Implement BGP EVPN Protected Overlay Segmentation on Catalyst 9000 Series Switches
Available Languages
Contents
Introduction
This document describes how to implement BGP EVPN VXLAN Protected Overlay Segmentation on Catalyst 9000 Series Switches.
Prerequisites
Requirements
Cisco recommends that you have knowledge of these topics:
- BGP EVPN VxLAN concepts
- BGP EVPN Unicast Troubleshooting
- BGP EVPN VxLAN routing policy
Components Used
The information in this document is based on these software and hardware versions:
- Catalyst 9300
- Catalyst 9400
- Catalyst 9500
- Catalyst 9600
- Cisco IOS® XE 17.12.1 and later versions
The information in this document was created from the devices in a specific lab environment. All of the devices used in this document started with a cleared (default) configuration. If your network is live, ensure that you understand the potential impact of any command.
Background Information
High Level Feauture Description
The protected segment feature is a security measure that prevents ports from forwarding traffic to each other, even if they are on the same VLAN and same switch
- This feature is similar to 'switchport protected' or private Vlans, but for EVPN fabrics.
- This design forces all traffic to the CGW where it can be inspected by a Firewall before being sent to its final destination.
- Traffic flows are controlled, deterministic, and easy to inspect using a centralized security appliance.
Document Details
This document is part 2 or 3 inter-related documents:
- Document 1: Implement BGP EVPN Routing Policy on Catalyst 9000 Series Switches covers how to control the BGP BUM traffic in the Overlay, and must be configured first
- Document 2: This document. Building upon the Overlay design and policy of document 1, this document describes the implementation of the 'protected' keyword
- Document 3: Coming Soon!
Caution: You must implement the configuration in document 1 prior to implementing protected segment configurations.
Protected Segment Types
Totally isolated
- Allows only North to South communication, and
- The gateway is advertised into the fabric with the 'default-gateway advertise' CLI
Mostly Isolated
- Allows North to South communication (in this use case East / West traffic flows are allowed based on firewall traffic policies)
- Allows East to West communication (based on firewall traffic policies)
- The gateway is external to the fabric & the SVI is not advertised using the 'default-gateway advertise' CLI
Switch Behavior
- Hosts cannot communicate with each other directly even if they are connected to the same switch (ARP request not sent to other ports on same switch when hosts are in the same VRF/Vlan/Segment)
- No BUM traffic between L2 VTEPs (IMET prefixes filtered using the routing policy configuration)
- All packets from the hosts are relayed to Border Leaf to be forwarded. (This means for Host 1 to communicate to host 2 on same leaf, traffic is hair pinned up to the CGW)
Route Type 2 Handling
- Access Leafs advertise local RT2 with E-Tree Extended Community and Leaf flag set.
- Access Leafs do not install any remote RT2 received with E-Tree Extended Community and Leaf flag set in data plane.
- Access Leafs do not install each others RT2 in data plane.
- Access Leafs and Border Leaf (CGW) install each others RT2 in data plane.
- No configuration change required on Access Leaf or Border Leaf.
Design Summary
- For broadcast (BUM) the RT3 topology is hub and spoke in order to force broadcast traffic such as ARP up to the GCW.
- To account for host mobility the RT2 are full mesh at the BGP control plane (when a host moves from one VTEP to another the Seq number is incremented in the RT2)
- The data plane selectively installs MAC addresses.
- A leaf installs only local MACs & RT2 which contain the DEF GW attribute
- The CGW does not have the protected KW and installs all local MAC & remote RT2 in its data plane.
Terminology
|
VRF |
Virtual Routing Forwarding |
Defines a layer 3 routing domain that be separated from other VRF and global IPv4/IPv6 routing domain |
|
AF |
Address Family |
Defines which type prefixes and routing info BGP handles |
|
AS |
Autonomous System |
A set of Internet routable IP prefixes that belong to a network or a collection of networks that are all managed, controlled and supervised by a single entity or organization |
|
EVPN |
Ethernet Virtual Private Network |
Extension that allows BGP to transport Layer 2 MAC and Layer 3 IP information is EVPN and uses Multi-Protocol Border Gateway Protocol (MP-BGP) as the protocol to distribute reachability information that pertains to the VXLAN overlay network. |
|
VXLAN |
Virtual Extensible LAN (Local Area Network) |
VXLAN is designed to overcome the inherent limitations of VLANs and STP. It is a proposed IETF standard [RFC 7348] to provide the same Ethernet Layer 2 network services as VLANs do, but with greater flexibility. Functionally, it is a MAC-in-UDP encapsulation protocol that runs as a virtual overlay on a Layer 3 underlay network. |
|
CGW |
Centralized Gateway |
And implementation of EVPN where the gateway SVI are not on each leaf. Instead, all routing is done by a specific leaf using asymmetric IRB (Integrated Routing and Bridging) |
|
DEF GW |
Default Gateway |
A BGP extended community attribute added to the MAC/IP prefix via the command "default-gateway advertise enable" under the 'l2vpn evpn' configuration section. |
|
IMET (RT3) |
Inclusive Multicast Ethernet Tag (Route) |
Also called BGP type-3 route. This route type is used in EVPN to deliver BUM (broadcast / unknown unicast / multicast) traffic between VTEPs. |
|
RT2 |
Route Type 2 |
BGP MAC or MAC/IP prefix that represents a host MAC or Gateway MAC-IP |
|
EVPN Mgr |
EVPN Manager |
Central management component for various other components (example: learns from SISF and signals to L2RIB) |
|
SISF |
Switch Integrated Security Feature |
An agnostic host tracking table that is used by EVPN to learn what local hosts are present on a Leaf |
|
L2RIB |
Layer 2 Routing Information Base |
In intermediate component for managing interactions between BGP, EVPN Mgr, L2FIB |
|
FED |
Forwarding Engine Driver |
Programs the ASIC (hardware) layer |
|
MATM |
Mac Address Table Manager |
IOS MATM: software table which installs only local addresses and FED MATM: hardware table which installs local and remote addresses learned from control plane, and is part of the hardware forwarding plane |
Flow Diagrams
Route-Type 2 (RT2) Diagram
This diagram shows the full mesh design of the type 2 MAC/MAC-IP host prefixes.
