当我们在做一些与网络相关实验的时候，可能会遇到需要模拟真实网络延迟的场景。这里记录下如何使用tc这个linux自带的高级流控指令模拟网络延迟。

实验环境

node1 Ubuntu 20.04.5 LTS 163.221.29.105
node2 Ubuntu 20.04.5 LTS 163.221.29.106
node3 Ubuntu 20.04.5 LTS 163.221.29.208

实验目的

我们希望在node1和node2之间制造50ms延迟，在node2和node3之间制造100ms延迟，在node1和node3之间制造150ms延迟。

实验手段

利用tc指令建立流队列接管各个节点网卡的流，并利用filter规则过滤流，然后制造延迟。
在node1上执行以下指令(注意:请替换ens160为你虚拟机的网卡)

$ sudo tc qdisc add dev ens160 root handle 1: prio
$ sudo tc filter add dev ens160 parent 1: protocol ip prio 1 u32 match ip dst 163.221.29.106 flowid 1:1
$ sudo tc qdisc add dev ens160 parent 1:1 handle 10: netem delay 50ms

在node2上执行以下指令

$ sudo tc qdisc add dev ens160 root handle 1: prio
$ sudo tc filter add dev ens160 parent 1: protocol ip prio 1 u32 match ip dst 163.221.29.208 flowid 1:1
$ sudo tc qdisc add dev ens160 parent 1:1 handle 10: netem delay 100ms

在node3上执行以下指令

$ sudo tc qdisc add dev ens160 root handle 1: prio
$ sudo tc filter add dev ens160 parent 1: protocol ip prio 1 u32 match ip dst 163.221.29.105 flowid 1:1
$ sudo tc qdisc add dev ens160 parent 1:1 handle 10: netem delay 150ms

现在你就全都配置好啦

实验结果

在node1上ping node2和node3

root@node1:~# ping 163.221.29.106 -c 4
PING 163.221.29.106 (163.221.29.106) 56(84) bytes of data.
64 bytes from 163.221.29.106: icmp_seq=1 ttl=64 time=50.3 ms
64 bytes from 163.221.29.106: icmp_seq=2 ttl=64 time=50.4 ms
64 bytes from 163.221.29.106: icmp_seq=3 ttl=64 time=50.3 ms
64 bytes from 163.221.29.106: icmp_seq=4 ttl=64 time=50.5 ms

--- 163.221.29.106 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 50.302/50.357/50.456/0.060 ms
root@node1:~# ping 163.221.29.208 -c 4
PING 163.221.29.208 (163.221.29.208) 56(84) bytes of data.
64 bytes from 163.221.29.208: icmp_seq=1 ttl=64 time=151 ms
64 bytes from 163.221.29.208: icmp_seq=2 ttl=64 time=151 ms
64 bytes from 163.221.29.208: icmp_seq=3 ttl=64 time=151 ms
64 bytes from 163.221.29.208: icmp_seq=4 ttl=64 time=151 ms

--- 163.221.29.208 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 150.548/150.604/150.681/0.053 ms

在node2上ping node1和node3

root@node2:~# ping 163.221.29.105 -c 4
PING 163.221.29.105 (163.221.29.105) 56(84) bytes of data.
64 bytes from 163.221.29.105: icmp_seq=1 ttl=64 time=50.5 ms
64 bytes from 163.221.29.105: icmp_seq=2 ttl=64 time=50.3 ms
64 bytes from 163.221.29.105: icmp_seq=3 ttl=64 time=50.5 ms
64 bytes from 163.221.29.105: icmp_seq=4 ttl=64 time=50.4 ms

--- 163.221.29.105 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 50.348/50.436/50.508/0.057 ms
root@node2:~# ping 163.221.29.208 -c 4
PING 163.221.29.208 (163.221.29.208) 56(84) bytes of data.
64 bytes from 163.221.29.208: icmp_seq=1 ttl=64 time=101 ms
64 bytes from 163.221.29.208: icmp_seq=2 ttl=64 time=101 ms
64 bytes from 163.221.29.208: icmp_seq=3 ttl=64 time=101 ms
64 bytes from 163.221.29.208: icmp_seq=4 ttl=64 time=100 ms

--- 163.221.29.208 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 100.394/100.506/100.573/0.068 ms

在node3上ping node1和node2

root@node3:~# ping 163.221.29.105 -c 4
PING 163.221.29.105 (163.221.29.105) 56(84) bytes of data.
64 bytes from 163.221.29.105: icmp_seq=1 ttl=64 time=150 ms
64 bytes from 163.221.29.105: icmp_seq=2 ttl=64 time=150 ms
64 bytes from 163.221.29.105: icmp_seq=3 ttl=64 time=150 ms
64 bytes from 163.221.29.105: icmp_seq=4 ttl=64 time=150 ms

--- 163.221.29.105 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 150.371/150.427/150.499/0.046 ms
root@node3:~# ping 163.221.29.106 -c 4
PING 163.221.29.106 (163.221.29.106) 56(84) bytes of data.
64 bytes from 163.221.29.106: icmp_seq=1 ttl=64 time=100 ms
64 bytes from 163.221.29.106: icmp_seq=2 ttl=64 time=101 ms
64 bytes from 163.221.29.106: icmp_seq=3 ttl=64 time=100 ms
64 bytes from 163.221.29.106: icmp_seq=4 ttl=64 time=101 ms

--- 163.221.29.106 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 100.443/100.563/100.661/0.097 ms

讨论

tc是一个神奇的linux指令，还有很多高级功能有待挖掘，对于正在研究SDN之类课题的人帮助极大。关于tc指令的更多细节请参考link。(ps:其实这也是我为了测试blog post水的first post😂)