实验步骤1-创建网络空间，设备，并配置
HOST_IP=$( hostname -I | awk '{print $1}')
POD1_IP=10.96.30.11
POD2_IP=10.96.30.12
GATEWAY_IP=169.254.1.1
#1. 创建network namespace
ip netns add pod1
ip netns add pod2
#2. 创建veth pair 并加入到namespace
ip link add eth0 type veth peer name eth0-p1
ip link set eth0 netns pod1
ip link add eth0 type veth peer name eth0-p2
ip link set eth0 netns pod2
#配置设备IP地址
ip netns exec pod1 ip addr add $POD1_IP/24 dev eth0
ip netns exec pod2 ip addr add $POD2_IP/24 dev eth0
#启动设备
ip netns exec pod1 ip link set eth0 up
ip netns exec pod2 ip link set eth0 up
ip link set eth0-p1 up
ip link set eth0-p2 up
#启动回环设备
ip netns exec pod1 ip link set lo up
ip netns exec pod2 ip link set lo up
#为设备添加路由--必须先添加默认设备端口，否则报网络不可达
ip netns exec pod1 ip route add $GATEWAY_IP dev eth0 scope link
ip netns exec pod2 ip route add $GATEWAY_IP dev eth0 scope link
ip netns exec pod1 ip route add default via $GATEWAY_IP dev eth0
ip netns exec pod2 ip route add default via $GATEWAY_IP dev eth0
#3. 设置veth pair 的proxy arp
echo 1 > /proc/sys/net/ipv4/conf/eth0-p1/proxy_arp
echo 1 > /proc/sys/net/ipv4/conf/eth0-p2/proxy_arp
#4. 设置主机路由
ip route add $POD1_IP dev eth0-p1
ip route add $POD2_IP dev eth0-p2
实验步骤2-在pod1中ping pod2测试两个ns是否连通，同时在四个设备上抓包：
ip netns exec pod1 ping -c 1 10.96.30.12 #pod1 eth0设备抓包 ip netns exec pod1 tcpdump -pne -i eth0 #eth0-p1设备抓包 tcpdump -pne -i eth0-p1 #pod2 eth0设备抓包 ip netns exec pod2 tcpdump -pne -i eth0 #eth0-p1设备抓包 tcpdump -pne -i eth0-p2
pod1 eth0设备抓包内容：
10:34:36.866502 d2:3a:94:8c:41:e5 > Broadcast, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.12 tell 10.96.30.11, length 28 10:34:36.921359 0e:05:60:f2:ac:47 > d2:3a:94:8c:41:e5, ethertype ARP (0x0806), length 42: Reply 10.96.30.12 is-at 0e:05:60:f2:ac:47, length 28 10:34:36.921395 d2:3a:94:8c:41:e5 > 0e:05:60:f2:ac:47, ethertype IPv4 (0x0800), length 98: 10.96.30.11 > 10.96.30.12: ICMP echo request, id 19231, seq 1, length 64 10:34:36.921484 0e:05:60:f2:ac:47 > d2:3a:94:8c:41:e5, ethertype IPv4 (0x0800), length 98: 10.96.30.12 > 10.96.30.11: ICMP echo reply, id 19231, seq 1, length 64 10:34:42.174413 0e:05:60:f2:ac:47 > d2:3a:94:8c:41:e5, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.11 tell 192.168.30.211, length 28 10:34:42.174449 d2:3a:94:8c:41:e5 > 0e:05:60:f2:ac:47, ethertype ARP (0x0806), length 42: Reply 10.96.30.11 is-at d2:3a:94:8c:41:e5, length 28
eth0-p1抓包内容：
10:34:36.866524 d2:3a:94:8c:41:e5 > Broadcast, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.12 tell 10.96.30.11, length 28 10:34:36.921343 0e:05:60:f2:ac:47 > d2:3a:94:8c:41:e5, ethertype ARP (0x0806), length 42: Reply 10.96.30.12 is-at 0e:05:60:f2:ac:47, length 28 10:34:36.921397 d2:3a:94:8c:41:e5 > 0e:05:60:f2:ac:47, ethertype IPv4 (0x0800), length 98: 10.96.30.11 > 10.96.30.12: ICMP echo request, id 19231, seq 1, length 64 10:34:36.921483 0e:05:60:f2:ac:47 > d2:3a:94:8c:41:e5, ethertype IPv4 (0x0800), length 98: 10.96.30.12 > 10.96.30.11: ICMP echo reply, id 19231, seq 1, length 64 10:34:42.174397 0e:05:60:f2:ac:47 > d2:3a:94:8c:41:e5, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.11 tell 192.168.30.211, length 28 10:34:42.174450 d2:3a:94:8c:41:e5 > 0e:05:60:f2:ac:47, ethertype ARP (0x0806), length 42: Reply 10.96.30.11 is-at d2:3a:94:8c:41:e5, length 28