Note: Full mesh is required to support mobility and roaming
Route-Type 3 (RT3) Diagram
This diagram shows the hub and spoke design of the broadcast IMET (RT3) tunnels
Note: Hub and spoke broadcast is required to prevent leafs with the same segment from sending broadcast to each other directly.
Address Resolution (ARP) Diagram
This diagram demonstrates that ARP is not allowed to reach any host in the same EPVN segment. When host ARPs for another host, only the CGW gets this ARP and replies
Note: This ARP behavior change is instantiated by the use of the 'protected' keyword.
Example: member evpn-instance 202 vni 20201 protected
Configure (Totally Isolated)
Network Diagram
Protected configuration keyword is applied on the Leaf switches. The CGW is a promiscuous device and installs all mac addresses.
Note: The routing policy community list & route-map configuration which controls the import/export of IMET prefixes is shown in Implement BGP EVPN Routing Policy on Catalyst 9000 Series Switches. Only protected segment differences are shown in this document.
Leaf-01 (Base EVPN Config)
Leaf-01#show run | sec l2vpn
l2vpn evpn
replication-type static
flooding-suppression address-resolution disable <-- Disables ARP caching so ARP is always sent up to the CGW
router-id Loopback1
l2vpn evpn instance 201 vlan-based encapsulation vxlan replication-type ingress <-- Sets segment to use Unicast replication of BUM traffic via RT3 type BGP routes multicast advertise enable
Leaf01#show run | sec vlan config vlan configuration 201 member evpn-instance 201 vni 20101 protected <-- protected keyword added
CGW (Base Config)
CGW#show running-config | beg l2vpn evpn instance 201 l2vpn evpn instance 201 vlan-based encapsulation vxlan replication-type ingress default-gateway advertise enable <-- adds the BGP attribute EVPN DEF GW:0:0 to the MAC/IP prefix multicast advertise enable
CGW#show running-config | sec vlan config vlan configuration 201
member evpn-instance 201 vni 20101
CGW#show run int nve 1 Building configuration... Current configuration : 313 bytes ! interface nve1 no ip address source-interface Loopback1 host-reachability protocol bgp
member vni 20101 ingress-replication local-routing <-- 'ingress-replication' (Unicast all BUM traffic) / 'local-routing' (Enables vxlan centralized gateway forwarding)
CGW#show run interface vlan 201 Building configuration... Current configuration : 231 bytes ! interface Vlan201 mac-address 0000.beef.cafe <-- MAC is static in this example for viewing simplicity. This is not required vrf forwarding red <-- SVI is in VRF red ip address 10.1.201.1 255.255.255.0 no ip redirects ip local-proxy-arp <-- Sets CGW to Proxy reply even for local subnet ARP requests ip pim sparse-mode ip route-cache same-interface <-- This is auto added when local-proxy-arp is configured. However, this is a legacy 'fast switching' command that is not used by CEF & is not required for forwarding ip igmp version 3 no autostate
Note: At the CGW there is no BGP policy applied. The CGW is allowed to receive and send all prefix types (RT2, RT5 / RT3).
Verify (Totally Isolated)
EVI Details
Leaf01#sh l2vpn evpn evi 201 detail
EVPN instance: 201 (VLAN Based)
RD: 172.16.254.3:201 (auto)
Import-RTs: 65001:201
Export-RTs: 65001:201
Per-EVI Label: none
State: Established
Replication Type: Ingress
Encapsulation: vxlan
IP Local Learn: Enabled (global)
Adv. Def. Gateway: Disabled (global)
Re-originate RT5: Disabled
Adv. Multicast: Enabled
AR Flood Suppress: Disabled (global)
Vlan: 201
Protected: True (local access p2p blocked) <-- Vlan 201 is in protected mode
<...snip...>
Local RT2 Generation (Local Host to RT2)
Verify the component dependency chain from local host learning to RT2 generation:
- SISF (While the Leaf does not have an SVI, SISF still gleans the host info via ARP frame from the host)
- EVPN Mgr
- L2RIB
- BGP
Tip: If a previous component is not properly programmed the whole dependency chain breaks (example: SISF does not have en entry then BGP cannot create an RT2).
SISF
Verify SISF has the host learned in DB (Host info learned from DHCP or ARP)
- SISF learns MAC entries from IOS-MATM learning then sends up to EVPN Mgr (must be MAC-REACHABLE with policy "evpn-sisf-policy").
- SISF gleans an IP/MAC binding on a local VTEP and using EVPN manager that information is expected to be programmed as a /32 route via BGP to other leafs.
Note: In this scenario the host has a static IP, so SISF uses ARP to glean the host details. In the Mostly Isolated section DHCP and DHCP snooping is shown.
