Add support for initial backoff to the apiserver call on recover (#476)

Stash will wait to connect to the apiserver until the delay expires. The default is set to forever as it's reasonable to expect that the apiserver will come back; until it doesn't the restore pod has nothing else to do, really.

Fixes #475

recover.go

@@ -1,6 +1,8 @@
 package cmds
 
 import (
+	"time"
+
 	"github.com/appscode/go/log"
 	"github.com/appscode/kutil/meta"
 	cs "github.com/appscode/stash/client/clientset/versioned/typed/stash/v1alpha1"
@@ -15,6 +17,7 @@ func NewCmdRecover() *cobra.Command {
 		masterURL      string
 		kubeconfigPath string
 		recoveryName   string
+		backoffMaxWait time.Duration
 	)
 
 	cmd := &cobra.Command{
@@ -29,13 +32,14 @@ func NewCmdRecover() *cobra.Command {
 			kubeClient := kubernetes.NewForConfigOrDie(config)
 			stashClient := cs.NewForConfigOrDie(config)
 
-			c := recovery.New(kubeClient, stashClient, meta.Namespace(), recoveryName)
+			c := recovery.New(kubeClient, stashClient, meta.Namespace(), recoveryName, backoffMaxWait)
 			c.Run()
 		},
 	}
 	cmd.Flags().StringVar(&masterURL, "master", masterURL, "The address of the Kubernetes API server (overrides any value in kubeconfig)")
 	cmd.Flags().StringVar(&kubeconfigPath, "kubeconfig", kubeconfigPath, "Path to kubeconfig file with authorization information (the master location is set by the master flag).")
 	cmd.Flags().StringVar(&recoveryName, "recovery-name", recoveryName, "Name of the Recovery CRD.")
+	cmd.Flags().DurationVar(&backoffMaxWait, "backoff-max-wait", 0, "Maximum wait for initial response from kube apiserver; 0 disables the timeout")
 
 	return cmd
 }