BGP additional-path over eBGP

[saccom]

Hi all,

i have a question about the additional-path feature in BGP. I have this topology:

a PE is learning 3 different path of the same prefix from different iBGP neighbors. Now, this PE have an eBGP session to a CE. Is there any method/way to advertise to the CE ALL these 3 paths?

Thanks a lot.

[nkarpysh]

Hello,

You can consider BGP multipath:

https://www.cisco.com/c/en/us/td/docs/routers/xr12000/software/xr12k_r4-1/routing/configuration/guide/routing_cg41xr12k_chapter1.html#concept_09F04520D4D94608AE232CDBCD9281E5

Niko

[saccom]

Hello,

thanks for your reply.

I don't think that it can be a valid solution, simply because the PE should have only one best path, and so it cannot do load balancing. 

Let's think in a VPNv4 topology: all PEs have all possibile paths by default, so if there is a VPNv4 inter-as interconnection, the external PE have all the paths that are present in the source AS.

Now, i would like to do this but on an ipv4 unicast address-family. 

Is that possible?

Thanks a lor

[nkarpysh]

So I read it wrong then,

Yes, in this case "additional path selection" with route policy in IPV4 AF should work. Only restriction - you cant do it per neighbor but only per whole AF.

Niko

[saccom]

Hello Niko,

thanks again for your reply.

Let's view it better. In the lab i have two XRs (R7 and R8), with are in different ASs (R7 in as 7, and R8 is in AS8). Beyond these routers, there are other routers that "inject" the same prefix, so on R7 i have this

Route Distinguisher: 7:7 (default for vrf X)
*>i7.7.7.7/32       155.1.156.1 0 100 0 ?
* i                        155.1.156.5 0 100 0 ?
* i                        155.1.156.6 0 100 0 ?
*> 8.8.8.8/32       155.1.78.8 0 8 ?

The prefix 7.7.7.7/32 has 3 path because each iBGP router (155.1.156.1, 155.1.156.5 and 155.1.156.6) send me that prefix. Now for this prefix i have 3 path, and R7 choose his best. If we go on R8, we see that 

Status codes: s suppressed, d damped, h history, * valid, > best
i - internal, r RIB-failure, S stale, N Nexthop-discard
Origin codes: i - IGP, e - EGP, ? - incomplete
Network Next Hop Metric LocPrf Weight Path
Route Distinguisher: 8:8 (default for vrf X)
*> 7.7.7.7/32           155.1.78.7 0 7 ?

Processed 1 prefixes, 1 paths

R7 is sending to R8 only one path (his best). Ok, now i've configured a route-policy

route-policy PATH
set path-selection all advertise
end-policy
!

And i've applied this under BGP

R7

router bgp 7

vrf X
address-family ipv4 unicast
additional-paths receive
additional-paths send
additional-paths selection route-policy PATH
!
neighbor 155.1.78.8     //is R8
remote-as 8
capability additional-paths receive
capability additional-paths send
address-family ipv4 unicast
route-policy PERMIT_ALL in
route-policy PERMIT_ALL out
!

However if i check under R8, i see that  

Origin codes: i - IGP, e - EGP, ? - incomplete
Network Next Hop Metric LocPrf Weight Path
Route Distinguisher: 8:8 (default for vrf PROVA)
*> 7.7.7.7/32       155.1.78.7 0 7 ?

Processed 1 prefixes, 1 paths

Nothig has changed.

Do you have any idea? Maybe i've simply misconfigured this.

PS. on R8 i've configured capability for receiving additional path, but i've read on forums that it's not supported for eBGP peer. 

Thanks a lot again

[nkarpysh]

Tested same in my lab and I see only one route advertised as well:
Route Distinguisher: 7:7 (default for vrf test)
*>i7.7.7.7/32 3.3.3.3 0 100 0 ?
* i 4.4.4.4 0 100 0 ?

Processed 1 prefixes, 2 paths
RP/0/0/CPU0:R7#
RP/0/0/CPU0:R7#
RP/0/0/CPU0:R7#
RP/0/0/CPU0:R7#
RP/0/0/CPU0:R7#
RP/0/0/CPU0:R7#sh bgp vrf test neighbors 192.168.78.8 advertised-r
Sat Jan 20 22:33:16.642 UTC
Network Next Hop From AS Path
Route Distinguisher: 7:7 (default for vrf test)
7.7.7.7/32 192.168.78.7 3.3.3.3 7?


Tried playing by moving config under VPNV4 which is recommended in some doc - als did not help.
For eBGP we have capability which you configured.
It seems it might not be implemented under VRF for eBGP. Can you please open new Service Request to explore it in more details.

Thanks
Niko

[saccom]

Hello Niko,

thanks a lot for your time.

In a VPNv4 inter-AS environment, this type of NLRI exchange (i mean multiple path) is already done by default, simply because (if there is retain route-target all) all router that have vpnv4 neighborship should have all the path, because  VPNv4 routers doesn't do best path selection. The problem come when a router must do best path selection, like in ipv4 unicast (vrf or not) scenario. In this case, the router should pick only one path, that can advertise to his eBGP neighbor (for iBGP neighbor i think that capability send/receive should work). 

Thanks a lot again Niko, i really appreciate.