Leaf01#show device-tracking database vlanid 201
vlanDB has 1 entries for vlan 201, 1 dynamic
Codes: L - Local, S - Static, ND - Neighbor Discovery, ARP - Address Resolution Protocol, DH4 - IPv4 DHCP, DH6 - IPv6 DHCP, PKT - Other Packet, API - API created
Preflevel flags (prlvl):
0001:MAC and LLA match 0002:Orig trunk 0004:Orig access
0008:Orig trusted trunk 0010:Orig trusted access 0020:DHCP assigned
0040:Cga authenticated 0080:Cert authenticated 0100:Statically assigned
Network Layer Address Link Layer Address Interface vlan prlvl age state Time left
ARP 10.1.201.10 0006.f601.cd43 Gi1/0/1 201 0005 3mn REACHABLE 86 s <-- Gleaned from local host ARP Request EVPN Manager
EVPN Mgr learns Local MAC and installs into L2RIB. EVPN Mgr also learns the Remote MAC from L2RIB, but entry is used only for processing MAC mobility
Confirm EVPN Mgr is updated with the SISF entry
Leaf01#show l2vpn evpn mac evi 201 MAC Address EVI VLAN ESI Ether Tag Next Hop(s) -------------- ----- ----- ------------------------ ---------- --------------- 0006.f601.cd43 201 201 0000.0000.0000.0000.0000 0 Gi1/0/1:201 <-- MAC in VLan 201 local interface Gi1/0/1:service instance 201 <...snip...>
L2RIB
- L2RIB learns local MAC from EVPN Mgr and sends to BGP and L2FIB.
- L2RIB is also responsible for learning remote MACs from BGP to update EVPN Mgr and L2FIB.
- L2RIB needs both Local and remote for other components to be properly updated.
- L2RIB component sits between local and remote MAC learning depending on which direction / component needs to be updated
Verify L2RIB is updated with the local MAC from EVPN Mgr
Leaf01#show l2route evpn mac topology 201 <-- View the overall topology for this segment EVI ETag Prod Mac Address Next Hop(s) Seq Number ----- ---------- ----- -------------- ---------------------------------------------------- ---------- 201 0 BGP 0000.beef.cafe V:20101 172.16.254.6 0 <-- produced by BGP who updated L2RIB (remote learn) 201 0 L2VPN 0006.f601.cd43 Gi1/0/1:201 0 <-- produced by EVPN Mgr who updated L2RIB (local learn) Leaf01#show l2route evpn mac mac-address 0006.f601.cd43 detail EVPN Instance: 201 Ethernet Tag: 0 Producer Name: L2VPN <-- Produced by local MAC Address: 0006.f601.cd43 <-- Host MAC Address Num of MAC IP Route(s): 1 Sequence Number: 0 ESI: 0000.0000.0000.0000.0000 Flags: B() Next Hop(s): Gi1/0/1:201 (E-LEAF) <-- Port:Instance and info about the Role (Leaf)
BGP
Verify BGP is updated by L2RIB
Leaf01#show bgp l2vpn evpn route-type 2 0 0006.f601.cd43 *
BGP routing table entry for [2][172.16.254.3:201][0][48][0006F601CD43][0][*]/20, version 268232
Paths: (1 available, best #1, table evi_201) <-- In the totally isolated evi context
Advertised to update-groups:
2
Refresh Epoch 1
Local
0.0.0.0 (via default) from 0.0.0.0 (172.16.255.3) <-- from 0.0.0.0 indicates local
Origin incomplete, localpref 100, weight 32768, valid, sourced, local, best <-- also indicates local
EVPN ESI: 00000000000000000000, Label1 20101
Extended Community: RT:65001:201 ENCAP:8 EVPN E-Tree:flag:1,label:0 <-- EVPN e-Tree attribute with Leaf flag = 1 (added to indicate this is a host address)
Local irb vxlan vtep:
vrf:not found, l3-vni:0
local router mac:0000.0000.0000
core-irb interface:(not found)
vtep-ip:172.16.254.3 <-- Local VTEP Loopback
rx pathid: 0, tx pathid: 0x0
Updated on Sep 14 2023 20:16:17 UTC
Remote RT2 Learning (Default Gateway RT2)
BGP
Verify BGP has learned the CGW RT2 prefix
Leaf01#show bgp l2vpn evpn route-type 2 0 0000.beef.cafe 10.1.201.1
BGP routing table entry for [2][172.16.254.3:201][0][48][0000BEEFCAFE][32][10.1.201.1]/24, version 114114
Paths: (1 available, best #1, table evi_201) <-- EVI context is 201
Flag: 0x100
Not advertised to any peer
Refresh Epoch 2
Local, imported path from [2][172.16.254.6:201][0][48][0000BEEFCAFE][32][10.1.201.1]/24 (global)
172.16.254.6 (metric 3) (via default) from 172.16.255.1 (172.16.255.1)
Origin incomplete, metric 0, localpref 100, valid, internal, best
EVPN ESI: 00000000000000000000, Label1 20101 <-- Correct segment identifier
Extended Community: RT:65001:201 ENCAP:8 EVPN DEF GW:0:0 <-- Default gateway attribute is added via the 'default gateway advertise CLI'
Originator: 172.16.255.6, Cluster list: 172.16.255.1
rx pathid: 0, tx pathid: 0x0
Updated on Sep 1 2023 15:27:45 UTC
L2RIB
Verify BGP updated L2RIB
- L2RIB learns local MAC from EVPN Mgr and sends to BGP and L2FIB. L2RIB is also responsible for learning remote MACs from BGP to update EVPN Mgr and L2FIB.
- L2RIB needs both Local and remote for other components to be properly updated.
- L2RIB component sits between local and remote MAC learning depending on which direction & component needs to be updated.
Leaf01#show l2route evpn default-gateway host-ip 10.1.201.1 EVI ETag Prod Mac Address Host IP Next Hop(s) ----- ---------- ----- -------------- --------------------------------------- -------------------------------------------------- 201 0 BGP 0000.beef.cafe 10.1.201.1 V:20101 172.16.254.6 <-- L2RIB has the MAC-IP of the Gateway programmed
L2FIB
Verify in L2FIB
- Component responsible for updating FED with the MACs to program in hardware.
- Remote MAC entries installed by L2FIB into FED-MATM are NOT punted to IOS-MATM. (IOS-MATM shows only local MACs, whereas FED-MATM displays both local and remote MAC).
- L2FIB output only shows remote MACs (It is not responsible for programming local MACs).
Leaf01#show l2fib bridge-domain 201 address unicast 0000.beef.cafe MAC Address : 0000.beef.cafe <-- CGW MAC Reference Count : 1 Epoch : 0 Producer : BGP <-- Learned from BGP Flags : Static Adjacency : VXLAN_UC PL:2973(1) T:VXLAN_UC [MAC]20101:172.16.254.6 <-- CGW Loopback IP PD Adjacency : VXLAN_UC PL:2973(1) T:VXLAN_UC [MAC]20101:172.16.254.6 Packets : 6979 Bytes : 0
FED
Verify in FED MATM
- At the hardware level of the Leafs configured with the 'protected keyword' you should only see the CGW default gateway MAC and the local host MACs.
- The switch looks at the RT2 prefix for the DEF GW attribute in order to determine what remote MAC is eligible to install.
Leaf01#show platform software fed switch active matm macTable vlan 201
VLAN MAC Type Seq# EC_Bi Flags machandle siHandle riHandle diHandle *a_time *e_time ports Con ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ 201 0000.beef.cafe 0x5000001 0 0 64 0x7a199d182498 0x7a199d183578 0x71e059173e08 0x0 0 82 VTEP 172.16.254.6 adj_id 9 No
<-- Only remote MAC installed in Fed is the Default Gateway (0x5000001 type) Conn = No (meaning not directly connected) 201 0006.f601.cd01 0x1 2458 0 0 0x7a199d1a2248 0x7a199d19eef8 0x0 0x7a199c6f7cd8 300 2 GigabitEthernet1/0/1 Yes 201 0006.f601.cd43 0x1 8131 0 0 0x7a199d195a98 0x7a199d19eef8 0x0 0x7a199c6f7cd8 300 84 GigabitEthernet1/0/1
<-- Two local MAC addresses (0x1 type) Conn = Yes (directly connected)
Total Mac number of addresses:: 5 Summary: Total number of secure addresses:: 0 Total number of drop addresses:: 0 Total number of lisp local addresses:: 0 Total number of lisp remote addresses:: 3 *a_time=aging_time(secs) *e_time=total_elapsed_time(secs) Type: MAT_DYNAMIC_ADDR 0x1 MAT_STATIC_ADDR 0x2 MAT_CPU_ADDR 0x4 MAT_DISCARD_ADDR 0x8 MAT_ALL_VLANS 0x10 MAT_NO_FORWARD 0x20 MAT_IPMULT_ADDR 0x40 MAT_RESYNC 0x80 MAT_DO_NOT_AGE 0x100 MAT_SECURE_ADDR 0x200 MAT_NO_PORT 0x400 MAT_DROP_ADDR 0x800 MAT_DUP_ADDR 0x1000 MAT_NULL_DESTINATION 0x2000 MAT_DOT1X_ADDR 0x4000 MAT_ROUTER_ADDR 0x8000 MAT_WIRELESS_ADDR 0x10000 MAT_SECURE_CFG_ADDR 0x20000 MAT_OPQ_DATA_PRESENT 0x40000 MAT_WIRED_TUNNEL_ADDR 0x80000 MAT_DLR_ADDR 0x100000 MAT_MRP_ADDR 0x200000 MAT_MSRP_ADDR 0x400000 MAT_LISP_LOCAL_ADDR 0x800000 MAT_LISP_REMOTE_ADDR 0x1000000 MAT_VPLS_ADDR 0x2000000 MAT_LISP_GW_ADDR 0x4000000
<-- the addition of these values = 0x5000001
MAT_LISP_REMOTE_ADDR 0x1000000
MAT_LISP_GW_ADDR 0x4000000
MAT_DYNAMIC_ADDR 0x1
Data Plane Adjacency
As a final step after confirming FED entry you can resolve the rewrite index (RI)
Leaf01#sh platform hardware fed switch active fwd-asic abstraction print-resource-handle 0x71e059173e08 0
<-- 0x71e059173e08 is taken from previous FED command riHandle for the CGW MAC Handle:0x71e059173e08 Res-Type:ASIC_RSC_RI Res-Switch-Num:255 Asic-Num:255 Feature-ID:AL_FID_L2_WIRELESS Lkp-ftr-id:LKP_FEAT_INVALID ref_count:1 priv_ri/priv_si Handle: 0x71e05917b8d8Hardware Indices/Handles: index0:0x38 mtu_index/l3u_ri_index0:0x0 index1:0x38 mtu_index/l3u_ri_index1:0x0 Features sharing this resource:58 (1)] Brief Resource Information (ASIC_INSTANCE# 0) ---------------------------------------- ASIC#:0 RI:56 Rewrite_type:AL_RRM_REWRITE_LVX_IPV4_L2_PAYLOAD_ENCAP_EPG(116) Mapped_rii:LVX_L3_ENCAP_L2_PAYLOAD_EPG(137) Src IP: 172.16.254.3 <-- source tunnel IP Dst IP: 172.16.254.6 <-- dest tunnel IP iVxlan dstMac: 0x9db:0x00:0x00 iVxlan srcMac: 0x00:0x00:0x00 IPv4 TTL: 0 iid present: 0 lisp iid: 20101 <-- Segment 20101 lisp flags: 0 dst Port: 4789 <-- VxLAN update only l3if: 0 is Sgt: 0 is TTL Prop: 0 L3if LE: 53 (0) Port LE: 281 (0) Vlan LE: 8 (0)
Note: You can also use 'show platform software fed switch active matm macTable vlan 201 detail' which chains this command with the FED command into one result
Configure (Partially Isolated)
Network Diagram
Note: This section only covers differences from Totally Isolated Segments.
