linuxexpert-practice-labs/labs/lfcs/2-7-bridges-bonding/README.md

LFCS 2.7: Bridges and bonding

An active-backup bond, a bridge carrying the host's address, and a guest attached to the bridge.

Before you start

Instead of extra network cards, the lab gives this VM four extra interfaces, each cabled to a small lab network:

• le-b0 and le-b1 both go to one switch, where a machine answers at 10.96.0.254.
• le-p0 goes to another network, 10.95.0.0/24, where a machine called lan has 10.95.0.2. le-p0 has this VM's address there, 10.95.0.1/24: a NetworkManager profile called le-p0 on the RHEL family, /etc/netplan/60-le-lab.yaml on Ubuntu.
• le-vnet0 is a guest's cable, the way libvirt connects a VM through a vnet interface. The guest has 10.95.0.50/24; it's connected to nothing yet.

Run commands on the switch, lan and the guest with sudo ip netns exec le-sw|le-lan|le-guest COMMAND. Your SSH connection doesn't use any of these interfaces.

Tasks

1. RHEL family: create an active-backup bond, bond0, from le-b0 and le-b1, with the static address 10.96.0.1/24, persistently.
2. Ubuntu: do the same with netplan.
3. Create a bridge br0 with le-p0 as its port and this VM's address on that network, 10.95.0.1/24, moved to the bridge, persistently.
4. Attach the guest to br0 (add le-vnet0 as a second port), so it's on the 10.95.0.0/24 network and lan can reach it at 10.95.0.50.

Hints

• nmcli connection add type bond ... bond.options "mode=...,miimon=...", then one type ethernet profile per member with master.
• bonds: with interfaces: and parameters: in netplan; members under ethernets:.
• cat /proc/net/bonding/bond0; to test failover, take the active member's other end down: sudo ip netns exec le-sw ip link set le-b0s down (or le-b1s), ping 10.96.0.254, then bring it back up.
• nmcli connection add type bridge ... bridge.stp no, or bridges: in netplan with stp: false; take the address off le-p0's own profile.
• bridge link, ip -br addr, sudo ip netns exec le-lan ping 10.95.0.50.