Alright, let's walk through setting up and managing this multi-container test setup step by step—here's how to make everything work smoothly:

First, you need to enable IP forwarding on Container2, which is critical for routing traffic between Container1 and Container3:

• Access Container2's shell: docker exec -it Container2 bash
• Enable IP forwarding temporarily: echo 1 > /proc/sys/net/ipv4/ip_forward
• To make this setting persistent across container restarts, edit /etc/sysctl.conf, uncomment the line net.ipv4.ip_forward=1, then run sysctl -p to apply changes.

Next, set the default gateway of Container1 and Container3 to point to Container2's IP:

• For Container1: docker exec -it Container1 ip route add default via <Container2_IP>
• For Container3: docker exec -it Container3 ip route add default via <Container2_IP>

To run commands on Container2 and Container3 via SSH, install and configure the SSH server in both containers:

• Access the target container (take Container3 as an example): docker exec -it Container3 bash
• Install openssh-server: apt update && apt install -y openssh-server
• Allow root login (this is convenient for testing, not recommended for production environments): Edit /etc/ssh/sshd_config, change PermitRootLogin prohibit-password to PermitRootLogin yes
• Set a root password: passwd root
• Restart the SSH service: service ssh restart

Now you can connect from your client (or the host machine) via SSH:

• Connect to Container2: ssh root@<Container2_IP>
• Connect to Container3: ssh root@<Container3_IP>
  You can also run single commands directly without entering the shell, e.g.: ssh root@<Container3_IP> "tc qdisc show"

Assuming you have a custom server program, here's how to run it and gather connection stats:

• Start the server in the background on Container3: nohup ./your-server-program > server-output.log 2>&1 &
• To collect connection statistics, use tools like ss or tcpdump:
  • View real-time connections: ssh root@<Container3_IP> "ss -tulnp | grep <your-server-port>"
  • Capture traffic to a file for later analysis: ssh root@<Container3_IP> "tcpdump -i any port <your-server-port> -w capture.pcap"
  • Alternatively, build statistics logic directly into your server program and output data to a log file for easy comparison.

tc netem is used to simulate network conditions like latency or packet loss. Here's how to update parameters and restart testing:

• First, clear any existing qdisc rules on Container3 (if present): ssh root@<Container3_IP> "tc qdisc del dev eth0 root"
• Add new network emulation rules, for example:
  • Add 100ms latency: ssh root@<Container3_IP> "tc qdisc add dev eth0 root netem delay 100ms"
  • Add 5% packet loss: ssh root@<Container3_IP> "tc qdisc add dev eth0 root netem loss 5%"
• Restart your server program to apply changes (or just initiate new test connections):
  • Find the server's PID: ssh root@<Container3_IP> "ps aux | grep your-server-program"
  • Kill the existing process: ssh root@<Container3_IP> "kill <PID>"
  • Restart the server: ssh root@<Container3_IP> "nohup ./your-server-program > new-test-output.log 2>&1 &"
• Finally, initiate new connections from your client and collect updated statistics to compare with previous test results.

内容的提问来源于stack exchange，提问作者AmithRc