Ethernet Virtual Private Network (evpn) Configuration Module

This configuration module configures the BGP EVPN address family to implement L2VPN or L3VPN. It supports:

VXLAN-based transport over IPv4 and IPv6
MPLS-based transport
VLAN-Based Service (bridging of a single VLAN within an EVPN Instance)
VLAN-Aware Bundle Service (bridging of multiple related VLANs inside a single EVPN Instance)
Symmetric and asymmetric IRB
Layer-3-only VPN (L3VPN) with symmetric IRB
Most EVPN design scenarios, including the convoluted ones like IBGP-over-EBGP or EBGP-over-EBGP.

Platform Support

The following table describes per-platform support of individual EVPN/VXLAN features:

Operating system	VLAN-based service	VLAN Bundle service	Asymmetric IRB	Symmetric IRB
Arista EOS	✅	✅	✅	✅
Aruba AOS-CX	✅	❌	✅	✅❗
Cisco Nexus OS	✅	❌	✅	✅
Cumulus Linux	✅	❌	✅	✅
Dell OS 10	✅	❌	✅	✅
FRR	✅	❌	✅	✅
Nokia SR Linux	✅	✅	✅	✅
Nokia SR OS	✅	❌	✅	✅
VyOS	✅	❌	✅	✅

The following table describes per-platform support of individual EVPN/MPLS features:

Operating system	VLAN-based service	VLAN Bundle service	Asymmetric IRB	Symmetric IRB
Arista EOS	✅	✅	❌	❌

Note

Arista EOS requires anycast gateway for EVPN/MPLS symmetric IRB configuration.

Devices supporting EVPN VLAN bundle services implement the following bundle service types (see RFC 7432 section 6 for more details):

Operating system	VLAN bundle	Port service	VLAN-aware bundle	Port-based VLAN bundle
Arista EOS	❌	❌	✅	❌
Nokia SR Linux	❌	❌	✅	❌

EVPN module supports IBGP- and EBGP-based EVPN:

Operating system	IBGP+IGP	BGP RR	EBGP-only	EBGP Unnumbered
Arista EOS	✅	✅	✅	✅
Aruba AOS-CX	✅	✅	❌	❌
Cisco Nexus OS	✅	✅	✅	❌
Cumulus Linux	✅	✅	✅	✅
Dell OS 10 ❗	✅	✅	✅	✅
FRR	✅	✅	✅	✅
Nokia SR Linux	✅	✅	❌	❌
Nokia SR OS	✅	✅	✅	✅
VyOS	✅	✅	❌	❌

With additional nerd knobs (more details), it’s possible to implement the more convoluted designs, including:

IBGP EVPN AF session established between loopback interfaces advertised with underlay EBGP IPv4 AF
EBGP EVPN AF session established between loopback interfaces advertised with underlay EBGP IPv4 AF (requires ebgp.multihop plugin)

Operating system	IBGP over EBGP	EBGP over EBGP
Arista EOS	✅	✅
Aruba AOS-CX	✅	❌
Cisco Nexus OS	❌	❌
Cumulus Linux	✅	✅
Dell OS 10	✅	❌
FRR	✅	✅
Nokia SR Linux	✅	❌
Nokia SR OS	✅	❌
VyOS	✅	❌

Most EVPN/VXLAN implementations support only IPv4 VXLAN transport; some can run VXLAN-over-IPv6:

Operating system	IPv4 VXLAN transport	IPv6 VXLAN transport
Arista EOS	✅	❌
Aruba AOS-CX	✅	❌
Cisco Nexus OS	✅	❌
Cumulus Linux	✅	❌
Dell OS 10	✅	❌
FRR	✅	✅
Nokia SR Linux	✅	❌
Nokia SR OS	✅	❌
VyOS	✅	❌ ❗

Global EVPN Parameters

EVPN module supports these default/global/node parameters:

evpn.vrfs (global or node parameter): A list of EVPN-enabled VRFs. The default value with VXLAN transport: all global VRFs with evpn.transit_vni parameter. There is no default value with MPLS transport.
evpn.vlans (global or node parameter): A list of EVPN-enabled VLANs. The default value with VXLAN transport: all global VLANs with vni parameter. There is no default value with MPLS transport.
evpn.session (global or node parameter): A list of BGP session types on which the EVPN address family is enabled (default: ibgp)
evpn.as (global parameter): Autonomous system number for VLAN and VRF route targets. Default value: bgp.as (when set globally) or vrf.as.
evpn.start_transit_vni (system default parameter) – the first symmetric IRB transit VNI, range 4096…16777215
evpn.start_transit_vlan (device-dependent node parameter) – the starting VLAN ID for VLANs used to map VXLAN transit VNIs

VLAN-Based Service Parameters

EVPN-related VLAN parameters are set on the vlans dictionary. You can set the following parameters for every VLAN using VLAN-Based Service:

evpn.evi: EVPN Instance identifier.
evpn.rd: EVPN Instance route distinguisher. Default: bgp.router_id:evpn.evi
evpn.import and evpn.export: Import and export route targets (not currently checked).

