Benoit Branciard wrote:
>> I suspected it may be useful to set sysctl 
>> net.ipv6.conf.<iface>.forwarding to 1, but :
>> - this turns the box in "router" mode, disabling RA acceptance from 
>> default router (annoying)
> 
> IPv6 forwarding is still mandatory, but this implies manual 
> configuration of IP address, netmask and default route on hardware node. 
> This last point may be annoying: in our case the default route is 
> hardware-dependent, so if some day we need to replace the router (or the 
> router interface), default route will become invalid and manual 
> reconfiguration of all openVZ hardware nodes will be needed.
> Too bad Linux kernel doesn't allow enabling forwarding without disabling 
> autoconfiguration.

I'm not sure, but I think that my setup is router independent. For my 
virtual servers I got prefix 2001:718:1:e::/64 (which is unbelivable 
waste of IP range but they say this is normal in IPv6). This network 
block is asigned to dedicated VLAN. IPv6 Gateway for all physical 
systems in that network is 2001:718:1:e::1.

My setup on HW node is:

iface eth0 inet6 static
         post-up /sbin/sysctl -w "net.ipv6.conf.eth0.autoconf=0"
         post-up /sbin/sysctl -w "net.ipv6.conf.all.forwarding=1"
         address 2001:718:1:e::23:3254
         gateway 2001:718:1:e::1
         netmask 64

I'm trying to stay away of MAC dependent automaticaly asigned IPs 
because of dependence on hardware.

I hope this might you, if needed I might try to ask our network 
specialists about more details...

Best regards
-- 
-----------------------
Jan Tomasek aka Semik
http://www.tomasek.cz/