[j-nsp] BGP full mesh or route reflector

[nsp at rhanssen.de]

Hi,

but please take into consideration that your iBGP Sessions will flap once you configure the first rr session on a box if you use vpn families:
https://www.juniper.net/documentation/us/en/software/junos/bgp/topics/topic-map/bgp-session-flaps.html

btw:
We run 14 routers (MX204-960, peak was 25 devices) with a full table each in a full meshed setup (IS-IS/RSVP/MPLS, no SR).
Works fine since many years and I remember no issue that a rr would have avoided.
On the other side we already had issues (long time ago in an other setup with quagga rr and Catalyst 6500 + x86 routers) with a rr setup in which router A was not able to reach router B but both reached the RR.
So I do not think that a full mesh is a no-go for small setups.

kind regards
Rolf

On 05/12/2025 23:20, Aaron1 via juniper-nsp wrote:
> Migrating to an RR design later would seem to add RR, full mesh to it, then one by one start moving routers to RR-client connections, til complete
> 
> Aaron
> 
>> On Dec 5, 2025, at 4:07 PM, Johan Borch via juniper-nsp <juniper-nsp at puck.nether.net> wrote:
>>
>> Thanks, this is a quite small network at the moment. Three edge routers
>> (with full tables) and a bunch (9) PE routers (these will be able to handle
>> full table). I guess my only option right now is to run RR on my three edge
>> routers, not sure if that is a good idea.
>>
>> A bunch of virtual RRs sound like a good solution. But we can't add more
>> cost at the moment, is it hard to migrate towards a RR design at a later
>> stage?
>>
>> Johan
>>
>>> On Fri, Dec 5, 2025 at 3:39 PM Saku Ytti <saku at ytti.fi> wrote:
>>>
>>> I would generally recommend RR on anything more than 2 router setup.
>>>
>>> RR gives redundancy on the signalling path, one iBGP flap doesn't
>>> cause an outage.
>>>
>>> With ORR and ADDPATH you're not really losing anything.
>>>
>>>> On Fri, 5 Dec 2025 at 14:36, Johan Borch via juniper-nsp
>>>> <juniper-nsp at puck.nether.net> wrote:
>>>>
>>>> Hi!
>>>>
>>>> In an SR/MP-BGP underlay, will it have a significant impact on device
>>>> performance if we use a full iBGP mesh instead of route reflectors or
>>> other
>>>> drawbacks? Let’s say we will end up with around 100 PE routers. These
>>>> routers will not carry an excessive number of prefixes (no full tables).
>>>> We can ignore the configuration part as configuration is auto-generated.
>>>>
>>>> Br
>>>> Johan
>>>> _______________________________________________
>>>> juniper-nsp mailing list juniper-nsp at puck.nether.net
>>>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>>>
>>>
>>>
>>> --
>>>   ++ytti
>>>
>> _______________________________________________
>> juniper-nsp mailing list juniper-nsp at puck.nether.net
>> https://puck.nether.net/mailman/listinfo/juniper-nsp
> 
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp