Skip communicating with kubernetes api failures

Multus is failing when can't communicate with kube-api, which fails
to create pods even in case of not using network attachment, this issue
appeared from commit c976f6b for
Add deviceid in clusterNetwork.

This change makes multus skip the failures of communicating with kube-api
and create the networks of delegates

Signed-off-by: Mamduh Alassi <mamduhala@mellanox.com>

pkg/multus/multus.go

@@ -549,8 +549,10 @@ func CmdAdd(args *skel.CmdArgs, exec invoke.Exec, kubeClient *k8s.ClientInfo) (c
 					return nil, cmdErr(k8sArgs, "error getting pod by service unavailable: %v", err)
 				}
 			} else {
-				// Other case, return error
-				return nil, cmdErr(k8sArgs, "error getting pod: %v", err)
+				// Other case, print warning and return nil
+				logging.Debugf("%v", cmdErr(k8sArgs, "error getting pod: %v", err))
+				kubeClient = nil
+				err = nil
 			}
 		}
 	}
@@ -670,7 +672,7 @@ func CmdAdd(args *skel.CmdArgs, exec invoke.Exec, kubeClient *k8s.ClientInfo) (c
 	}
 
 	// set the network status annotation in apiserver, only in case Multus as kubeconfig
-	if n.Kubeconfig != "" && kc != nil {
+	if n.Kubeconfig != "" && kc != nil && kubeClient != nil {
 		if !types.CheckSystemNamespaces(string(k8sArgs.K8S_POD_NAME), n.SystemNamespaces) {
 			err = k8s.SetNetworkStatus(kubeClient, k8sArgs, netStatus, n)
 			if err != nil {
@@ -777,8 +779,10 @@ func CmdDel(args *skel.CmdArgs, exec invoke.Exec, kubeClient *k8s.ClientInfo) er
 				// If not found, proceed to remove interface with cache
 				pod = nil
 			} else {
-				// Other case, return error
-				return cmdErr(k8sArgs, "error getting pod: %v", err)
+				// Other case, print warning and return nil
+				logging.Debugf("%v", cmdErr(k8sArgs, "error getting pod: %v", err))
+				kubeClient = nil
+				err = nil
 			}
 		}
 	}
@@ -837,7 +841,7 @@ func CmdDel(args *skel.CmdArgs, exec invoke.Exec, kubeClient *k8s.ClientInfo) er
 	}
 
 	// unset the network status annotation in apiserver, only in case Multus as kubeconfig
-	if in.Kubeconfig != "" {
+	if in.Kubeconfig != "" && kubeClient != nil {
 		if netnsfound {
 			if !types.CheckSystemNamespaces(string(k8sArgs.K8S_POD_NAMESPACE), in.SystemNamespaces) {
 				err := k8s.SetNetworkStatus(kubeClient, k8sArgs, nil, in)