time.Duration,
nodeStartupGracePeriod time.Duration,
nodeMonitorPeriod time.Duration,
clusterCIDR *net.IPNet,
serviceCIDR *net.IPNet,
nodeCIDRMaskSize int,
allocateNodeCIDRs bool) (*NodeController, error) {
eventBroadcaster := record.NewBroadcaster()
recorder := eventBroadcaster.NewRecorder(api.EventSource{Component: "controllermanager"})
eventBroadcaster.StartLogging(glog.Infof)
if kubeClient != nil {
glog.V(0).Infof("Sending events to api server.")
eventBroadcaster.StartRecordingToSink(&unversionedcore.EventSinkImpl{Interface: kubeClient.Core().Events("")})
} else {
glog.V(0).Infof("No api server defined - no events will be sent to API server.")
}
if kubeClient != nil && kubeClient.Core().GetRESTClient().GetRateLimiter() != nil {
metrics.RegisterMetricAndTrackRateLimiterUsage("node_controller", kubeClient.Core().GetRESTClient().GetRateLimiter())
}
if allocateNodeCIDRs {
if clusterCIDR == nil {
glog.Fatal("NodeController: Must specify clusterCIDR if allocateNodeCIDRs == true.")
}
mask := clusterCIDR.Mask
if maskSize, _ := mask.Size(); maskSize > nodeCIDRMaskSize {
glog.Fatal("NodeController: Invalid clusterCIDR, mask size of clusterCIDR must be less than nodeCIDRMaskSize.")
}
}
nc := &NodeController{
cloud: cloud,
knownNodeSet: make(map[string]*api.Node),
kubeClient: kubeClient,
recorder: recorder,
podEvictionTimeout: podEvictionTimeout,
maximumGracePeriod: 5 * time.Minute,
zonePodEvictor: make(map[string]*RateLimitedTimedQueue),
zoneTerminationEvictor: make(map[string]*RateLimitedTimedQueue),
nodeStatusMap: make(map[string]nodeStatusData),
nodeMonitorGracePeriod: nodeMonitorGracePeriod,
nodeMonitorPeriod: nodeMonitorPeriod,
nodeStartupGracePeriod: nodeStartupGracePeriod,
lookupIP: net.LookupIP,
now: unversioned.Now,
clusterCIDR: clusterCIDR,
serviceCIDR: serviceCIDR,
allocateNodeCIDRs: allocateNodeCIDRs,
forcefullyDeletePod: func(p *api.Pod) error { return forcefullyDeletePod(kubeClient, p) },
nodeExistsInCloudProvider: func(nodeName string) (bool, error) { return nodeExistsInCloudProvider(cloud, nodeName) },
evictionLimiterQPS: evictionLimiterQPS,
secondaryEvictionLimiterQPS: secondaryEvictionLimiterQPS,
largeClusterThreshold: largeClusterThreshold,
unhealthyZoneThreshold: unhealthyZoneThreshold,
zoneStates: make(map[string]zoneState),
}
nc.enterPartialDisruptionFunc = nc.ReducedQPSFunc
nc.enterFullDisruptionFunc = nc.HealthyQPSFunc
nc.computeZoneStateFunc = nc.ComputeZoneState
podInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
AddFunc: nc.maybeDeleteTerminatingPod,
UpdateFunc: func(_, obj interface{}) { nc.maybeDeleteTerminatingPod(obj) },
})
nc.podStore.Indexer = podInformer.GetIndexer()
nc.podController = podInformer.GetController()
nodeEventHandlerFuncs := cache.ResourceEventHandlerFuncs{}
if nc.allocateNodeCIDRs {
nodeEventHandlerFuncs = cache.ResourceEventHandlerFuncs{
AddFunc: func(obj interface{}) {
node := obj.(*api.Node)
err := nc.cidrAllocator.AllocateOrOccupyCIDR(node)
if err != nil {
glog.Errorf("Error allocating CIDR: %v", err)
}
},
UpdateFunc: func(_, obj interface{}) {
node := obj.(*api.Node)
// If the PodCIDR is not empty we either:
// - already processed a Node that already had a CIDR after NC restarted
// (cidr is marked as used),
// - already processed a Node successfully and allocated a CIDR for it
// (cidr is marked as used),
// - already processed a Node but we did saw a "timeout" response and
// request eventually got through in this case we haven't released
// the allocated CIDR (cidr is still marked as used).
