如何测试Linux主备模式（Active/Backup）绑定网卡（bond）接口

Alright, let's tackle this problem. You've got an Active/Backup bond (bond0) running with eth0 as the primary active slave, and you need to verify eth1's data path without messing up your live services. Here are the safest, non-intrusive ways to get this done:

First, let's confirm your current bond state to make sure we're working with the right setup. Run this command to check which slave is active and verify bond configuration:

cat /proc/net/bonding/bond0

Look for the "Active Slave" line (should be eth0) and confirm the mode is set to active-backup (mode=1).

Method 1: Temporarily Switch Active Slave to eth1 (Recommended) #

This is the most straightforward approach since it uses the bond driver's native functionality, and the switchover is nearly instantaneous—most workloads won't notice any disruption.

• Switch active slave to eth1:
  Run this command to force the bond to use eth1 as the active interface:

  echo eth1 > /sys/class/net/bond0/bonding/active_slave
  

  Re-run cat /proc/net/bonding/bond0 to confirm eth1 is now listed as the Active Slave.

• Test eth1's connectivity:

  • Use ping to verify basic reachability (specify bond0 as the interface to ensure traffic goes through the bond stack):

    ping -I bond0 <your-test-target-ip>
    

  • Use tcpdump on eth1 to confirm traffic is passing through it:

    tcpdump -i eth1 -n host <your-test-target-ip>
    

    You should see ICMP (ping) packets or other test traffic flowing through eth1.
  • For throughput testing, use iperf3 with a remote server (bind to bond0 to use the bond interface):

    iperf3 -c <iperf-server-ip> -B <bond0-ip>
    

• Switch back to eth0:
  Once testing is done, revert to your original active slave with:

  echo eth0 > /sys/class/net/bond0/bonding/active_slave
  

  Note: If your bond configuration has a primary parameter set (e.g., in your distro's network config file like /etc/sysconfig/network-scripts/ifcfg-bond0), the bond will automatically revert to eth0 after a reboot even if you forget to run this command.

Method 2: Test Link Health Without Switching Active Slave #

If you just want to verify eth1's physical link and basic data path without sending production traffic through it, use these checks:

• Verify eth1's link state is up:

  ip link show eth1
  

  Look for UP in the output.
• Use ethtool to check link details (speed, duplex, carrier status):

  ethtool eth1
  

• Send test traffic directly to eth1's MAC address to confirm physical path functionality:
  First, get eth1's MAC with ip link show eth1, then use arping to send ARP requests through eth1 to your gateway:

  arping -I eth1 -c 5 <gateway-ip>
  

  This confirms eth1 can send and receive frames on the network.

Key Notes #

• Both methods avoid modifying permanent network configurations, so your setup will revert to normal if you reboot or restart the network service.
• The active slave switchover is designed to be seamless—most applications won't even notice the brief transition.

内容的提问来源于stack exchange，提问作者Zaboj Campula