16 – SONiC BGP Unnumbered Configuration

In this article we will explore how to configure BGP using unnumbered interfaces on Enterprise SONiC. We will cover:

BGP unnumbered overview
Enabling BGP on unnumbered interfaces
Configuring neighbors with interface references
Verification

BGP Unnumbered Overview

BGP unnumbered allows routers to establish BGP sessions without assigning an IPv4 address to each point-to-point interface. Instead, it relies on IPv6 link-local addresses and the extended next-hop encoding defined in RFC 5549, which enables IPv4 routes to be advertised with an IPv6 next hop. When IPv6 is enabled on an interface, a link-local address is automatically created and exchanged using IPv6 router advertisements (RAs). BGP peers discover each other by parsing these RAs and use the learned link-local addresses to form BGP sessions over TCP. The primary benefit of this approach is operational simplicity: it eliminates the need to allocate and manage IPv4 addresses on every inter-router link, reduces configuration overhead, and prevents address exhaustion. In large enterprise and data-center networks built on SONiC, BGP unnumbered enables cleaner, more scalable routing designs while maintaining full IPv4 reachability.

Topology

We will use the same topology used in Article 15 – SONiC BGP configuration and we will use BGP unnumbered for eBGP configuration between Switch-3 and Switch-5.

Configuration

We will remove the IP interfaces from the following interface:

Switch-3 / Ethernet 2
Switch-5 / Ethernet 2

Switch 3

Switch-3# show running-configuration interface Ethernet 2
!
interface Ethernet2
 mtu 9100
 speed 25000
 unreliable-los auto
 no shutdown
 ip address 192.168.3.1/30


Switch-3# configure terminal 
Switch-3(config)# interface Ethernet2
Switch-3(config-if-Ethernet2)# no ip address 
Switch-3(config-if-Ethernet2)# ipv6 enable 
Switch-3(config-if-Ethernet2)#

Switch 5

Switch-5# show running-configuration interface Ethernet 2
!
interface Ethernet2
 mtu 9100
 speed 25000
 unreliable-los auto
 no shutdown
 ip address 192.168.3.2/30


Switch-5# configure terminal
Switch-5(config)# interface Ethernet2
Switch-5(config-if-Ethernet2)# no ip address
Switch-5(config-if-Ethernet2)# ipv6 enable 
Switch-5(config-if-Ethernet2)#

We will change the router BGP configuration to use BPG unnumbered.

Switch 3 – Before

router bgp 200
 router-id 3.3.3.3
 log-neighbor-changes
 timers 60 180
 !
 neighbor 1.1.1.1
  remote-as 200
  update-source interface Loopback 0
  !
  address-family ipv4 unicast
   activate
 !
 neighbor 2.2.2.2
  remote-as 200
  update-source interface Loopback 0
  !
  address-family ipv4 unicast
   activate
 !
 neighbor 4.4.4.4
  remote-as 200
  update-source interface Loopback 0
  !
  address-family ipv4 unicast
   activate
 !
 neighbor 192.168.3.2
  remote-as 100
  !
  address-family ipv4 unicast
   activate

Switch 3 – After

router bgp 200
 router-id 3.3.3.3
 log-neighbor-changes
 timers 60 180
 !
 neighbor 1.1.1.1
  remote-as 200
  update-source interface Loopback 0
  !
  address-family ipv4 unicast
   activate
 !
 neighbor 2.2.2.2
  remote-as 200
  update-source interface Loopback 0
  !
  address-family ipv4 unicast
   activate
 !
 neighbor 4.4.4.4
  remote-as 200
  update-source interface Loopback 0
  !
  address-family ipv4 unicast
   activate
 !
 neighbor interface Ethernet2
  remote-as 100
  capability extended-nexthop
  !
  address-family ipv4 unicast
   activate

We apply similar configuration on Switch-5.

Verification

Use the command show bgp ipv4 unicast summary to verify BGP peers.

Switch-5

Switch-5# show bgp ipv4 unicast summary 
BGP router identifier 5.5.5.5, local AS number 100 VRF default
Neighbor      V   AS         MsgRcvd   MsgSent   InQ     OutQ    Up/Down         State/PfxRcd   
Ethernet2     4   200        9         11        0       0       00:02:54        4              
 
Total number of neighbors 1
Total number of neighbors established 1

Notice that we have BGP neighbor on interface Ethernet 2 in AS 200.

Use the command show ip routeto verify the routing table.

Switch-5# show ip route 
Codes:  K - kernel route, C - connected, S - static, B - BGP, O - OSPF, A - attached-host
        > - selected route, * - FIB route, q - queued route, r - rejected route, b - backup
       Destination        Gateway                                                                    Dist/Metric   Last Update 
--------------------------------------------------------------------------------------------------------------------------------      
 B>*   10.0.0.0/28        via fe80::e2d:4aff:fed0:a       Ethernet2                                  20/0          00:05:45 ago
 B>*   172.16.1.0/28      via fe80::e2d:4aff:fed0:a       Ethernet2                                  20/0          00:05:45 ago
 B>*   192.168.1.8/30     via fe80::e2d:4aff:fed0:a       Ethernet2                                  20/0          00:05:45 ago
 B>*   192.168.1.16/30    via fe80::e2d:4aff:fed0:a       Ethernet2                                  20/0          00:05:45 ago
 C>*   192.168.3.8/30     Direct                          Ethernet5                                  0/0           00:14:59 ago

Notice that the BGP-learned routes use IPv6 link-local addresses as next hops (fe80::/10), even though the routes themselves are IPv4. This is a direct result of BGP unnumbered with RFC 5549 extended next-hop encoding, where IPv4 prefixes are resolved over IPv6 link-local next hops on the point-to-point interface.

Notes

IPv6 is required – BGP unnumbered relies on IPv6 link-local addresses and router advertisements, even when carrying only IPv4 routes.
Use only on point-to-point links – Shared or multi-access segments can cause ambiguous neighbor discovery and should be avoided.
Ensure RFC 5549 support end-to-end – All BGP peers must support extended next-hop encoding for IPv4 over IPv6 to work correctly.
Expect different troubleshooting workflows – There are no IPv4 interface addresses, so debugging relies on IPv6 link-local and BGP-specific tools.