// There's a possible error here:
// - NC sees a new Node and assigns a CIDR X to it,
// - Update Node call fails with a timeout,
// - Node is updated by some other component, NC sees an update and
// assigns CIDR Y to the Node,
// - Both CIDR X and CIDR Y are marked as used in the local cache,
// even though Node sees only CIDR Y
// The problem here is that in in-memory cache we see CIDR X as marked,
// which prevents it from being assigned to any new node. The cluster
// state is correct.
// Restart of NC fixes the issue.
if node.Spec.PodCIDR == "" {
err := nc.cidrAllocator.AllocateOrOccupyCIDR(node)
if err != nil {
glog.Errorf("Error allocating CIDR: %v", err)
}
}
},
DeleteFunc: func(obj interface{}) {
node, isNode := obj.(*api.Node)
// We can get DeletedFinalStateUnknown instead of *api.Node here and we need to handle that correctly. #34692
if !isNode {
deletedState, ok := obj.(cache.DeletedFinalStateUnknown)
if !ok {
glog.Errorf("Received unexpected object: %v", obj)
return
}
node, ok = deletedState.Obj.(*api.Node)
if !ok {
glog.Errorf("DeletedFinalStateUnknown contained non-Node object: %v", deletedState.Obj)
return
}
}
if err := nc.cidrAllocator.ReleaseCIDR(node); err != nil {
glog.Errorf("Error releasing CIDR: %v", err)
}
},
}
}
nc.nodeStore.Store, nc.nodeController = cache.NewInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return nc.kubeClient.Core().Nodes().List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return nc.kubeClient.Core().Nodes().Watch(options)
},
},
&api.Node{},
controller.NoResyncPeriodFunc(),
nodeEventHandlerFuncs,
)
nc.daemonSetStore.Store, nc.daemonSetController = cache.NewInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return nc.kubeClient.Extensions().DaemonSets(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return nc.kubeClient.Extensions().DaemonSets(api.NamespaceAll).Watch(options)
},
},
&extensions.DaemonSet{},
controller.NoResyncPeriodFunc(),
cache.ResourceEventHandlerFuncs{},
)
if allocateNodeCIDRs {
var nodeList *api.NodeList
var err error
// We must poll because apiserver might not be up. This error causes
// controller manager to restart.
if pollErr := wait.Poll(10*time.Second, apiserverStartupGracePeriod, func() (bool, error) {
nodeList, err = kubeClient.Core().Nodes().List(api.ListOptions{
FieldSelector: fields.Everything(),
LabelSelector: labels.Everything(),
})
if err != nil {
glog.Errorf("Failed to list all nodes: %v", err)
return false, nil
}
return true, nil
}); pollErr != nil {
return nil, fmt.Errorf("Failed to list all nodes in %v, cannot proceed without updating CIDR map", apiserverStartupGracePeriod)
}
nc.cidrAllocator, err = NewCIDRRangeAllocator(kubeClient, clusterCIDR, serviceCIDR, nodeCIDRMaskSize, nodeList)
if err != nil {
return nil, err
}
}
return nc, nil
}
func NewDaemonSetsController(podInformer cache.SharedIndexInformer, kubeClient clientset.Interface, resyncPeriod controller.ResyncPeriodFunc, lookupCacheSize int) *DaemonSetsController {
eventBroadcaster := record.NewBroadcaster()
eventBroadcaster.StartLogging(glog.Infof)
// TODO: remove the wrapper when every clients have moved to use the clientset.
