[c-nsp] unknown interface for ipv6 route in 4.2.3

[Tassos Chatzithomaoglou]

For everyone interested, this got id CSCue96950

Bug Details
Outgoing Interface for IPv6 BGP Routes Became Unknown

Symptom:
IPv6 routes pointing to *unknown* interface after interface flapping, disturbing the traffic forwarding and requires hard reset of BGP session:


RP/0/RP0/CPU0:CISCO#show route ipv6 bgp
Mon Mar 4 16:13:45.458 CET
B 2001:600::/32
[200/0] via fe80::d2d0:fdff:fe0d:9979, 00:01:05, Unknown


Conditions:
"bgp fast-external-fallover disable" is configured and
IPv6 BGP Session is up and
Fast deletion and reconfiguration of an interface.


Workaround:
1- Enable "bgp fast-external-fallover" or
2- Reset BGP session

--
Tassos

Tassos Chatzithomaoglou wrote on 30/1/2013 21:51:
> tbh, this is something i didn't even think about...and i don't know why they chose to go
> with /48.
>
> https://www.peeringdb.com/private/exchange_view.php?id=347
>
>
> PS: sorry for the previous "pgp" email.
>
>
> --
> Tassos
>
>
> Gert Doering wrote on 30/1/2013 21:09:
>> Hi,
>>
>> On Wed, Jan 30, 2013 at 08:57:25PM +0200, Tassos Chatzithomaoglou wrote:
>>> Yes, that's the LAN of the local IX.
>>>
>>> interface TenGigE0/2/0/3.290851
>>>  description ** GRIX **
>>>  ipv4 address x.x.x.x/24
>>>  ipv6 address 2001:648:2100::x/48
>> This... is amazing.  And then people complain that a /64 is too big for
>> a single LAN.
>>
>> I'd expect more bugs and unexpected behaviour - implementations get tested
>> with "LAN = /64", and sometimes with "/112" or "/124", but I'd expect
>> most interesting results for connected networks with shorter masks.
>>
>> gert