EVPN configuration module sets the following default EVI/RD/RT values for EVPN-enabled VLANs that are not part of a bundle service:

evpn.evi: vlan-id
evpn.rd: router-id:evi (according to Section 7.9 of RFC 7432 as the evpn.evi is set to vlan.id)
evpn.import and evpn.export: as:vlan-id (according to Section 7.10 of RFC 7432 and Section 5.1.2.1 of RFC 8365)[1]

EVPN Bundle Services

VLAN-Aware Bundle Service uses VRF configuration (and thus requires VRF configuration module). All EVPN-enabled VLANs belonging to a single VRF are configured as a bundle service, modeled as a single EVPN Instance.

RD and RT values assigned by VRF module are used to configure the VLAN bundle; you can set evpn.evi VRF parameter to set the EVPN Instance identifier. The default value of the VRF EVPN Instance identifier is the vrf.id.

The EVPN bundle service is enabled with evpn.bundle VRF parameter that can take one of the following values:

vlan – VLAN bundle service (RFC 7432 section 6.2)
port – Port-based service (RFC 7432 section 6.2.1)
vlan_aware – VLAN-aware bundle service (RFC 7432 section 6.3)
port_aware – Port-based VLAN-aware service (RFC 7432 section 6.3.1)

Integrated Routing and Bridging

IRB is configured whenever EVPN-enabled VLANs in a VRF contain IPv4 or IPv6 addresses:

Asymmetric IRB requires no extra parameters (see Asymmetric IRB section for more details)
Symmetric IRB used with VXLAN transport needs a transit VNI that has to be set with the evpn.transit_vni parameter.
You can set the VRF EVI value with evpn.evi parameter.

The evpn.transit_vni parameter must specify a globally unique VNI value. It could be set to:

True: EVPN configuration module auto-assigns a unique VNI to the VRF.
An integer value: static VNI assignment, checked for uniqueness
Name of another VRF: the evpn.transit_vni value is copied from that VRF. Use this setting for complex topologies where VRFs with different connectivity requirements have to share the transit VXLAN segment.

Asymmetric IRB

Asymmetric IRB is a forwarding paradigm in which the ingress PE device performs routing between source and destination VLAN followed by EVPN bridging. The egress PE device bridges between EPVN transport (VXLAN or MPLS pseudowire) and destination VLAN.

To make asymmetric IRB work, all EVPN-enabled VLANs participating in a routing domain must be present on all participating PE devices. The EVPN configuration module strictly enforces that requirement; every EVPN-enabled VLAN belonging to a VRF that uses asymmetric IRB must be present on every node on which the parent VRF is defined.

While you could define VLANs with a vlans attribute on every participating device, it’s much easier to meet those requirements with a group of PE devices, listing the VLANs participating in an asymmetric IRB routing domain in the group vlans attribute:

groups:
  hosts:
    members: [ h1, h2 ]
    device: linux
  switches:
    members: [ s1,s2 ]
    module: [ vlan,vxlan,ospf,bgp,evpn,vrf ]
    bgp.as: 65000
    vrfs:
      tenant:
        ospf: False
    vlans:
      red:
        vrf: tenant
      blue:
        vrf: tenant

Tip

Disable OSPF in a VRF using asymmetric IRB unless you connected external router(s) to one of the participating VLANs

Convoluted EVPN Designs

Implementing mainstream EVPN designs (IBGP+IGP, EBGP) with netlab topologies is trivial; more convoluted designs require additional BGP features or plugins:

IBGP-over-EBGP design assigns the same BGP AS number to all nodes (triggering IBGP EVPN sessions) and uses bgp.local_as on links to force EBGP sessions between adjacent nodes. Furthermore, you must limit the activation of IPv4 address family to EBGP sessions, and turn off IBGP-without-IGP warning (sample topology)
EBGP-over-EBGP design uses ebgp.multihop plugin to establish additional EBGP sessions between device loopbacks. You have to activate EVPN AF on multihop EBGP sessions and limit IPv4 AF to direct EBGP sessions (sample topology).