- Routing-policy to mark the GCW gateway MAC IP with the DEF GW attribute
- Custom Device tracking policy required to prevent MAC flaps
- Static device-tracking binding for the GW MAC IP
Leaf-01 (Base EVPN Config)
Leaf-01#show run | sec l2vpn
l2vpn evpn
replication-type static
flooding-suppression address-resolution disable <-- Disables ARP caching so ARP is always sent up to the CGW
router-id Loopback1 l2vpn evpn instance 202 vlan-based encapsulation vxlan replication-type ingress multicast advertise enable
Leaf01#show run | sec vlan config vlan configuration 202 member evpn-instance 202 vni 20201 protected <-- protected keyword added
CGW (Base Config)
Set the replication mode under the nve
CGW#show run int nve 1 Building configuration... Current configuration : 313 bytes ! interface nve1 no ip address source-interface Loopback1 host-reachability protocol bgp member vni 20201 ingress-replication local-routing <-- 'ingress-replication' (Unicast all BUM traffic) / 'local-routing' (Enables vxlan centralized gateway forwarding) end
Configure the external gateway SVI
CGW#show run interface vlan 2021 Building configuration... Current configuration : 231 bytes ! interface Vlan2021 mac-address 0000.beef.cafe <-- MAC is static in this example for viewing simplicity. This is not required vrf forwarding pink <-- SVI is in VRF pink ip address 10.1.202.1 255.255.255.0 no ip redirects ip local-proxy-arp <-- Sets CGW to Proxy reply even for local subnet ARP requests ip pim sparse-mode ip route-cache same-interface <-- This is auto added when local-proxy-arp is configured. However, this is a legacy 'fast switching' command that is not used by CEF & is not required for forwarding ip igmp version 3 no autostate end
Create a policy with gleaning disabled
device-tracking policy dt-no-glean <-- Configure device tracking policy to prevent MAC-IP flapping security-level glean no protocol ndp no protocol dhcp6 no protocol arp no protocol dhcp4
Attach to externalgatewayevi/vlans
CGW#show running-config | sec vlan config vlan configuration 202 member evpn-instance 202 vni 20201
device-tracking attach-policy dt-no-glean <-- apply the new device tracking policy to the vlan configuration
Add static entries into device tracking table for externalgateway mac-ip
device-tracking binding vlan 202 10.1.202.1 interface TwentyFiveGigE1/0/1 0000.beef.cafe
<-- All static entries in device tracking table should be for external gateway mac-ip’s.
If there is any other static entry in device tracking table, match ip/ipv6 configurations in route map with prefix lists (permit/deny) are required to attach default gateway extended community to external gateway mac-ip routes only.
Create BGP route map to match RT2 MAC-IP prefixes and set the default gateway extendedcommunity
route-map CGW_DEF_GW permit 10
match evpn route-type 2-mac-ip <-- match RT2 type MAC-IP
set extcommunity default-gw <-- Set Default-gateway (DEF GW 0:0) extended community
route-map CGW_DEF_GW permit 20
Apply route-map to BGP Route Reflector neighbors
CGW#sh run | s r bgp
address-family l2vpn evpn
neighbor 172.16.255.1 activate
neighbor 172.16.255.1 send-community both
neighbor 172.16.255.1 route-map CGW_DEF_GW out <-- Sets the DEF GW Community when it advertises MAC-IP type RT2 to the RR
neighbor 172.16.255.2 activate
neighbor 172.16.255.2 send-community both
neighbor 172.16.255.2 route-map CGW_DEF_GW out <-- Sets the DEF GW Community when it advertises MAC-IP type RT2 to the RRVerify (Partially Isolated)
EVI Details
Leaf01#show l2vpn evpn evi 202 detail
EVPN instance: 202 (VLAN Based)
RD: 172.16.254.3:202 (auto)
Import-RTs: 65001:202
Export-RTs: 65001:202
Per-EVI Label: none
State: Established
Replication Type: Ingress
Encapsulation: vxlan
IP Local Learn: Enabled (global)
Adv. Def. Gateway: Enabled (global)
Re-originate RT5: Disabled
Adv. Multicast: Enabled
Vlan: 202
Protected: True (local access p2p blocked) <-- Vlan 202 is in protected mode
<...snip...>Local RT2 Generation (Local Host to RT2)
Covered in previous Totally Isolated Example
Remote RT2 Learning (Default Gateway RT2)
Covers the differences from Totally Isolated
CGW Default Gateway Prefix (Leaf)
Check that the prefix has the appropriate attribute in order to be eligible to be installed into hardware
Note: This is critical for DHCP L2 Relay to function
Leaf01#show bgp l2vpn evpn route-type 2 0 0000.beef.cafe 10.1.202.1
