it would be very nice to assing vlans to bonded slaves in active-backup mode.
In this way when using bonding together with vlans (a very common setup) it
would be possible to have load-balancing or active-active together with true
fault-tolerance even with two different switches!
The setup is following:
Physical bonding VLANs
eth0 |== bond0 ==| bond0.1 Traffic for VLAN1 goes over eth0
eth1 | | bond0.2 Traffic for VLAN1 goes over eth1
When one of the physical interfaces fails boths VLANs travel over the surviving
interface. This setup is common - for example servers on a storage network with
NFS or iSCSI and a user network. We would have a setup like in the cluster world
with two different services and in case of hardware failure both services are
done by the surviving node - its called active-active cluster. There would be
no trunc configuration needed on the switches - only standard VLAN setup and
we could use different switches for both interfaces which would be true hardware
fault tolerance and could still use the bandwidth of both interfaces.