BGP Auto-peering

Auto-peering is a network of cooperating interconnected devices that create an AutoBGP for any topology, providing fully redundant, multipath routing. The fabric grows dynamically and freely, not bound to any well-known topology such as Clos or Leaf/Spine.

Auto-peering nodes build a secure network by running the very scalable Border Gateway Protocol (BGP) to exchange topology and host information about IP networks. It uses IPv6 as the network layer to transport IPv4 and IPv6 traffic.

Any device connecting to an auto-peering device is an attachment point to the network. This network provides the underlay for services such as VXLAN, policy, VRF, and Fabric Attach. The AutoBGP device applies policy rules as it discovers external devices. These devices can be any IP host, LAG-attached servers and bridges, or gateway routers. By default, auto-peering allows connectivity for all attached hosts, allowing for a controller-less operation. However, interconnection (trunk ports) between fabric nodes should not be LAG ports; you should not enable port sharing on trunk ports.

BGP auto-peering includes the following features:

EBGP:
- Facilitates IP host routing on Default VR
- Single command (no IP address assignments required for interlinks)
- IPv6 link locals for interlink addresses
- LLDP for discovery (proprietary, but RFC is in draft to be standardized)
- BGP peering on link-local addresses
- Automatic EVPN peering
Route redistribution on Default VR:
- Static routes
- OSPFv2/v3
Routing in-and-out of tunnels (RIOT) with redundant attachments (see Virtual Extensible LAN (VXLAN) Tunnel Improvements)
- MLAG (active/active)
- EasyLAG (active/standby)—default when no MLAG is configured
Multicast (PIM-DM)

Supported Platforms

Summit X670-G2, X770, and the ExtremeSwitching X690, X870 series switches.

This feature requires the Advanced Edge license. For more information about licenses, see the ExtremeXOS 22.5 Feature License Requirements.

Limitations

The following features are not supported in BGP auto-peering:

Stacking
AutoBGP LAG with Extended Edge Switching
Static LAG attachments to AutoBGP
AutoBGP links on Extended Edge Switching ports
MPLS, VPLS, L2VPN, L3VPN
PIM Snooping, PIM-SM, SSM mode
OSPF, OSPFv3, ISIS, RIP, RIPNG per VRF
IPv6 within VRF
VRF route leaking
EVPN multi-homing Ethernet segments
Explicit-remotes mode used for VNET flooding
VXLAN symmetric routing
AutoBGP MLAG
EVPN does not interoperate with third-party devices

The following limitations apply:

Manual configuration of MLAG is supported. One MLAG peer per leaf node.
Loss of first multicast packet in the flow is expected due to slow path forwarding.
Static router must be an external router per VRF.
VLANS spanning multiple bridges, where each bridge is AutoBGP LAG connected, must be VXLAN-based, or AutoBGP LAG replaced with MLAG.
VLANs behind a bridge that is AutoBGP LAG connected and AutoBGP node has same VLAN ports must be VXLAN-based or AutoBGP LAG replaced with MLAG.

New CLI Commands

create auto-peering bgp vlans vlan_list routerid ipaddress AS-number asNumber

delete auto-peering

show auto-peering

Changed CLI Commands

Changes are underlined.

show iproute ipv6 origin [auto-peering direct | static | blackhole | ripng | ospfv3 | ospfv3-intra | ospv3-inter | ospfv3-extern1 | ospfv3-extern2 | isis | isis-leve1-1 | isis-level-2 | isis-level-1-external | isis-level-2-external | bgp | ibgp | ebgp | bootp | host-mobility] {vr vr_name}

The following show command was changed to show auto-peering information:

show iproute {ipv4} {priority | vlan vlan_name | permanent | ip_address netmask | summary} {multicast | unicast} {vr vrname}}

Published June 2019prev | next

Email this topic

Print this page Print this page

Leave feedback Feedback