BGP routing table entry for [2][172.16.254.3:202][0][48][0000BEEFCAFE][32][10.1.202.1]/24, version 184602
Paths: (1 available, best #1, table evi_202) <-- the EVI context of 202 which matches the Vlan/EVI we are concerned about
Not advertised to any peer
Refresh Epoch 2
Local, imported path from [2][172.16.254.6:202][0][48][0000BEEFCAFE][32][10.1.202.1]/24 (global)
172.16.254.6 (metric 3) (via default) from 172.16.255.1 (172.16.255.1)
Origin incomplete, metric 0, localpref 100, valid, internal, best
EVPN ESI: 00000000000000000000, Label1 20201 <-- Correct Segment ID
Extended Community: RT:65001:202 ENCAP:8 EVPN DEF GW:0:0 <-- prefix has the Default GW attribute added
Originator: 172.16.255.6, Cluster list: 172.16.255.1
rx pathid: 0, tx pathid: 0x0
Updated on Sep 7 2023 19:56:43 UTCFED MATM (Leaf)
F241.03.23-9300-Leaf01#show platform software fed active matm macTable vlan 202 mac 0000.beef.cafe VLAN MAC Type Seq# EC_Bi Flags machandle siHandle riHandle diHandle *a_time *e_time ports Con ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ 202 0000.beef.cafe 0x5000001 0 0 64 0x71e058da7858 0x71e05916c0d8 0x71e059171678 0x0 0 5 VTEP 172.16.254.6 adj_id 651 No
<-- MAC of Default GW is installed in FED
SISF (CGW)
CGW#sh device-tracking database vlanid 202
vlanDB has 1 entries for vlan 202, 0 dynamic
Codes: L - Local, S - Static, ND - Neighbor Discovery, ARP - Address Resolution Protocol, DH4 - IPv4 DHCP, DH6 - IPv6 DHCP, PKT - Other Packet, API - API created
Preflevel flags (prlvl):
0001:MAC and LLA match 0002:Orig trunk 0004:Orig access
0008:Orig trusted trunk 0010:Orig trusted access 0020:DHCP assigned
0040:Cga authenticated 0080:Cert authenticated 0100:Statically assigned
Network Layer Address Link Layer Address Interface vlan prlvl age state Time left
S 10.1.202.1 0000.beef.cafe Twe1/0/1 202 0100 13006mn STALE N/A
IOS MATM (CGW)
CGW#show mac address-table address 0000.beef.cafe
Mac Address Table
-------------------------------------------
Vlan Mac Address Type Ports
---- ----------- -------- -----
201 0000.beef.cafe STATIC Vl201
2021 0000.beef.cafe STATIC Vl2021 <-- The Vlan 2021 SVI MAC advertised out Tw1/0/1
202 0000.beef.cafe DYNAMIC Twe1/0/1 <-- The Vlan 2021 SVI MAC learned dynamically after passing through the Firewall and the Dot1q tag swapped to fabric Vlan 202Troubleshoot
Address Resolution (ARP)
General steps for isolating ARP issues
- Confirm IMET tunnel is ready
- Capture on CGW Uplink to verify ARP received Encapsulated from Leaf
- If no ARP seen arriving encap on uplink
- Verify IMET tunnel is ready on both Leaf and CGW
- Capture on Leaf uplinks to confirm ARP is encapsulated and sent
- Troubleshoot intermediate path
- If ARP arrives on Border IMET tunnel capture but not programmed in VRF ARP table
- Troubleshoot CPU/CoPP punt path to confirm ARP punted to CPU
- Confirm IP address / client info is correct
- Debug ARP in VRF to see what might be impacting ARP process
- Verify CGW MAC installed as next hop / dest mac on the hosts
- Confirm CGW has both ARP entries with the real host MACs
- Verify firewall policy allows this tyep of traffic
Caution: Be careful when enabling debugs!
Ensure you have disabled flooding suppression
Leaf-01#show run | sec l2vpn
l2vpn evpn
replication-type static
flooding-suppression address-resolution disable <-- This CLI prevents a VTEP from trying to unicast other VTEPs based on its EVPN table
When host off Leaf-02 resolves ARP for host off Leaf-01 the ARP request is not broadcast to Leaf-01 directly
- The ARP is instead passed up the only BUM tunnel programmed on Leaf-02 toward the CGW
- The CGW does not forward this to Leaf-01, and instead replies with its own MAC
- This causes all communication to be passed up to the CGW then routed to between the hosts
- CGW routes packets, even when they are on the same local subnet
This diagram is to help visualize the flow of the ARP resolution process described in this section.
The ARP Request is show in purple
- This ARP request is to resolve the MAC address of the hos 10.1.202.10 off Leaf-01
- Notice that the purple line terminates at the CGW, and does not reach Leaf-01
The ARP Reply is shown in green
- The reply contains the MAC of the CGW SVI for Vlan 202
- Notice that the green line comes from the CGW, not from the actual host
Note: The red X is to indicate that this communication did not involve sending traffic to Leaf-01.
