Merge pull request #16413 from ddysher/controller-manager-start

cmd/kube-controller-manager/app/controllermanager.go

@@ -55,6 +55,7 @@ import (
 	"k8s.io/kubernetes/pkg/healthz"
 	"k8s.io/kubernetes/pkg/master/ports"
 	"k8s.io/kubernetes/pkg/util"
+	"k8s.io/kubernetes/pkg/util/wait"
 
 	"github.com/golang/glog"
 	"github.com/prometheus/client_golang/prometheus"
@@ -309,7 +310,13 @@ func (s *CMServer) Run(_ []string) error {
 
 	resourcequotacontroller.NewResourceQuotaController(kubeClient).Run(s.ResourceQuotaSyncPeriod)
 
-	versionStrings, err := client.ServerAPIVersions(kubeconfig)
+	// If apiserver is not running we should wait for some time and fail only then. This is particularly
+	// important when we start apiserver and controller manager at the same time.
+	var versionStrings []string
+	err = wait.PollImmediate(time.Second, 10*time.Second, func() (bool, error) {
+		versionStrings, err = client.ServerAPIVersions(kubeconfig)
+		return err == nil, err
+	})
 	if err != nil {
 		glog.Fatalf("Failed to get api versions from server: %v", err)
 	}

pkg/registry/service/ipallocator/controller/repair.go

@@ -27,6 +27,7 @@ import (
 	"k8s.io/kubernetes/pkg/registry/service"
 	"k8s.io/kubernetes/pkg/registry/service/ipallocator"
 	"k8s.io/kubernetes/pkg/util"
+	"k8s.io/kubernetes/pkg/util/wait"
 )
 
 // Repair is a controller loop that periodically examines all service ClusterIP allocations
@@ -84,13 +85,10 @@ func (c *Repair) RunOnce() error {
 	// important when we start apiserver and etcd at the same time.
 	var latest *api.RangeAllocation
 	var err error
-	for i := 0; i < 10; i++ {
-		if latest, err = c.alloc.Get(); err != nil {
-			time.Sleep(time.Second)
-		} else {
-			break
-		}
-	}
+	err = wait.PollImmediate(time.Second, 10*time.Second, func() (bool, error) {
+		latest, err = c.alloc.Get()
+		return err == nil, err
+	})
 	if err != nil {
 		return fmt.Errorf("unable to refresh the service IP block: %v", err)
 	}