func (gc *GarbageCollector) Run(workers int, stopCh <-chan struct{}) {
glog.Infof("Garbage Collector: Initializing")
for _, monitor := range gc.monitors {
go monitor.controller.Run(stopCh)
}
wait.PollInfinite(10*time.Second, func() (bool, error) {
for _, monitor := range gc.monitors {
if !monitor.controller.HasSynced() {
glog.Infof("Garbage Collector: Waiting for resource monitors to be synced...")
return false, nil
}
}
return true, nil
})
glog.Infof("Garbage Collector: All monitored resources synced. Proceeding to collect garbage")
// worker
go wait.Until(gc.propagator.processEvent, 0, stopCh)
for i := 0; i < workers; i++ {
go wait.Until(gc.worker, 0, stopCh)
go wait.Until(gc.orphanFinalizer, 0, stopCh)
}
Register()
<-stopCh
glog.Infof("Garbage Collector: Shutting down")
gc.dirtyQueue.ShutDown()
gc.orphanQueue.ShutDown()
gc.propagator.eventQueue.ShutDown()
}
func (im *realImageGCManager) Start() {
go wait.Until(func() {
// Initial detection make detected time "unknown" in the past.
var ts time.Time
if im.initialized {
ts = time.Now()
}
err := im.detectImages(ts)
if err != nil {
glog.Warningf("[imageGCManager] Failed to monitor images: %v", err)
} else {
im.initialized = true
}
}, 5*time.Minute, wait.NeverStop)
// Start a goroutine periodically updates image cache.
// TODO(random-liu): Merge this with the previous loop.
go wait.Until(func() {
images, err := im.runtime.ListImages()
if err != nil {
glog.Warningf("[imageGCManager] Failed to update image list: %v", err)
} else {
im.imageCache.set(images)
}
}, 30*time.Second, wait.NeverStop)
}
// Runs controller blocks until stopCh is closed
func (e *TokensController) Run(workers int, stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
// Start controllers (to fill stores, call informers, fill work queues)
go e.serviceAccountController.Run(stopCh)
go e.secretController.Run(stopCh)
// Wait for stores to fill
for !e.serviceAccountController.HasSynced() || !e.secretController.HasSynced() {
time.Sleep(100 * time.Millisecond)
}
// Spawn workers to process work queues
for i := 0; i < workers; i++ {
go wait.Until(e.syncServiceAccount, 0, stopCh)
go wait.Until(e.syncSecret, 0, stopCh)
}
// Block until stop channel is closed
<-stopCh
// Shut down queues
e.syncServiceAccountQueue.ShutDown()
e.syncSecretQueue.ShutDown()
}
// 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(workers int) {
defer runtime.HandleCrash()
go s.serviceController.Run(wait.NeverStop)
for i := 0; i < workers; i++ {
go wait.Until(s.worker, time.Second, wait.NeverStop)
}
nodeLW := cache.NewListWatchFromClient(s.kubeClient.Core().RESTClient(), "nodes", v1.NamespaceAll, fields.Everything())
cache.NewReflector(nodeLW, &v1.Node{}, s.nodeLister.Store, 0).Run()
go wait.Until(s.nodeSyncLoop, nodeSyncPeriod, wait.NeverStop)
}
// Run starts an asynchronous loop that monitors the status of cluster nodes.
func (nc *NodeController) Run() {
go func() {
defer utilruntime.HandleCrash()
if !cache.WaitForCacheSync(wait.NeverStop, nc.nodeInformer.Informer().HasSynced, nc.podInformer.Informer().HasSynced, nc.daemonSetInformer.Informer().HasSynced) {
utilruntime.HandleError(errors.New("NodeController timed out while waiting for informers to sync..."))
return
}
// Incorporate the results of node status pushed from kubelet to master.
go wait.Until(func() {
if err := nc.monitorNodeStatus(); err != nil {
glog.Errorf("Error monitoring node status: %v", err)
}
}, nc.nodeMonitorPeriod, wait.NeverStop)
// Managing eviction of nodes:
// When we delete pods off a node, if the node was not empty at the time we then
// queue an eviction watcher. If we hit an error, retry deletion.
