正文
修改docker daemon启动参数，增加以下参数后重启docker daemon：
node1: --bip=192.1.78.1/24
node2: --bip=192.1.87.1/24
这时node1和node2的容器之间还不能直接通信， node1也不能跨主机和node2上的容器直接通信，反之node2也无法直接和node1上的容器通信.。
第二步、创建VTEPs
在node1上执行以下命令：
PREFIX=vxlan
IP=$external-ip-of-node-1
DESTIP=$external-ip-of-node-2
PORT=8579
VNI=1
SUBNETID=78
SUBNET=192.$VNI.0.0/16
VXSUBNET=192.$VNI.$SUBNETID.0/32
DEVNAME=$PREFIX.$VNI
ip link delete $DEVNAME
ip link add $DEVNAME type vxlan id $VNI dev eth0 local $IP dstport $PORT nolearning
echo '3' > /proc/sys/net/ipv4/neigh/$DEVNAME/app_solicit
ip address add $VXSUBNET dev $DEVNAME
ip link set $DEVNAME up
ip route delete $SUBNET dev $DEVNAME scope global
ip route add $SUBNET dev $DEVNAME scope global
node2上执行以下命令：
PREFIX=vxlan
IP=$external-ip-of-node-2
DESTIP=$external-ip-of-node-1
VNI=1
SUBNETID=87
PORT=8579
SUBNET=192.$VNI.0.0/16
VXSUBNET=192.$VNI.$SUBNETID.0/32
DEVNAME=$PREFIX.$VNI
ip link delete $DEVNAME
ip link add $DEVNAME type vxlan id $VNI dev eth0 local $IP dstport $PORT nolearning
echo '3' > /proc/sys/net/ipv4/neigh/$DEVNAME/app_solicit
ip -d link show
ip addr add $VXSUBNET dev $DEVNAME
ip link set $DEVNAME up
ip route delete $SUBNET dev $DEVNAME scope global
ip route add $SUBNET dev $DEVNAME scope global
第三步、为VTEP配置forward table
# node1
node1$ bridge fdb add $mac-of-vtep-on-node-2 dev $DEVNAME dst $DESTIP
node2
node2$ bridge fdb add $mac-of-vtep-on-node-1 dev $DEVNAME dst $DESTIP
第四步、配置Neighbors，IPv4中为ARP Table