eventBroadcaster.StartRecordingToSink(&unversionedcore.EventSinkImpl{Interface: kubeClient.Core().Events("")})
if kubeClient != nil && kubeClient.Core().GetRESTClient().GetRateLimiter() != nil {
metrics.RegisterMetricAndTrackRateLimiterUsage("daemon_controller", kubeClient.Core().GetRESTClient().GetRateLimiter())
}
dsc := &DaemonSetsController{
kubeClient: kubeClient,
eventRecorder: eventBroadcaster.NewRecorder(api.EventSource{Component: "daemonset-controller"}),
podControl: controller.RealPodControl{
KubeClient: kubeClient,
Recorder: eventBroadcaster.NewRecorder(api.EventSource{Component: "daemon-set"}),
},
burstReplicas: BurstReplicas,
expectations: controller.NewControllerExpectations(),
queue: workqueue.NewNamed("daemonset"),
}
// Manage addition/update of daemon sets.
dsc.dsStore.Store, dsc.dsController = cache.NewInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return dsc.kubeClient.Extensions().DaemonSets(api.NamespaceAll).List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return dsc.kubeClient.Extensions().DaemonSets(api.NamespaceAll).Watch(options)
},
},
&extensions.DaemonSet{},
// TODO: Can we have much longer period here?
FullDaemonSetResyncPeriod,
cache.ResourceEventHandlerFuncs{
AddFunc: func(obj interface{}) {
ds := obj.(*extensions.DaemonSet)
glog.V(4).Infof("Adding daemon set %s", ds.Name)
dsc.enqueueDaemonSet(ds)
},
UpdateFunc: func(old, cur interface{}) {
oldDS := old.(*extensions.DaemonSet)
curDS := cur.(*extensions.DaemonSet)
// We should invalidate the whole lookup cache if a DS's selector has been updated.
//
// Imagine that you have two RSs:
// * old DS1
// * new DS2
// You also have a pod that is attached to DS2 (because it doesn't match DS1 selector).
// Now imagine that you are changing DS1 selector so that it is now matching that pod,
// in such case we must invalidate the whole cache so that pod could be adopted by DS1
//
// This makes the lookup cache less helpful, but selector update does not happen often,
// so it's not a big problem
if !reflect.DeepEqual(oldDS.Spec.Selector, curDS.Spec.Selector) {
dsc.lookupCache.InvalidateAll()
}
glog.V(4).Infof("Updating daemon set %s", oldDS.Name)
dsc.enqueueDaemonSet(curDS)
},
DeleteFunc: dsc.deleteDaemonset,
},
)
// Watch for creation/deletion of pods. The reason we watch is that we don't want a daemon set to create/delete
// more pods until all the effects (expectations) of a daemon set's create/delete have been observed.
podInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
AddFunc: dsc.addPod,
UpdateFunc: dsc.updatePod,
DeleteFunc: dsc.deletePod,
})
dsc.podStore.Indexer = podInformer.GetIndexer()
dsc.podController = podInformer.GetController()
dsc.podStoreSynced = podInformer.HasSynced
// Watch for new nodes or updates to nodes - daemon pods are launched on new nodes, and possibly when labels on nodes change,
dsc.nodeStore.Store, dsc.nodeController = cache.NewInformer(
&cache.ListWatch{
ListFunc: func(options api.ListOptions) (runtime.Object, error) {
return dsc.kubeClient.Core().Nodes().List(options)
},
WatchFunc: func(options api.ListOptions) (watch.Interface, error) {
return dsc.kubeClient.Core().Nodes().Watch(options)
},
},
&api.Node{},
resyncPeriod(),
cache.ResourceEventHandlerFuncs{
AddFunc: dsc.addNode,
UpdateFunc: dsc.updateNode,
},
)
dsc.syncHandler = dsc.syncDaemonSet
dsc.lookupCache = controller.NewMatchingCache(lookupCacheSize)
return dsc
}