iBGP

[Alexander V. Chernikov]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

fredrik danerklint wrote:
> Setup:
> 
> Transit <-> CR2 <-> ER2
> 
> Loopback addressess:
> CR2 has the ip adress of xx.xx.xx.2/32.
> ER2 has the ip adress of xx.xx.xx.4/32.
> 
> Network link:
> CR2 has the ip address xx.xx.xx.201/30
> ER2 has the ip address xx.xx.xx.202/30
> 
> All announced via ospf so ping and traceroute does work between each other.
> 
> bird> show route  for 77.73.232.6 all
> 77.73.232.0/21     via xx.xx.xx.201 on igb2 [b4_sekkaa_1_cr2 2011-09-30 
Real (announced to kernel) nexthop is xx.xx.xx.201
> 22:18:34 from xx.xx.xx.2] * (100/6) [AS42678i]
>         Type: BGP unicast univ
>         BGP.origin: IGP
>         BGP.as_path: 35706 8359 25513 42678
>         BGP.next_hop: xx.xx.xx.2
CR2 announces you its source (loopback) address which is OK
>         BGP.med: 0
>         BGP.local_pref: 100

Since your session is multihop and (at least FreeBSD) kernel can't
resolve recursive routes this is done by bird (gateway recursive is set
in this case). It changes received gateway with gateway via which
xx.xx.xx.2 is directly reachable. This is xx.xx.xx.201.

I can't see any problem here. Please correct me if I'm wrong.

> 
> $netstat -rn | grep 77.73.232.0/21
> 77.73.232.0/21     xx.xx.xx.201      UG1         0        0   igb2
> 
> protocol bgp b4_sekkaa_1_cr2 
> {
>     description "sekkaa-1-cr2";
>     import filter {
>         accept;
>     };
>     export filter {
>         accept;
>     };
>     
>     interpret communities off;
>     local as my_as;
>     default bgp_med 0;
>     default bgp_local_pref 200;
>     neighbor xx.xx.xx.2 as 52011;
>     source address xx.xx.xx.4;
>         start delay time 30;
>         next hop self;
>         multihop 4;
> }
> 
> This is from the internal router ER2. 
> 
> 
> 
> 
> 
> fredrik danerklint wrote:
>>>> I already have the 'next hop self;' set in both of the routers. still no
>>>> luck.
> Can you show BGP configuration block on both routers?
> 
> It would be great to see
> 'show route [table XXX] for some_prefix_advertised_by_router1 all' on
> both routers, too
> 
>>>> fredrik danerklint wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> FreeBSD 8.2 and Bird 1.3.3.
>>>>>>>
>>>>>>> I do my internal peering between routes on loopback interfaces.
>>>>>>> All the loopback interfaces is announced over ospf.
>>>>>>>
>>>>>>> To be able to get an bgp session the 'multi hop 4;' is needed.
>>>>>>>
>>>>>>> The question I have is what value should the 'next hop' be? I would
>>>>>>> like to have it to the neighbour address of the peering session but
>>>>>>> it seems it always the nearest physical network interface address (to
>>>>>>> the neighbour) that's 'next hop'.
>>>> RFC recommendation for IBGP session tells us:
>>>>
>>>> When sending a message to an internal peer, if the route is not
>>>>
>>>>          locally originated, the BGP speaker SHOULD NOT modify the
>>>>          NEXT_HOP attribute unless it has been explicitly configured to
>>>>          announce its own IP address as the NEXT_HOP
>>>>
>>>> (RFC 4271 clause 5.1.3)
>>>>
>>>> You can alter next hop via 'next hop self' config option. It will set
>>>> next hop address to whatever you configure in 'source address' option.
>>>>
>>>>>>> I did even try with setting bgp_next_hop=10.10.10.11; within the
>>>>>>> filter of the bgp protocol but that did not work.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.14 (FreeBSD)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk6GR2kACgkQwcJ4iSZ1q2lDXQCfQLuz4UEQWdr8AOnGKjJVlewD
6P4AnR9mPE06cgvTKDFWmamAxECHjYh0
=KNFI
-----END PGP SIGNATURE-----