ebgp_unn_vxlan

SR Linux BGP IPv6 Unnumbered VXLAN DC Fabric running in Containerlab

Topology

• Leaf, Spines, Super/Spines and Internet/router are SR Linux nodes (24.7.x software version)
• Core switches are Cumulus nodes
• Clients are simulated with FRR docker images: connected to Leaf and Core3 switches
• Uses Netmiko to configure SR Linux switches

All BGP sessions are established through IPv6 link-local addresses except the overlay EVPN:

• FRRClient <-> leaf: IPv4 and IPv6 families.
• Underlay Leaf <-> Spine: IPv4 and IPv6 families.
• Underlay Spine <-> SuperSpine: EVPN, IPv4 and IPv6 families.
• Underlay SuperSpine <-> Core: EVPN, IPv4 and IPv6 families.
• Overlay between Leaf/Spine/SuperSpine and Core: EBGP Multihop EVPN
• Core <-> Core: EVPN, IPv4 and IPv6 families.
• Core <-> InternetRouter: IPv4 and IPv6 families.

VXLAN VTEP termination in Leaf (SR Linux) and Core3 (Cumulus) switches.

Requirements

• Containerlab
• Docker
• SR Linux Container image
• Cumulus Container image
• FRR
• Netmiko

Deploying the lab

# clone this repository
git clone https://github.com/michelredondo/ebgp_unn_vxlan && cd ebgp_unn_vxlan

# deploy containerlab topology
clab deploy

# Underlay SR Linux switches configuration
python create_underlay.py

# Overlay/Client Services SR Linux switches configuration
python create_services.py

Client Connectivity use cases

Multihomed routed port in GRT

FRR client is connected to a leaf-pair by using two routed interfaces. Two BGP unnumbered sessions (one per interface) are used to exchange v4/v6 prefixes.

Leaf switches announce default v4/v6 route. FRR clients announces local network. Connection is terminated in GTR (network-instance default).

Multihomed bridged port in GRT

FRR client is connected to a leaf-pair by using two bridged interfaces (using IRB). Two BGP unnumbered sessions (one per interface) are used to exchange v4/v6 prefixes.

Leaf switches announce default v4/v6 route. FRR clients announces local network Connection is terminated in GTR (network-instance default).

Lag port in GRT

FRR client is connected to a leaf-pair by using a bridged lag interface (using IRB). v4/v6 subnet is directly connected to Leaf-pair. Lag interface uses Multihoming all-active EVPN Ethernet Segment procedures.

Leaf switches announce directly connected network. FRR clients uses static routing. Connection is terminated in GTR (network-instance default).

Lag port in mac-vrf (Layer 2 EVPN)

FRR client is connected to a leaf-pair by using a bridged lag interface. L2 traffic is terminated in mac-vrf MAC3 and transported to inf-core3 switch by using VXLAN. An unnumbered BGP session is established between client connected to leaf switches and client connected to inf-core3 switch (Cumulus).

Lag interface uses Multihoming all-active EVPN Ethernet Segment procedures.

Lag port in mac-vrf&ip-vrf (Layer 3 EVPN Anycast gateways)

FRR client is connected to a leaf-pair by using a bridged lag interface (using IRB). Traffic is terminated in mac-vrf MAC4 and connected to ip-vrf1 by using IRB interface. All IRB interfaces facing the clients have the same IP address and MAC (Anycast GW configuration).

Lag interface uses Multihoming all-active EVPN Ethernet Segment procedures.

Summary of services