Observe the ARP entries on each respective host
Leaf02-HOST#sh ip arp 10.1.202.10 Protocol Address Age (min) Hardware Addr Type Interface Internet 10.1.202.10 1 0000.beef.cafe ARPA Vlan202 <-- MAC address for Leaf01 host is CGW MAC Leaf01-HOST#sh ip arp 10.1.202.11 Protocol Address Age (min) Hardware Addr Type Interface Internet 10.1.202.11 7 0000.beef.cafe ARPA Vlan202 <-- MAC address for Leaf02 host is CGW MAC
Observe on CGW the RT2 prefixes are learned. This is required for the CGW to route packets
CGW#sh bgp l2vpn evpn route-type 2 0 0006.f617.eec4 * <-- Leaf02 actual MAC
BGP routing table entry for [2][172.16.254.6:202][0][48][0006F617EEC4][0][*]/20, version 235458
Paths: (1 available, best #1, table evi_202)
Not advertised to any peer
Refresh Epoch 2
Local, imported path from [2][172.16.254.4:202][0][48][0006F617EEC4][0][*]/20 (global)
172.16.254.4 (metric 3) (via default) from 172.16.255.1 (172.16.255.1)
Origin incomplete, metric 0, localpref 100, valid, internal, best
EVPN ESI: 00000000000000000000, Label1 20201 <-- correct segment identifier
Extended Community: RT:65001:202 ENCAP:8 EVPN E-Tree:flag:1,label:0 <-- prefix contains the Leaf flag indicating this is a normal host
Originator: 172.16.255.4, Cluster list: 172.16.255.1
rx pathid: 0, tx pathid: 0x0
Updated on Apr 9 2025 17:11:22 UTC
CGW#sh bgp l2vpn evpn route-type 2 0 0006.f601.cd44 * <-- Leaf01 actual MAC
BGP routing table entry for [2][172.16.254.6:202][0][48][0006F601CD44][0][*]/20, version 235521
Paths: (1 available, best #1, table evi_202)
Not advertised to any peer
Refresh Epoch 2
Local, imported path from [2][172.16.254.3:202][0][48][0006F601CD44][0][*]/20 (global)
172.16.254.3 (metric 3) (via default) from 172.16.255.1 (172.16.255.1)
Origin incomplete, metric 0, localpref 100, valid, internal, best
EVPN ESI: 00000000000000000000, Label1 20201 <-- correct segment identifier
Extended Community: RT:65001:202 ENCAP:8 EVPN E-Tree:flag:1,label:0 <-- prefix contains the Leaf flag indicating this is a normal host
Originator: 172.16.255.3, Cluster list: 172.16.255.1
rx pathid: 0, tx pathid: 0x0
Updated on Apr 9 2025 17:17:06 UTC
Capture the ARP exchange on the uplinks to confirm bi-directional communication
- You can use Embedded Packet Capture (EPC) on the Fabric uplinks
- This scenario shows EPC on the Leaf01 Uplink. Repeat this same process on CGW if necessary
Configure the EPC
Leaf01#monitor capture 1 interface range te 1/1/2 , te 1/1/4 both match any buffer size 100 <-- both Uplinks toward fabric included
Start the capture
Leaf01#monitor capture 1 start
Inititate ping to trigger the ARP request (In this case ping is from Leaf01 host 10.1.201.10 to Leaf02 host 10.1.201.11)
Leaf01-HOST#ping vrf red 10.1.201.11
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.1.201.11, timeout is 2 seconds:
...!!
Success rate is 40 percent (2/5), round-trip min/avg/max = 1/1/1 ms
Stop Capture & Check for the ARP frames
Leaf01#mon cap 1 stop
F241.03.23-9300-Leaf01#show mon cap 1 buff br | i ARP
11 8.153510 00:06:f6:01:cd:42 -> ff:ff:ff:ff:ff:ff ARP 110 Who has 10.1.201.11? Tell 10.1.201.10 <-- .10 requests .11 MAC (this is Frame 11)
12 8.154030 00:00:be:ef:ca:fe -> 00:06:f6:01:cd:42 ARP 110 10.1.201.11 is at 00:00:be:ef:ca:fe <-- CGW replies with its MAC
View the capture packets in detail. If you want to see more info about the packest, use the detail option of EPC
- Be aware that this output is clipped in various places for brevity
Leaf01#show mon cap 1 buffer detailed | beg Frame 11 <-- begin detail result from Frame 11 (ARP Request)
Frame 11: 110 bytes on wire (880 bits), 110 bytes captured (880 bits) on interface /tmp/epc_ws/wif_to_ts_pipe, id 0
Ethernet II, Src: 00:00:00:00:00:00 (00:00:00:00:00:00), Dst: 00:00:00:00:00:00 (00:00:00:00:00:00) <-- Expected to see all zeros for outbound VxLAN encapped packet
Destination: 00:00:00:00:00:00 (00:00:00:00:00:00)
Address: 00:00:00:00:00:00 (00:00:00:00:00:00)
.... ..0. .... .... .... .... = LG bit: Globally unique address (factory default)
.... ...0 .... .... .... .... = IG bit: Individual address (unicast)
Source: 00:00:00:00:00:00 (00:00:00:00:00:00)
Address: 00:00:00:00:00:00 (00:00:00:00:00:00)
.... ..0. .... .... .... .... = LG bit: Globally unique address (factory default)
.... ...0 .... .... .... .... = IG bit: Individual address (unicast)
Type: IPv4 (0x0800)
Internet Protocol Version 4, Src: 172.16.254.3, Dst: 172.16.254.6 <--- Outer tunnel IP header
Source: 172.16.254.3
Destination: 172.16.254.6
User Datagram Protocol, Src Port: 65483, Dst Port: 4789 <-- VXLAN Dest port
Virtual eXtensible Local Area Network
VXLAN Network Identifier (VNI): 20101 <-- Verify the VNI for the segment you are investigating
Reserved: 0
Ethernet II, Src: 00:06:f6:01:cd:42 (00:06:f6:01:cd:42), Dst: ff:ff:ff:ff:ff:ff (ff:ff:ff:ff:ff:ff) <-- Start of inner payload info
Type: ARP (0x0806)
Trailer: 000000000000000000000000000000000000
Address Resolution Protocol (request) <-- is an ARP request
Hardware type: Ethernet (1)
Protocol type: IPv4 (0x0800)
Hardware size: 6
Protocol size: 4
Opcode: request (1)