go wait.Until(func() {
nc.evictorLock.Lock()
defer nc.evictorLock.Unlock()
for k := range nc.zonePodEvictor {
nc.zonePodEvictor[k].Try(func(value TimedValue) (bool, time.Duration) {
obj, exists, err := nc.nodeStore.GetByKey(value.Value)
if err != nil {
glog.Warningf("Failed to get Node %v from the nodeStore: %v", value.Value, err)
} else if !exists {
glog.Warningf("Node %v no longer present in nodeStore!", value.Value)
} else {
node, _ := obj.(*v1.Node)
zone := utilnode.GetZoneKey(node)
EvictionsNumber.WithLabelValues(zone).Inc()
}
nodeUid, _ := value.UID.(string)
remaining, err := deletePods(nc.kubeClient, nc.recorder, value.Value, nodeUid, nc.daemonSetStore)
if err != nil {
utilruntime.HandleError(fmt.Errorf("unable to evict node %q: %v", value.Value, err))
return false, 0
}
if remaining {
glog.Infof("Pods awaiting deletion due to NodeController eviction")
}
return true, 0
})
}
}, nodeEvictionPeriod, wait.NeverStop)
}()
}
// Run starts all of this controller's control loops
func (ctrl *PersistentVolumeController) Run(stopCh <-chan struct{}) {
glog.V(1).Infof("starting PersistentVolumeController")
ctrl.initializeCaches(ctrl.volumeSource, ctrl.claimSource)
go ctrl.volumeController.Run(stopCh)
go ctrl.claimController.Run(stopCh)
go ctrl.classReflector.RunUntil(stopCh)
go wait.Until(ctrl.volumeWorker, time.Second, stopCh)
go wait.Until(ctrl.claimWorker, time.Second, stopCh)
<-stopCh
ctrl.claimQueue.ShutDown()
ctrl.volumeQueue.ShutDown()
}
func (fdc *DeploymentController) Run(workers int, stopCh <-chan struct{}) {
go fdc.deploymentController.Run(stopCh)
fdc.fedDeploymentInformer.Start()
fdc.fedPodInformer.Start()
fdc.deploymentDeliverer.StartWithHandler(func(item *fedutil.DelayingDelivererItem) {
fdc.deploymentWorkQueue.Add(item.Key)
})
fdc.clusterDeliverer.StartWithHandler(func(_ *fedutil.DelayingDelivererItem) {
fdc.reconcileDeploymentsOnClusterChange()
})
// Wait until the cluster is synced to prevent the update storm at the very beginning.
for !fdc.isSynced() {
time.Sleep(5 * time.Millisecond)
glog.V(3).Infof("Waiting for controller to sync up")
}
for i := 0; i < workers; i++ {
go wait.Until(fdc.worker, time.Second, stopCh)
}
fedutil.StartBackoffGC(fdc.deploymentBackoff, stopCh)
<-stopCh
glog.Infof("Shutting down DeploymentController")
fdc.deploymentDeliverer.Stop()
fdc.clusterDeliverer.Stop()
fdc.deploymentWorkQueue.ShutDown()
fdc.fedDeploymentInformer.Stop()
fdc.fedPodInformer.Stop()
}
// Runs e; will not return until stopCh is closed. workers determines how many
// endpoints will be handled in parallel.
func (e *EndpointController) Run(workers int, stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
defer e.queue.ShutDown()
go e.serviceController.Run(stopCh)
go e.podController.Run(stopCh)
if !cache.WaitForCacheSync(stopCh, e.podStoreSynced) {
return
}
for i := 0; i < workers; i++ {
go wait.Until(e.worker, time.Second, stopCh)
}
go func() {
defer utilruntime.HandleCrash()
time.Sleep(5 * time.Minute) // give time for our cache to fill
e.checkLeftoverEndpoints()
}()
if e.internalPodInformer != nil {
go e.internalPodInformer.Run(stopCh)
}
<-stopCh
}
func startKubelet(k kubelet.KubeletBootstrap, podCfg *config.PodConfig, kubeCfg *componentconfig.KubeletConfiguration, kubeDeps *kubelet.KubeletDeps) {
// start the kubelet
go wait.Until(func() { k.Run(podCfg.Updates()) }, 0, wait.NeverStop)
// start the kubelet server
if kubeCfg.EnableServer {
go wait.Until(func() {
k.ListenAndServe(net.ParseIP(kubeCfg.Address), uint(kubeCfg.Port), kubeDeps.TLSOptions, kubeDeps.Auth, kubeCfg.EnableDebuggingHandlers)
}, 0, wait.NeverStop)
}
if kubeCfg.ReadOnlyPort > 0 {
go wait.Until(func() {
k.ListenAndServeReadOnly(net.ParseIP(kubeCfg.Address), uint(kubeCfg.ReadOnlyPort))
}, 0, wait.NeverStop)
}
}
// Start starts resource collector and connects to the standalone Cadvisor pod
// then repeatedly runs collectStats.
func (r *ResourceCollector) Start() {
// Get the cgroup container names for kubelet and docker
kubeletContainer, err := getContainerNameForProcess(kubeletProcessName, "")
dockerContainer, err := getContainerNameForProcess(dockerProcessName, dockerPidFile)
if err == nil {
systemContainers = map[string]string{
stats.SystemContainerKubelet: kubeletContainer,
stats.SystemContainerRuntime: dockerContainer,
}
} else {
framework.Failf("Failed to get docker container name in test-e2e-node resource collector.")
