On Mon Jan 12, 2026 at 12:17 PM CET, Stuart Henderson wrote:
> On 2026-01-11, Remi Locherer <[email protected]> wrote:
>> Hi,
>>
>> For a BGP route reflector setup I want to set next hop self
>> towards iBGP peers. But only for routes learnt through eBGP.
>>
>> In the RIB I can see this:
>>
>> dut# bgpctl sh rib det 172.0.42.0/24  
>>
>> BGP routing table entry for 172.0.42.0/24
>>     Nexthop 10.100.0.10 (via 10.1.0.2) Neighbor r1 (10.100.0.10)
>>     Origin IGP, metric 0, localpref 100, weight 0, ovs not-found, avs 
>> unknown, internal, valid, best
>>                                                                              
>>   ^^^^^^^^
>>     Last update: 00:58:22 ago
>>     Originator Id: 10.100.0.10
>>     Cluster Id List:  10.0.0.42
>>
>> BGP routing table entry for 172.0.42.0/24
>>     Nexthop 10.100.0.10 (via 10.1.0.2) Neighbor rr2 (10.100.0.2)
>>     Origin IGP, metric 0, localpref 100, weight 0, ovs not-found, avs 
>> unknown, internal, valid, ecmp
>>                                                                              
>>   ^^^^^^^^
>>     Last update: 00:58:24 ago
>>     Originator Id: 10.100.0.10
>>     Cluster Id List:  10.0.0.42 10.0.0.43
>>
>> BGP routing table entry for 172.0.42.0/24
>>     65101 65000
>>     Nexthop 10.1.0.14 (via 10.1.0.14) Neighbor x1 (172.42.42.1)
>>     Origin IGP, metric 0, localpref 100, weight 0, ovs not-found, avs 
>> unknown, external
>>                                                                              
>>   ^^^^^^^^
>>     Last update: 00:58:24 ago
>>     Originator Id: 10.100.0.1
>>     Cluster Id List:  10.0.0.42
>> dut#
>>
>> With "set nexthop self" for a given neighbor the nexthop is changed
>> unconditinally for this neighbor. How can I match only on external
>> routes?
>
> I have this, is that what you're after? (obviously skip the community
> setting if you don't want it).
>
> match from ebgp set { community $ASN:neighbor-as, nexthop self }

I was looking for a way to define it per peer. But actually it works fine
with "match from ebgp ...".

Thank you!

Reply via email to