RAdv protocol in failover router setup

[Tore Anderson]

Hi Nico,

> we are running active-active routers with bird, wihout VRRP as follows:
> 
> - router1: high priority
> - router2: low priority
> 
> Sample config:
> 
> protocol radv {
>     # Internal
>     interface "eth1.10" {
>          max ra interval 5;      # Fast failover with more routers
>          other config yes;       # dhcpv6 boot
>          default preference high;
>          default lifetime 600;
>     };
>     # Server
>     interface "eth1.11" {
>          max ra interval 5;
>          default preference high;
>          default lifetime 600;
>     };
> }

If I am not mistaken «default lifetime 600» here implies that it would take
ten minutes after «router1» has crashed for its high-priority default route to
expire from downstream nodes, which is neccessary in order for them to move on
to using «router2»'s lower priority route instead.

We need this to happen in a couple of seconds at most.

> Routing via EUI-64 based link local addresses.
> 
> This works fine with Linux servers, however Windows/Linux clients do not
> necessarily seem to accept the higher/lower preference route.

Well, that is precisely what I want to avoid. Depending on IPv6 ND for
failover does not work too well with many clients, so ensuring that there is
only a single unambiguous next-hop being advertised (i.e., fe80::1 in my case)
is necessary in order for failovers to be reliable and fast.

Otherwise we tend to run into client bugs, such as the one described here:
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1800836

Tore