}
wait.Poll(1*time.Second, 1*time.Minute, func() (bool, error) {
var err error
r.client, err = cadvisorclient.NewClient(fmt.Sprintf("http://localhost:%d/", cadvisorPort))
if err == nil {
return true, nil
}
return false, err
})
Expect(r.client).NotTo(BeNil(), "cadvisor client not ready")
r.request = &cadvisorapiv2.RequestOptions{IdType: "name", Count: 1, Recursive: false}
r.stopCh = make(chan struct{})
oldStatsMap := make(map[string]*cadvisorapiv2.ContainerStats)
go wait.Until(func() { r.collectStats(oldStatsMap) }, r.pollingInterval, r.stopCh)
}
// RunUntil starts the controller until the provided ch is closed.
func (c *Repair) RunUntil(ch chan struct{}) {
wait.Until(func() {
if err := c.RunOnce(); err != nil {
runtime.HandleError(err)
}
}, c.interval, ch)
}
func (frsc *ReplicaSetController) Run(workers int, stopCh <-chan struct{}) {
go frsc.replicaSetController.Run(stopCh)
frsc.fedReplicaSetInformer.Start()
frsc.fedPodInformer.Start()
frsc.replicasetDeliverer.StartWithHandler(func(item *fedutil.DelayingDelivererItem) {
frsc.replicasetWorkQueue.Add(item.Key)
})
frsc.clusterDeliverer.StartWithHandler(func(_ *fedutil.DelayingDelivererItem) {
frsc.reconcileReplicaSetsOnClusterChange()
})
for !frsc.isSynced() {
time.Sleep(5 * time.Millisecond)
}
for i := 0; i < workers; i++ {
go wait.Until(frsc.worker, time.Second, stopCh)
}
fedutil.StartBackoffGC(frsc.replicaSetBackoff, stopCh)
<-stopCh
glog.Infof("Shutting down ReplicaSetController")
frsc.replicasetDeliverer.Stop()
frsc.clusterDeliverer.Stop()
frsc.replicasetWorkQueue.ShutDown()
frsc.fedReplicaSetInformer.Stop()
frsc.fedPodInformer.Stop()
}
// Run the main goroutine responsible for watching and syncing jobs.
func (jm *CronJobController) Run(stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
glog.Infof("Starting CronJob Manager")
// Check things every 10 second.
go wait.Until(jm.SyncAll, 10*time.Second, stopCh)
<-stopCh
glog.Infof("Shutting down CronJob Manager")
}
// RunUntil starts a watch and handles watch events. Will restart the watch if it is closed.
// RunUntil starts a goroutine and returns immediately. It will exit when stopCh is closed.
func (r *Reflector) RunUntil(stopCh <-chan struct{}) {
glog.V(3).Infof("Starting reflector %v (%s) from %s", r.expectedType, r.resyncPeriod, r.name)
go wait.Until(func() {
if err := r.ListAndWatch(stopCh); err != nil {
utilruntime.HandleError(err)
}
}, r.period, stopCh)
}
// Run starts a watch and handles watch events. Will restart the watch if it is closed.
// Run starts a goroutine and returns immediately.
func (r *Reflector) Run() {
glog.V(3).Infof("Starting reflector %v (%s) from %s", r.expectedType, r.resyncPeriod, r.name)
go wait.Until(func() {
if err := r.ListAndWatch(wait.NeverStop); err != nil {
utilruntime.HandleError(err)
}
}, r.period, wait.NeverStop)
}
// Run begins quota controller using the specified number of workers
func (rq *ResourceQuotaController) Run(workers int, stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
go rq.rqController.Run(stopCh)
// the controllers that replenish other resources to respond rapidly to state changes
for _, replenishmentController := range rq.replenishmentControllers {
go replenishmentController.Run(stopCh)
}
// the workers that chug through the quota calculation backlog
for i := 0; i < workers; i++ {
go wait.Until(rq.worker(rq.queue), time.Second, stopCh)
go wait.Until(rq.worker(rq.missingUsageQueue), time.Second, stopCh)
}
// the timer for how often we do a full recalculation across all quotas
go wait.Until(func() { rq.enqueueAll() }, rq.resyncPeriod(), stopCh)
<-stopCh
glog.Infof("Shutting down ResourceQuotaController")
rq.queue.ShutDown()
}
// Run starts observing the system with the specified number of workers.
