Fix SDN exponential backoff timeouts

- First iteration of wait.ExponentialBackoff() will not wait, so
  the times will be:
  0, a, a * r, a * r^2, ... a * r^n-2 [a: duration, r: factor, n: steps]
  Total = a * (r^(n-1) - 1)/(r-1))

Author: Ravi Sankar Penta

pkg/network/node/runtime.go

@@ -23,7 +23,7 @@ func (node *OsdnNode) getRuntimeService() (kubeletapi.RuntimeService, error) {
 		kwait.Backoff{
 			Duration: 100 * time.Millisecond,
 			Factor:   1.2,
-			Steps:    23,
+			Steps:    24,
 		},
 		func() (bool, error) {
 			runtimeService, err := kubeletremote.NewRemoteRuntimeService(node.runtimeEndpoint, node.runtimeRequestTimeout)

pkg/network/node/sdn_controller.go

@@ -30,10 +30,10 @@ func (plugin *OsdnNode) getLocalSubnet() (string, error) {
 	// unexpectedly long though, so give it plenty of time before returning an error
 	// (since that will cause the node process to exit).
 	backoff := utilwait.Backoff{
-		// A bit over 1 minute total
+		// ~2 mins total
 		Duration: time.Second,
 		Factor:   1.5,
-		Steps:    8,
+		Steps:    11,
 	}
 	err := utilwait.ExponentialBackoff(backoff, func() (bool, error) {
 		var err error
@@ -106,10 +106,11 @@ func (plugin *OsdnNode) alreadySetUp(localSubnetGatewayCIDR string, clusterNetwo
 }
 
 func deleteLocalSubnetRoute(device, localSubnetCIDR string) {
+	// ~1 sec total
 	backoff := utilwait.Backoff{
 		Duration: 100 * time.Millisecond,
 		Factor:   1.25,
-		Steps:    6,
+		Steps:    7,
 	}
 	err := utilwait.ExponentialBackoff(backoff, func() (bool, error) {
 		l, err := netlink.LinkByName(device)

pkg/oc/admin/diagnostics/diagnostics/network/run_pod.go

@@ -143,7 +143,7 @@ func (d *NetworkDiagnostic) runNetworkDiagnostic() {
 		return
 	}
 	// Wait for network diagnostic pod completion (timeout: ~3 mins)
-	backoff := wait.Backoff{Steps: 38, Duration: 500 * time.Millisecond, Factor: 1.1}
+	backoff := wait.Backoff{Steps: 39, Duration: 500 * time.Millisecond, Factor: 1.1}
 	if err := d.waitForNetworkPod(d.nsName1, util.NetworkDiagPodNamePrefix, backoff, []kapi.PodPhase{kapi.PodSucceeded, kapi.PodFailed}); err != nil {
 		d.res.Error("DNet2007", err, err.Error())
 		return
@@ -164,7 +164,7 @@ func (d *NetworkDiagnostic) runNetworkDiagnostic() {
 	}
 
 	// Wait for network diagnostic pod to start (timeout: ~5 mins)
-	backoff = wait.Backoff{Steps: 36, Duration: time.Second, Factor: 1.1}
+	backoff = wait.Backoff{Steps: 37, Duration: time.Second, Factor: 1.1}
 	if err := d.waitForNetworkPod(d.nsName1, util.NetworkDiagPodNamePrefix, backoff, []kapi.PodPhase{kapi.PodRunning, kapi.PodFailed, kapi.PodSucceeded}); err != nil {
 		d.res.Error("DNet2010", err, err.Error())
 		// Do not bail out here, collect what ever info is available from all valid nodes

pkg/oc/admin/diagnostics/diagnostics/network/setup.go

@@ -155,7 +155,7 @@ func (d *NetworkDiagnostic) waitForTestPodAndService(nsList []string) error {
 	errList := []error{}
 	validPhases := []kapi.PodPhase{kapi.PodRunning, kapi.PodSucceeded, kapi.PodFailed}
 	for _, name := range nsList {
-		backoff := wait.Backoff{Steps: 36, Duration: time.Second, Factor: 1.1} // timeout: ~5 mins
+		backoff := wait.Backoff{Steps: 37, Duration: time.Second, Factor: 1.1} // timeout: ~5 mins
 		if err := d.waitForNetworkPod(name, util.NetworkDiagTestPodNamePrefix, backoff, validPhases); err != nil {
 			errList = append(errList, err)
 		}