Hello,

I've been searching everywhere i could but i didn't find a large scale solution for IPv6 only network with RouterOS.

In fact, according to the informations i got, with IPv6, you can't have Framed-IP-Address because IPv6 Address don't rely on PPP but on IPv6 itself:

1) You have to use Framed-IPv6-Prefix but the client side has no idea of its attributed prefix

2) You can set attributed Prefix statically on client LAN interface(s) on initial provisioning but your router won't be accessible through a "WAN" IP, because the PPPoE attributed IPv6 address is local (fe80)

3) You can use DHCPv6-PD but it is totally static (and buggy), each pppoe server should use a static pool and prefixes are then affected to clients dynamically, but because DHCPv6 doesn't rely on radius, you have no idea on which prefix from the pool is affected to which router.

So, if you want to manage your router, you should use a static IPv6 address you have set manually on the LAN side (at provisioning time), but if you then want to change it... problems begins because you don't have other ip to manage your router.

You then have to use radius to know the NAS address, then script to search DHCPv6 bindings that matches your router (if you find it through the UID that might be different than the ethernet mac used for pppoe login).

I think, if all these facts are true, that's not a large scale optimized situation. Worse, if you suceed in implementing these tweaks, your lucky if they work....

If i'm wrong somewhere or if somebody have ideas on how to achieve large scale IPv6 only networks, please let me know...

If Mikrotik have prevision on a date to get, at least, RADIUS accounting for DHCPv6-PD with information on pppoe interface / username source of the DHCPv6 client request, so it would be easier to know which router got which IP address, that would be great.

Maybe the actual thinking is to use IPv6 only for Clients LAN side, and keep IPv4 for router managing purposes, but that's a bit risky for the future evolutions.

Thank you for your comments and your help

Daren

Try to do a bgp to route subnets to the pppoe terminators. Then it don't matter, what IP is to where. its routed dynamic. all pppoe terminators could have the same big subnet.

Ok for the server side but the biggest problem is the client side...
Backbone routing is not an issue.

Thank you for your reply

I've tested it on the backbone static IPv6 routing does work and I can make an IPv6 tunnel over IPv4 as well, but I am still a bit cautious to push it any larger - ROS quality is still too sporadic for that. Let them tinker with it a year or two.

That would be interesting to get a Mikrotik view of the future of ROS for IPv6, like a roadmap for example...

But what route do you actually need if you setup BGP between nas'es, and give out from one big pool to the clients? then the pppoe, who become the clients default gateway, will fix all the routing (if you see the nas'es as backbone with bgp)

With IPv4, your router get a WAN IP Address automatically through PPPoE (with radius, a Framed-IP-Address),
you only have to set router IP addresses if the client have a routed subnet (with radius, a Framed-Route)
When you need to manage the router, you connect to the WAN IP Address

With IPv6, you have a routed subnet, (with radius Framed-IPv6-Prefix), but the router don't get a WAN (routable) IP Address unless you use DHCPv6-PD.
But even if you use DHCPv6-PD, you can't use radius to decide or to know which prefix has been delegated to which client (with radius accounting for example), so router management is not easy...

Is it easier to understand what i mean with this comparison?

I understand. and I know. I have implemented this my self.

Still, the NAS will know the subnet given to the customer, and will know the "route" by the interface (pppoe) used to give out the net.

The cpe you give one ip. (Not the whole subnet in first place), at pppoe interface. You put the same address to another interface (internal) so the subnet is working. This now should be ND enabled to give internal units its ipv6 address from the subnet provided by pppoe.

The NAS wil have the route to that subnet trough the pppoe interface. If the NAS use bgp, all the other router's will know that the first nas, have the route trough the pppoe interface. ipv4 is a bit different, Couse you need the route. ipv6 have the subnet to the interface. not just a single IP.

Its not that different from ipv4, and I cannot see why radius, and radius settings, will do to much to help you to route. I know its a bit different to use the framed route, but I cannot see the use of this, since dhcpv6 give out subnets. Not single address. This means, the router it self, know about the route, trough the interface. ITs easier.