func (nm *NamespaceController) Run(workers int, stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
go nm.controller.Run(stopCh)
for i := 0; i < workers; i++ {
go wait.Until(nm.worker, time.Second, stopCh)
}
<-stopCh
glog.Infof("Shutting down NamespaceController")
nm.queue.ShutDown()
}
// Run begins watching and syncing.
func (dc *DeploymentController) Run(workers int, stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
defer dc.queue.ShutDown()
defer dc.progressQueue.ShutDown()
glog.Infof("Starting deployment controller")
if !cache.WaitForCacheSync(stopCh, dc.dListerSynced, dc.rsListerSynced, dc.podListerSynced) {
return
}
for i := 0; i < workers; i++ {
go wait.Until(dc.worker, time.Second, stopCh)
}
go wait.Until(dc.progressWorker, time.Second, stopCh)
<-stopCh
glog.Infof("Shutting down deployment controller")
}
// Run begins watching and syncing.
func (e *quotaEvaluator) run() {
defer utilruntime.HandleCrash()
for i := 0; i < e.workers; i++ {
go wait.Until(e.doWork, time.Second, e.stopCh)
}
<-e.stopCh
glog.Infof("Shutting down quota evaluator")
e.queue.ShutDown()
}
// RunKubernetesNamespaces periodically makes sure that all internal namespaces exist
func (c *Controller) RunKubernetesNamespaces(ch chan struct{}) {
wait.Until(func() {
// Loop the system namespace list, and create them if they do not exist
for _, ns := range c.SystemNamespaces {
if err := c.CreateNamespaceIfNeeded(ns); err != nil {
runtime.HandleError(fmt.Errorf("unable to create required kubernetes system namespace %s: %v", ns, err))
}
}
}, c.SystemNamespacesInterval, ch)
}
// Run begins watching and syncing.
func (cc *ClusterController) Run() {
defer utilruntime.HandleCrash()
go cc.clusterController.Run(wait.NeverStop)
// monitor cluster status periodically, in phase 1 we just get the health state from "/healthz"
go wait.Until(func() {
if err := cc.UpdateClusterStatus(); err != nil {
glog.Errorf("Error monitoring cluster status: %v", err)
}
}, cc.clusterMonitorPeriod, wait.NeverStop)
}
// RunKubernetesService periodically updates the kubernetes service
func (c *Controller) RunKubernetesService(ch chan struct{}) {
wait.Until(func() {
// Service definition is not reconciled after first
// run, ports and type will be corrected only during
// start.
if err := c.UpdateKubernetesService(false); err != nil {
runtime.HandleError(fmt.Errorf("unable to sync kubernetes service: %v", err))
}
}, c.EndpointInterval, ch)
}
func (cm *containerManagerImpl) Start(node *v1.Node) error {
// cache the node Info including resource capacity and
// allocatable of the node
cm.nodeInfo = node
// Setup the node
if err := cm.setupNode(); err != nil {
return err
}
// Don't run a background thread if there are no ensureStateFuncs.
hasEnsureStateFuncs := false
for _, cont := range cm.systemContainers {
if cont.ensureStateFunc != nil {
hasEnsureStateFuncs = true
break
}
}
if hasEnsureStateFuncs {
// Run ensure state functions every minute.
go wait.Until(func() {
for _, cont := range cm.systemContainers {
if cont.ensureStateFunc != nil {
if err := cont.ensureStateFunc(cont.manager); err != nil {
glog.Warningf("[ContainerManager] Failed to ensure state of %q: %v", cont.name, err)
}
}
}
}, time.Minute, wait.NeverStop)
}
if len(cm.periodicTasks) > 0 {
go wait.Until(func() {
for _, task := range cm.periodicTasks {
if task != nil {
task()
}
}
}, 5*time.Minute, wait.NeverStop)
}
return nil
}
func (dc *DisruptionController) Run(stopCh <-chan struct{}) {
glog.V(0).Infof("Starting disruption controller")
if dc.kubeClient != nil {
glog.V(0).Infof("Sending events to api server.")
dc.broadcaster.StartRecordingToSink(&v1core.EventSinkImpl{Interface: dc.kubeClient.Core().Events("")})
} else {
glog.V(0).Infof("No api server defined - no events will be sent to API server.")