eth0-p2抓包内容：
10:34:36.921439 d2:e5:fb:a0:e6:c7 > Broadcast, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.12 tell 192.168.30.211, length 28 10:34:36.921453 3e:63:d4:1b:a8:b0 > d2:e5:fb:a0:e6:c7, ethertype ARP (0x0806), length 42: Reply 10.96.30.12 is-at 3e:63:d4:1b:a8:b0, length 28 10:34:36.921455 d2:e5:fb:a0:e6:c7 > 3e:63:d4:1b:a8:b0, ethertype IPv4 (0x0800), length 98: 10.96.30.11 > 10.96.30.12: ICMP echo request, id 19231, seq 1, length 64 10:34:36.921478 3e:63:d4:1b:a8:b0 > d2:e5:fb:a0:e6:c7, ethertype IPv4 (0x0800), length 98: 10.96.30.12 > 10.96.30.11: ICMP echo reply, id 19231, seq 1, length 64 10:34:42.174426 3e:63:d4:1b:a8:b0 > d2:e5:fb:a0:e6:c7, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.11 tell 10.96.30.12, length 28 10:34:42.174456 d2:e5:fb:a0:e6:c7 > 3e:63:d4:1b:a8:b0, ethertype ARP (0x0806), length 42: Reply 10.96.30.11 is-at d2:e5:fb:a0:e6:c7, length 28
pod2 eth0抓包内容：
10:34:36.921442 d2:e5:fb:a0:e6:c7 > Broadcast, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.12 tell 192.168.30.211, length 28 10:34:36.921451 3e:63:d4:1b:a8:b0 > d2:e5:fb:a0:e6:c7, ethertype ARP (0x0806), length 42: Reply 10.96.30.12 is-at 3e:63:d4:1b:a8:b0, length 28 10:34:36.921456 d2:e5:fb:a0:e6:c7 > 3e:63:d4:1b:a8:b0, ethertype IPv4 (0x0800), length 98: 10.96.30.11 > 10.96.30.12: ICMP echo request, id 19231, seq 1, length 64 10:34:36.921477 3e:63:d4:1b:a8:b0 > d2:e5:fb:a0:e6:c7, ethertype IPv4 (0x0800), length 98: 10.96.30.12 > 10.96.30.11: ICMP echo reply, id 19231, seq 1, length 64 10:34:42.174421 3e:63:d4:1b:a8:b0 > d2:e5:fb:a0:e6:c7, ethertype ARP (0x0806), length 42: Request who-has 10.96.30.11 tell 10.96.30.12, length 28 10:34:42.174457 d2:e5:fb:a0:e6:c7 > 3e:63:d4:1b:a8:b0, ethertype ARP (0x0806), length 42: Reply 10.96.30.11 is-at d2:e5:fb:a0:e6:c7, length 28
实验分析：
1.pod1 发送icmp包
pod1不知道10.96.30.12这个ip地址对应的mac地址，无法完成以太网帧，因此会先通过arp协议转换
1.1 pod1中的eth0网卡发出arp请求（为什么从eth0发出？pod1配置了默认路由,从eth0接口发出）
1.2 eth0-p1所在的宿主机因为收到arp请求, 会自动更新自己的arp表项：(10.96.30.11) at d2:3a:94:8c:41:e5 [ether] on eth0-p1
1.3 eth0-p1开启了 proxy arp，因此用自己的mac回复该arp请求
1.4 pod1得到回复后，更新自己的arp表，表项为：(10.96.30.12) at 0e:05:60:f2:ac:47 [ether] on eth0 。其实eth0被eth0-p1骗了
2. pod1的arp完成，紧接着发出icmp包（ping echo）,该网络包自然会到达eth0-p1
3. eth0-p1是宿主机网卡，宿主机收到该icmp包，并且由于宿主机开启了ip forward 因此会转发该包。
宿主机转发icmp包时，宿主机自己也不知道10.96.30.12的mac,因此宿主机发出arp广播请求
3.1. 该arp广播请求从eth0-p2发出，为什么？因为宿主机配置了路由，所有到10.96.30.12的包从eth0-p2接口出去
3.2. eth0-p2发出arp广播包（Request who-has 10.96.30.12 tell 192.168.30.211），pod2中的eth0回复该请求（Reply 10.96.30.12 is-at 3e:63:d4:1b:a8:b0）
3.3. eth0-p2收到回复，亦即宿主机收到回复，宿主机更新自己的arp表，表项为：(10.96.30.12) at 3e:63:d4:1b:a8:b0 [ether] on eth0-p2
3.4 宿主机将icmp包装入以太网帧并发出
4. pod2中的eth0收到该icmp包，发现是在ping自己，因此回复该icmp包
pod2自己也不知道10.96.30.11的mac,因此发出arp广播请求
4.1. 该arp广播请求从eth0发出
4.2. eth0-p2收到arp广播包，由于自己开启了proxy arp 因此eth0-p2用自己的mac回复该arp广播
4.3. pod2中的eth0收到回复，更新自己的arp表，表项为： (10.96.30.11) at d2:e5:fb:a0:e6:c7 [ether] on eth0
4.4 pod2将icmp回复包装入以太网帧并发出给eth0-p2
8. 宿主机的eth0-p2收到pod2回复的icmp包，由于自己开启了ip forward，因此会转发该包，同时由于宿主机知道10.96.30.11的mac(arp表项已经存在)，便直接封装成以太网帧通过eth0-p1发送（路由表知道10.96.30.11需要通过该接口发送，arp表中也有）
9. pod1中eth0收到icmp回复，至此ping完成。
标签：ip,ipip,length,实验,10.96,e5,d2,calico,eth0 From： https://www.cnblogs.com/lucidar/p/18070107