Sender MAC address: 00:06:f6:01:cd:42 (00:06:f6:01:cd:42) <-- Sending host
Sender IP address: 10.1.201.10
Target MAC address: 00:00:00:00:00:00 (00:00:00:00:00:00) <-- Trying to resolve MAC for host
Target IP address: 10.1.201.11
Frame 12: 110 bytes on wire (880 bits), 110 bytes captured (880 bits) on interface /tmp/epc_ws/wif_to_ts_pipe, id 0 <-- ARP reply
Ethernet II, Src: dc:77:4c:8a:6d:7f (dc:77:4c:8a:6d:7f), Dst: 68:2c:7b:f8:87:48 (68:2c:7b:f8:87:48) <-- Underlay MACs
Internet Protocol Version 4, Src: 172.16.254.6, Dst: 172.16.254.3 <-- From CGW loopback to Leaf01 loopback
User Datagram Protocol, Src Port: 65410, Dst Port: 4789
Virtual eXtensible Local Area Network
VXLAN Network Identifier (VNI): 20101
Reserved: 0
Ethernet II, Src: 00:00:be:ef:ca:fe (00:00:be:ef:ca:fe), Dst: 00:06:f6:01:cd:42 (00:06:f6:01:cd:42) <-- Start of payload
Type: ARP (0x0806)
Trailer: 000000000000000000000000000000000000
Address Resolution Protocol (reply) <-- is an ARP reply
Hardware type: Ethernet (1)
Protocol type: IPv4 (0x0800)
Hardware size: 6
Protocol size: 4
Opcode: reply (2)
Sender MAC address: 00:00:be:ef:ca:fe (00:00:be:ef:ca:fe) <-- Reply is that of the CGW MAC due to local proxy on SVI
Sender IP address: 10.1.201.11
Target MAC address: 00:06:f6:01:cd:42 (00:06:f6:01:cd:42)
Target IP address: 10.1.201.10
CGW RT2 Gateway Prefix
Gateway Prefix Missing
As mentioned in the previous section on Partially Isolated segments the MAC is required to be learned in the fabric Vlan
- This issue can manifest if there s no traffic destined for the gateway for longer than the MAC aging timer.
- If the CGW Gateway prefix is missing, you need to confirm the MAC is present
CGW#show bgp l2vpn evpn route-type 2 0 0000.beef.cafe 10.1.202.1
% Network not in table <-- RT2 not generated on CGW
CGW#show mac address-table address 0000.beef.cafe
Mac Address Table
-------------------------------------------
Vlan Mac Address Type Ports
---- ----------- -------- -----
201 0000.beef.cafe STATIC Vl201
2021 0000.beef.cafe STATIC Vl2021
<-- MAC is not learned in Fabric Vlan 202
Total Mac Addresses for this criterion: 2
Gateway Prefix Missing Remediation
In most production networks there is likely to be some traffic at all times. However, if you are having this issue you can use one of these options to remediate the issue:
- Add static MAC entry such as 'mac address-table static 0000.beef.cafe vlan 202 interface TwentyFiveGigE1/0/1'
- Increase the MAC aging timer with 'mac address-table aging-time <seconds>'. (Keep in mind this increases the aging time for all MAC addresses, so the static MAC option is preferred)
Missing DEF GW Attribute
With Partially Isolated Segments there are a number of additional configurations to add this attribute.
Missing DEF GW Attribute Remediation
Confirm these details:
- You are running 17.12.1 or later
- The SISF (Device-Tracking) CLI is present in the configuration
- The route-map match & set commands are configured and route-map is applied to the BGP neighbors
- You have refreshed the BGP advertisements (you must clear BGP to re-advertise the prefix with the new attribute
Wireless Roaming
Frequent roaming can cause BGP to update too frequently & roaming per time interval should be increased before switch declares it owns the MAC and sends RT2 Update
- This occurs when a host moves between two APs that are on different switches.
- Default limit for roam is 5 per 180 seconds
Leaf01#sh run | sec l2vpn
l2vpn evpn
replication-type static
flooding-suppression address-resolution disable
ip duplication limit 10 time 180 <--- You can adjust this default in the global l2vpn section
mac duplication limit 10 time 180
Leaf01#sh l2vpn evpn summary
L2VPN EVPN
EVPN Instances (excluding point-to-point): 4
VLAN Based: 4
Vlans: 4
BGP: ASN 65001, address-family l2vpn evpn configured
Router ID: 172.16.254.3
Global Replication Type: Static
ARP/ND Flooding Suppression: Disabled
Connectivity to Core: UP
MAC Duplication: seconds 180 limit 10
MAC Addresses: 13
Local: 6
Remote: 7
Duplicate: 0
IP Duplication: seconds 180 limit 10
IP Addresses: 7
Local: 4
Remote: 3
Duplicate: 0
<...snip...>Commands to Collect for TAC
In the event this guide did not resolve your issue please collect the command list shown and attach them to your TAC service request.
Minimum info to collect
(limited time to gather data prior to reload/recovery action)
- Show tech evpn
- Show tech
- Show tech sisf
Detailed info to collect
(If there is time to collect more complete data, this is preferred)
-
show tech
-
show tech evpn
-
show tech platform evpn_vxlan switch <number>
-
show tech platform
-
show tech resource
-
show tech sisf
-
show tech isis
-
show tech bgp
-
show monitor event-trace evpn event all
-
show monitor event-trace evpn error all
-
request platform software trace archive
Related Information
- Implement BGP EVPN Routing Policy on Catalyst 9000 Series Switches
- DHCP Layer 2 Relay (coming soon)
Revision History
| Revision | Publish Date | Comments |
|---|---|---|
1.0 |
20-Sep-2023 |
Initial Release |
FeedbackContact Cisco
- Open a Support Case
- (Requires a Cisco Service Contract)