}
go dc.pdbController.Run(stopCh)
go dc.podController.Run(stopCh)
go dc.rcController.Run(stopCh)
go dc.rsController.Run(stopCh)
go dc.dController.Run(stopCh)
go dc.ssController.Run(stopCh)
go wait.Until(dc.worker, time.Second, stopCh)
go wait.Until(dc.recheckWorker, time.Second, stopCh)
<-stopCh
glog.V(0).Infof("Shutting down disruption controller")
}
// Run runs the statefulset controller.
func (psc *StatefulSetController) Run(workers int, stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
glog.Infof("Starting statefulset controller")
go psc.podController.Run(stopCh)
go psc.psController.Run(stopCh)
for i := 0; i < workers; i++ {
go wait.Until(psc.worker, time.Second, stopCh)
}
<-stopCh
glog.Infof("Shutting down statefulset controller")
psc.queue.ShutDown()
}
func (m *containerManager) Start() error {
// TODO: check if the required cgroups are mounted.
if len(m.cgroupsName) != 0 {
manager, err := createCgroupManager(m.cgroupsName)
if err != nil {
return err
}
m.cgroupsManager = manager
}
go wait.Until(m.doWork, 5*time.Minute, wait.NeverStop)
return nil
}
// nodesSyncLoop lists nodes every 15 seconds, calling Update() on the TunnelList
// each time (Update() is a noop if no changes are necessary).
func (c *SSHTunneler) nodesSyncLoop() {
// TODO (cjcullen) make this watch.
go wait.Until(func() {
addrs, err := c.getAddresses()
glog.V(4).Infof("Calling update w/ addrs: %v", addrs)
if err != nil {
glog.Errorf("Failed to getAddresses: %v", err)
}
c.tunnels.Update(addrs)
atomic.StoreInt64(&c.lastSync, c.clock.Now().Unix())
}, 15*time.Second, c.stopChan)
}
// Create a new Cacher responsible from service WATCH and LIST requests from its
// internal cache and updating its cache in the background based on the given
// configuration.
func NewCacherFromConfig(config CacherConfig) *Cacher {
watchCache := newWatchCache(config.CacheCapacity, config.KeyFunc, config.GetAttrsFunc)
listerWatcher := newCacherListerWatcher(config.Storage, config.ResourcePrefix, config.NewListFunc)
// Give this error when it is constructed rather than when you get the
// first watch item, because it's much easier to track down that way.
if obj, ok := config.Type.(runtime.Object); ok {
if err := runtime.CheckCodec(config.Codec, obj); err != nil {
panic("storage codec doesn't seem to match given type: " + err.Error())
}
}
stopCh := make(chan struct{})
cacher := &Cacher{
ready: newReady(),
storage: config.Storage,
objectType: reflect.TypeOf(config.Type),
watchCache: watchCache,
reflector: cache.NewReflector(listerWatcher, config.Type, watchCache, 0),
versioner: config.Versioner,
triggerFunc: config.TriggerPublisherFunc,
watcherIdx: 0,
watchers: indexedWatchers{
allWatchers: make(map[int]*cacheWatcher),
valueWatchers: make(map[string]watchersMap),
},
// TODO: Figure out the correct value for the buffer size.
incoming: make(chan watchCacheEvent, 100),
dispatchTimeoutBudget: newTimeBudget(stopCh),
// We need to (potentially) stop both:
// - wait.Until go-routine
// - reflector.ListAndWatch
// and there are no guarantees on the order that they will stop.
// So we will be simply closing the channel, and synchronizing on the WaitGroup.
stopCh: stopCh,
}
watchCache.SetOnEvent(cacher.processEvent)
go cacher.dispatchEvents()
cacher.stopWg.Add(1)
go func() {
defer cacher.stopWg.Done()
wait.Until(
func() {
if !cacher.isStopped() {
cacher.startCaching(stopCh)
}
}, time.Second, stopCh,
)
}()
return cacher
}
// Run the main goroutine responsible for watching and syncing jobs.
func (cc *CertificateController) Run(workers int, stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
defer cc.queue.ShutDown()
go cc.csrController.Run(stopCh)
glog.Infof("Starting certificate controller manager")
for i := 0; i < workers; i++ {
go wait.Until(cc.worker, time.Second, stopCh)
}
<-stopCh
glog.Infof("Shutting down certificate controller")
}
func (c *APIServiceRegistrationController) Run(stopCh <-chan struct{}) {
defer utilruntime.HandleCrash()
defer c.queue.ShutDown()
defer glog.Infof("Shutting down APIServiceRegistrationController")
glog.Infof("Starting APIServiceRegistrationController")
// only start one worker thread since its a slow moving API and the discovery server adding bits
// aren't threadsafe
go wait.Until(c.runWorker, time.Second, stopCh)
<-stopCh
}
