Questions on IPv6 stateless autoconfiguration

[Jan Lühr <ff@jluehr.de>]

Hello,

I got stuck at some IPv6 deployment scenarios I haven't found any
solution yet. Maybe you can help me.

Consider three Debian servers (A,R1,R2) and two networks.
A is connected to both networks using different NICs.
R1 is a router in network 1, while R2 is a router in network 2. Both
provide IPv6 connectivity using radvd / stateless auto configuration).

Assuming, that R1, R2 announce global prefixes:
-> Are both of them installed in A's routing table as default routers
using their link local address? - Or - can radvd be configured to
provide stateless auto configuration for global prefixes without being
considered as a default router? If later, how? Does it change when using
ULA addresses?

-> What happens, if A fires up a web browser trying to reach
http://ipv6.google.com. Afaik A's sender address is chosen randomly,
using either R1's or R2's network. Is the corresponding router used,
accordingly - or - may it occur, that R1 is used in conjunction with an
IPv6 address from R2's network? If later: How can I force A to use the
corresponding router?

Assuming, R1 crashes. Does A detect this? Is R1's default route removed
from the routing table, then?

Assuming, that R1 is a wifi hotspot router, and R2 a vpn endpoint
(OpenVPN, tap, tls-server).
-> How can I force the client to use the VPN whenever possible, while
using the hotspot network for reaching the VPN-endpoint while using the
sender addresses accordingly?

Any help is appreciated :-)

Thanks in advance,
Jan