Hi,

We have a lot of RB1100 with RouterOS 5.15 at or DSL POPs for PPPoE termination (PPPoE Server) and we need to roll out IPv6 over PPPoE.
I tried to use "Framed-IPv6-Prefix" via Radius. The user route is added to the ipv6 routing table and everything is working so far. because we have more than one routers as PPPoE servers I need to have these ipv6 customer prefixes in OSPFv3. So I activated "redistribute static routes as type 2" in my OSPFv3 instance. When the users connects, the prefix gets redistributed and is seen in our Cisco core and other mikrotik routers.

The problem arises when the user disconnects (and possibly connects on another PPPoE router later): The ipv6 route is deleted from the ipv6 routing table but not from OSPFv3, OSPF still announces the prefix even if it's not in the local routing table anymore. this leads to routing loops because packets for the prefix are sent to the mikrotik router and back to the default gw because the route is not known locally on the mikrotik then.

This is clearly a bug I think. Redistribution should always happen via routing table, if a prefix is not (anymore) in the routing table it is not allowed to be inserted into ospf / needs to be deleted from ospf.

an example:
in the routing table i see the prefix (currently connected user)
15 ADS 2a03:5480:4002::/48 <pppoe-USER1> 1

but in the ospf routes i see also an old prefix, from a user which is no longer connected
15 2a03:5480:4001::/48 imported-ext-2 20
16 2a03:5480:4002::/48 imported-ext-2 20

when i disabled the "redistribute static routes" these routes disappear but as soon I re-enable it they are back (even the one from the not connected user). so where does the route come from when there is no such route? Just a reboot deletes these old prefixes, even a disable ospfv3 instance does not help.



MikroTik support, can you please tell us, if this bug is known and when it will be fixed? We need to roll out IPv6 soon and it won't work without OSPF.


thanks a lot
Marco

I've just opened this bug report:
http://bugs.mikrotik-routeros.com/view.php?id=24

OOI, did you get anywhere with this? I've been trying to get them to fix ospfv3 for months (will not reliably maintain adjacency) and so far it's just not been a priority for them at all :/

It's very bugged first don't use other areas than backbone / 0, any other area (also on other non MT routers) will break adjs when MT receives an inter-area route. that's what broke adjs for me all the time...

MT support told me the bug about redistribution from "user-static" routes is known and will be fixed in Version 6 of RouterOS, they have no plan when RouterOS6 is released though.

I only have a single area in the test setup - 0.

Pretty dire, but you get what you pay for I guess. I just wish they wouldn't advertise features that simply don't work - even when just setting up peering between two routerboards.