Bird does not prefer older eBGP route - RFC5004 and "older prefer on"

Mazur, Dariusz dmazur at akamai.com
Mon Jul 3 13:46:55 CEST 2023


Hello Maria,
Thanks for response. Attributes look the same. Maybe the problem is these routes are learnt via iBGP, what is not visible in "show route  all'

Our simplified topology looks like below:

1.Host_1 and Host_2 announces 172.232.160.0/19
2.Host_1 and Host_2 have different ASNs and they use eBGP to peer with ToRs
3.All tors are connected to r01.leaf over iBGP

Host1----eBGP----r01a.tor----iBGP-----r01.leaf
           ----eBGP----r01b.tor----iBGP

Host2----eBGP----r02a.tor----iBGP-----r01.leaf
          ----eBGP----r02b.tor----iBGP

Show route all from r01.leaf

172.232.160.0/19     unicast [192.168.196.1__r01a.tor106 2023-06-30] * (100) [AS4250627481?]
                via 192.168.196.1 on vlan.201
                Type: BGP univ
                BGP.origin: Incomplete
                BGP.as_path: 4250627481
                BGP.next_hop: 192.168.196.1
                BGP.med: 0
                BGP.local_pref: 400
                BGP.atomic_aggr:
                BGP.aggregator: 23.219.179.225 AS4250627481
                BGP.community: (63949,1000) (63949,1002) (63949,1004) (63949,1005) (65110,31107) (65310,31107) (65518,31107)
                     unicast [192.168.196.3__r01b.tor106 2023-06-30] (100) [AS4250627481?]
                via 192.168.196.3 on vlan.202
                Type: BGP univ
                BGP.origin: Incomplete
                BGP.as_path: 4250627481
                BGP.next_hop: 192.168.196.3
                BGP.med: 0
                BGP.local_pref: 400
                BGP.atomic_aggr:
                BGP.aggregator: 23.219.179.225 AS4250627481
                BGP.community: (63949,1000) (63949,1002) (63949,1004) (63949,1005) (65110,31107) (65310,31107) (65518,31107)
                     unicast [192.168.196.5__r02a.tor106 2023-06-30] (100) [AS4250627482?]
                via 192.168.196.5 on vlan.203
                Type: BGP univ
                BGP.origin: Incomplete
                BGP.as_path: 4250627482
                BGP.next_hop: 192.168.196.5
                BGP.med: 0
                BGP.local_pref: 400
                BGP.atomic_aggr:
                BGP.aggregator: 23.219.179.226 AS4250627482
                BGP.community: (63949,1000) (63949,1002) (63949,1004) (63949,1005) (65110,31107) (65310,31107) (65518,31107)
                     unicast [192.168.196.7__r02b.tor106 2023-06-30] (100) [AS4250627482?
                via 192.168.196.7 on vlan.204
                Type: BGP univ
                BGP.origin: Incomplete
                BGP.as_path: 4250627482
                BGP.next_hop: 192.168.196.7
                BGP.med: 0
                BGP.local_pref: 400
                BGP.atomic_aggr:
                BGP.aggregator: 23.219.179.226 AS4250627482
                BGP.community: (63949,1000) (63949,1002) (63949,1004) (63949,1005) (65110,31107) (65310,31107) (65518,31107)

Thanks,
Dariusz

From: Maria Matejka via Bird-users <bird-users at network.cz>
Reply-To: Maria Matejka <maria.matejka at nic.cz>
Date: Friday, June 30, 2023 at 12:53 PM
To: "bird-users at network.cz" <bird-users at network.cz>
Subject: Re: Bird does not prefer older eBGP route - RFC5004 and "older prefer on"


Hello!

I suspect that the routes either aren't all external, or are otherwise compared different before it comes to breaking ties. Could you please share the `show route all` output to see all the relevant BGP attributes?

Maria
On 6/30/23 11:43, Mazur, Dariusz via Bird-users wrote:
Hello Bird users,
Have you ever use RFC 5004 and “older prefer” knob. I am trying to use it but it seems not to work:

1.Router learns the same route from different ebgp peers, it prefers route from r01a  and this route is exported to BGP peers
172.232.160.0/19     unicast [192.168.196.129__r01a.tor106 2023-06-14] * (100) [AS4250627481?]
                via 192.168.196.129 on vlan.201
                     unicast [192.168.196.131__r01b.tor106 2023-06-14] (100) [AS4250627481?]
                via 192.168.196.131 on vlan.202
                     unicast [192.168.196.133__r02a.tor106 2023-06-14] (100) [AS4250627482?]
                via 192.168.196.133 on vlan.203


2.Once we lose links to r01a  and r01b  route from r02a is preferred and exported to BGP peers. It is expected
172.232.160.0/19     unicast [192.168.196.133__r02a.tor106 2023-06-14] * (100) [AS4250627482?]
                via 192.168.196.133 on vlan.203

3.When links to r01a  and r01b. are again online route, route from r01a is pricked as primary and exported to BGP. It causes route oscillation
172.232.160.0/19     unicast [192.168.196.129__r01a.tor106 09:14:19.982] * (100) [AS4250627481?]
                via 192.168.196.129 on vlan.201
                             unicast [192.168.196.131__r01b.tor106 09:14:19.896] (100) [AS4250627481?]
                via 192.168.196.131 on vlan.202
                     unicast [192.168.196.133__r02a.tor106 2023-06-14] (100) [AS4250627482?]
                via 192.168.196.133 on vlan.203

4.I believe it is default behavior not to prefer older path. According to documentation RFC 5004 and "prefer older on" should fix my problem, but it does not work.
Bird doc says:
prefer older switch
Standard route selection algorithm breaks ties by comparing router IDs. This changes the behavior to prefer older routes (when both are external and from different peer). For details, see RFC 5004. Default: off.

5. According to documentation RFC 5004 and "prefer older on" should fix my problem, but it does not work.

a) added "prefer older on", bgp flapped and routes were re-learnt
172.232.160.0/19     unicast [192.168.196.129__r01a.tor106 09:22:12.949] * (100) [AS4250627481?]
                via 192.168.196.129 on vlan.201
                             unicast [192.168.196.131__r01b.tor106 09:22:13.527] (100)   [AS4250627481?]
                via 192.168.196.131 on vlan.202
                     unicast [192.168.196.133__r02a.tor106 09:22:12.683] (100) [AS4250627482?]
                via 192.168.196.133 on vlan.203

b) shut links to r01a and r01b
172.232.160.0/19     unicast [192.168.196.133__r02a.tor106 09:22:12.683] * (100) [AS4250627482?]
                via 192.168.196.133 on vlan.203

c) unshut links to r01a and r01b, route from r01a is again  preferred, so looks like "older" knob does not work

172.232.160.0/19     unicast [192.168.196.129__r01a.tor106 09:27:55.841] * (100) [AS4250627481?]
                via 192.168.196.129 on vlan.201
                     unicast [192.168.196.131__r01b.tor106 09:27:54.448] (100) [AS4250627481?]
                via 192.168.196.131 on vlan.202
                     unicast [192.168.196.133__r02a.tor106 09:22:12.683] (100) [AS4250627482?]. ---> this route is older
                via 192.168.196.133 on vlan.203

Thanks,
Dariusz

--

Maria Matejka (she/her) | BIRD Team Leader | CZ.NIC, z.s.p.o.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://trubka.network.cz/pipermail/bird-users/attachments/20230703/ac395bfb/attachment.htm>


More information about the Bird-users mailing list