[6bone] AS1654/SICS announce 178 prefixes !

[Lars Albertsson]

Gert Doering <gert@space.net> writes:

> Hi,
> 
> On Sun, Jun 08, 2003 at 09:17:27PM +0200, Nicolas DEFFAYET wrote:
> >    Network          Next Hop            Metric LocPrf Weight Path
> > *  2001:210::/35    2001:748:200:a0::6
> >                                                   100      0 5430 13285
> > 6939 6939 3320 9112 2847 1654 i
> 
> Seconded, I see that as well - 1654 is originating half of the prefixes
> as their own.  BGP bug or human goof...
> 
> The interesting part is that 1654 does have a number of links, but these
> prefixes only go out via 2847 1654.  Someone from 2847 around who can 
> figure out what's happening and make it stop?
> 
> Does anybody know the background of this?  Any peers of 1654 around?

I am sorry if we are causing problems again. I try to keep a very
simple configuration in order to avoid routing problems, and I don't
understand the cause for this. AS2847 (Litnet) is one of our
delegations. We announce the full IPv6 routing table, including the
prefixes mentioned by Nicolas, to our downstream sites, but prefixes
other than ours are of course not meant to be announced with our AS as
last hop, and I don't understand why they end up that way. And I
expect the sites downstream from us not to announce prefixes outside
their address space to other sites.

I looked through our configuration files and bgpd log, and couldn't
see anything suspicious. I restarted zebra and monitored the traffic
to litnet's router with tcpdump, and it seems we are announcing
correct AS paths to them:

15:22:38.755586 alt.sics.se > gw.ipv6.lt: alt-litnet.ipv6.sics.se.3895 > 3ffe:20
0:1:e::2.bgp: P 8029:8588(559) ack 29731 win 355: BGP [|BGP] (UPDATE: (Path attr
ibutes: (ORIGIN[T] IGP)
                (AS_PATH[T] 1654 6342 2549)
                (MP_REACH_NLRI[O] IPv6 Unicast, nexthop alt-litnet.ipv6.sics.se,
 NLRI 2001:b18::/32)))
        (UPDATE: (Path attributes: (ORIGIN[T] IGP)
                (AS_PATH[T] 1654 3274 16023)
                (MP_REACH_NLRI[O] IPv6 Unicast, nexthop alt-litnet.ipv6.sics.se,
 NLRI 2001:b30::/32)))
        (UPDATE: (Path attributes: (ORIGIN[T] IGP)
                (AS_PATH[T] 1654 3274 790 6667)
                (MP_REACH_NLRI[O] IPv6 Unicast, nexthop alt-litnet.ipv6.sics.se,
 NLRI yggdrasil-gw.ipv6.sevenlevels.net/32)))
        (UPDATE: (Path attributes: (ORIGIN[T] IGP)
                (AS_PATH[T] 1654 6342 6435 13944 25396)
                (MP_REACH_NLRI[O] IPv6 Unicast, nexthop alt-litnet.ipv6.sics.se,
 NLRI 2001:b70::/32)))
        (UPDATE: (Path attributes: (ORIGIN[T] IGP)
                (AS_PATH[T] 1654 6342 2549 3320)
                (MP_REACH_NLRI[O] IPv6 Unicast, nexthop alt-litnet.ipv6.sics.se,
 NLRI 2001:bb8::/32)))
        (UPDATE: (Path attributes: (ORIGIN[T] IGP)
                (AS_PATH[T] 1654 6342)
                (MP_REACH_NLRI[O] IPv6 Unicast, nexthop alt-litnet.ipv6.sics.se,
 NLRI 2001:c00::/32)))
        (UPDATE: (Path attributes: (ORIGIN[T] IGP)
                (AS_PATH[T] 1654 6342 109)
                (MP_REACH_NLRI[O] IPv6 Unicast, nexthop alt-litnet.ipv6.sics.se,
NLRI 2001:c08::/32))) [hlim 1] (len 579) (ttl 45, id 53505)


I am clueless on what could be wrong, and where. If anyone has any
ideas on something I can look for, feel free to speak up. If you want,
you can also look at the state and configuration for our router at
http://www.ipv6.sics.se/6bone_config/.

The last Merit routing report
(http://www.merit.edu/mail.archives/html/6bone-routing-report/msg00100.html)
claims that weird prefixes from us are leaking through AS64734 as
well. It is owned by VSH, which is one of our NLAs. I am therefore
CC:ing their contact person.

Regards,

Lalle