interface teaming / bonding + vlan under linux / debian

goal:

layer-2 failover without any special features on the switch level.
after recent hang of on of switches i’d like to improve reliability of connectivity within a server rack.
i already have all servers connected to two different switches… now it’s time for automated failover.

i’m experimenting with interface teaming in active-backup mode.

that’s my /etc/network/interfaces:

auto bond0
iface bond0 inet manual
        slaves eth0 eth1
        bond_mode active-backup
        bond_primary eth0
        bond_arp_interval 1000
        bond_arp_ip_target 192.168.0.10 10.16.10.1

auto bond0.22
iface bond0.22 inet static
        address 192.168.0.11
        netmask 255.255.255.0

auto bond0.11
iface bond0.11 inet static
        address 10.16.10.177
        netmask 255.255.255.0
        gateway 10.16.10.1

.. and so far it seems to work fine in the test lab

failure is not detected by the link status [ that would be quite useless – during my last hang all switche’s ports were up ], instead it sends arp-who-has to group of hosts defined in bond_arp_ip_target

20110606 – interestingly option that worked for me fine underl lenny:

bond_arp_validate 3

no longer works for me under squeeze. if i set it bond shows status of both interfaces as down.

20141023 – setup as described above has been working fine for years in production using the following topology: