Merge pull request kubernetes#30138 from gmarek/flags

Automatic merge from submit-queue Expose flags for new NodeEviction logic in NodeController Fix kubernetes#28832 Last PR from the NodeController NodeEviction logic series. cc @davidopp @lavalamp @mml
lavalamp · Aug 18, 2016 · f9190ed · f9190ed
2 parents d576486 + 4cf698e
commit f9190ed
Show file tree

Hide file tree

Showing 10 changed files with 1,037 additions and 829 deletions.
diff --git a/cmd/kube-controller-manager/app/controllermanager.go b/cmd/kube-controller-manager/app/controllermanager.go
@@ -236,7 +236,7 @@ func StartControllers(s *options.CMServer, kubeClient *client.Client, kubeconfig
 		glog.Warningf("Unsuccessful parsing of service CIDR %v: %v", s.ServiceCIDR, err)
 	}
 	nodeController, err := nodecontroller.NewNodeController(sharedInformers.Pods().Informer(), cloud, clientset.NewForConfigOrDie(restclient.AddUserAgent(kubeconfig, "node-controller")),
-		s.PodEvictionTimeout.Duration, s.DeletingPodsQps, s.NodeMonitorGracePeriod.Duration,
+		s.PodEvictionTimeout.Duration, s.NodeEvictionRate, s.SecondaryNodeEvictionRate, s.LargeClusterSizeThreshold, s.UnhealthyZoneThreshold, s.NodeMonitorGracePeriod.Duration,
 		s.NodeStartupGracePeriod.Duration, s.NodeMonitorPeriod.Duration, clusterCIDR, serviceCIDR,
 		int(s.NodeCIDRMaskSize), s.AllocateNodeCIDRs)
 	if err != nil {

diff --git a/cmd/kube-controller-manager/app/options/options.go b/cmd/kube-controller-manager/app/options/options.go
@@ -141,6 +141,7 @@ func (s *CMServer) AddFlags(fs *pflag.FlagSet) {
 	fs.DurationVar(&s.DeploymentControllerSyncPeriod.Duration, "deployment-controller-sync-period", s.DeploymentControllerSyncPeriod.Duration, "Period for syncing the deployments.")
 	fs.DurationVar(&s.PodEvictionTimeout.Duration, "pod-eviction-timeout", s.PodEvictionTimeout.Duration, "The grace period for deleting pods on failed nodes.")
 	fs.Float32Var(&s.DeletingPodsQps, "deleting-pods-qps", 0.1, "Number of nodes per second on which pods are deleted in case of node failure.")
+	fs.MarkDeprecated("deleting-pods-qps", "This flag is currently no-op and will be deleted.")
 	fs.Int32Var(&s.DeletingPodsBurst, "deleting-pods-burst", 0, "Number of nodes on which pods are bursty deleted in case of node failure. For more details look into RateLimiter.")
 	fs.MarkDeprecated("deleting-pods-burst", "This flag is currently no-op and will be deleted.")
 	fs.Int32Var(&s.RegisterRetryCount, "register-retry-count", s.RegisterRetryCount, ""+
@@ -173,5 +174,10 @@ func (s *CMServer) AddFlags(fs *pflag.FlagSet) {
 	fs.DurationVar(&s.ControllerStartInterval.Duration, "controller-start-interval", s.ControllerStartInterval.Duration, "Interval between starting controller managers.")
 	fs.BoolVar(&s.EnableGarbageCollector, "enable-garbage-collector", s.EnableGarbageCollector, "Enables the generic garbage collector. MUST be synced with the corresponding flag of the kube-apiserver. WARNING: the generic garbage collector is an alpha feature.")
 	fs.Int32Var(&s.ConcurrentGCSyncs, "concurrent-gc-syncs", s.ConcurrentGCSyncs, "The number of garbage collector workers that are allowed to sync concurrently.")
+	fs.Float32Var(&s.NodeEvictionRate, "node-eviction-rate", 0.1, "Number of nodes per second on which pods are deleted in case of node failure when a zone is healthy (see --unhealthy-zone-threshold for definition of healthy/unhealthy). Zone refers to entire cluster in non-multizone clusters.")
+	fs.Float32Var(&s.SecondaryNodeEvictionRate, "secondary-node-eviction-rate", 0.01, "Number of nodes per second on which pods are deleted in case of node failure when a zone is unhealthy (see --unhealthy-zone-threshold for definition of healthy/unhealthy). Zone refers to entire cluster in non-multizone clusters. This value is implicitly overridden to 0 if the cluster size is smaller than --large-cluster-size-threshold.")
+	fs.Int32Var(&s.LargeClusterSizeThreshold, "large-cluster-size-threshold", 50, "Number of nodes from which NodeController treats the cluster as large for the eviction logic purposes. --secondary-node-eviction-rate is implicitly overridden to 0 for clusters this size or smaller.")
+	fs.Float32Var(&s.UnhealthyZoneThreshold, "unhealthy-zone-threshold", 0.55, "Fraction of Nodes in a zone which needs to be not Ready (minimum 3) for zone to be treated as unhealthy. ")
+
 	leaderelection.BindFlags(&s.LeaderElection, fs)
 }
diff --git a/contrib/mesos/pkg/controllermanager/controllermanager.go b/contrib/mesos/pkg/controllermanager/controllermanager.go
@@ -159,7 +159,7 @@ func (s *CMServer) Run(_ []string) error {
 	_, clusterCIDR, _ := net.ParseCIDR(s.ClusterCIDR)
 	_, serviceCIDR, _ := net.ParseCIDR(s.ServiceCIDR)
 	nodeController, err := nodecontroller.NewNodeControllerFromClient(cloud, clientset.NewForConfigOrDie(restclient.AddUserAgent(kubeconfig, "node-controller")),
-		s.PodEvictionTimeout.Duration, s.DeletingPodsQps,
+		s.PodEvictionTimeout.Duration, s.NodeEvictionRate, s.SecondaryNodeEvictionRate, s.LargeClusterSizeThreshold, s.UnhealthyZoneThreshold,
 		s.NodeMonitorGracePeriod.Duration, s.NodeStartupGracePeriod.Duration, s.NodeMonitorPeriod.Duration, clusterCIDR, serviceCIDR, int(s.NodeCIDRMaskSize), s.AllocateNodeCIDRs)
 	if err != nil {
 		glog.Fatalf("Failed to initialize nodecontroller: %v", err)

diff --git a/hack/verify-flags/known-flags.txt b/hack/verify-flags/known-flags.txt
@@ -276,6 +276,7 @@ kubelet-sync-frequency
 kubelet-timeout
 kubernetes-service-node-port
 label-columns
+large-cluster-size-threshold
 last-release-pr
 leader-elect
 leader-elect-lease-duration
@@ -335,6 +336,7 @@ network-plugin-dir
 no-headers
 no-suggestions
 node-cidr-mask-size
+node-eviction-rate
 node-instance-group
 node-ip
 node-labels
@@ -433,6 +435,7 @@ scheduler-name
 schema-cache-dir
 scopes
 seccomp-profile-root
+secondary-node-eviction-rate
 secure-port
 serialize-image-pulls
 server-start-timeout
@@ -492,6 +495,7 @@ ttl-secs
 type-src
 udp-port
 udp-timeout
+unhealthy-zone-threshold
 unix-socket
 update-period
 upgrade-target