Just don't set the cluster CIDR on kube proxy right now

pkg/model/components/kubeproxy.go

@@ -59,18 +59,20 @@ func (b *KubeProxyOptionsBuilder) BuildOptions(o interface{}) error {
 	// * dns is set up by dns-controller
 	// * dns-controller talks to the API using the kube-proxy configured kubernetes service
 
-	if config.ClusterCIDR == "" {
-		// If we're using the AmazonVPC networking, we should omit the ClusterCIDR
-		// because pod IPs are real, routable IPs in the VPC, and they are not in a specific
-		// CIDR range that allows us to distinguish them from other IPs.  Omitting the ClusterCIDR
-		// causes kube-proxy never to SNAT when proxying clusterIPs, which is the behavior
-		// we want for pods.
-		// If we're not using the AmazonVPC networking, and the KubeControllerMananger has
-		// a ClusterCIDR, use that because most networking plug ins draw pod IPs from this range.
-		if clusterSpec.Networking.AmazonVPC == nil && clusterSpec.KubeControllerManager != nil {
-			config.ClusterCIDR = clusterSpec.KubeControllerManager.ClusterCIDR
+	/*
+		if config.ClusterCIDR == "" {
+			// If we're using the AmazonVPC networking, we should omit the ClusterCIDR
+			// because pod IPs are real, routable IPs in the VPC, and they are not in a specific
+			// CIDR range that allows us to distinguish them from other IPs.  Omitting the ClusterCIDR
+			// causes kube-proxy never to SNAT when proxying clusterIPs, which is the behavior
+			// we want for pods.
+			// If we're not using the AmazonVPC networking, and the KubeControllerMananger has
+			// a ClusterCIDR, use that because most networking plug ins draw pod IPs from this range.
+			if clusterSpec.Networking.AmazonVPC == nil && clusterSpec.KubeControllerManager != nil {
+				config.ClusterCIDR = clusterSpec.KubeControllerManager.ClusterCIDR
+			}
 		}
-	}
+	*/
 
 	// Set the kube-proxy hostname-override (actually the NodeName), to avoid #2915 et al
 	cloudProvider := kops.CloudProviderID(clusterSpec.CloudProvider)