This feature provides coexistence between traditional Default Gateways using First Hop Gateway Protocol (HSRP being the mode supported in this release), and Distributed Anycast Gateway (DAG) for VXLAN EVPN fabrics. Instead of a disruptive cut-over or inefficient hair pinning, Default Gateways with HSRP can now be active at the same time as VXLAN EVPNs DAG, as long as the common Default Gateway MAC and IP is configured. The functionality as part of this feature provides ease for migration and coexistence between Classic Ethernet / FabricPath and VXLAN EVPN fabrics. This functionality is solely enabled on the VXLAN EVPN side, more specifically on the Border nodes neighboring the Classic Ethernet / FabricPath network. This feature allows more efficient routing and less disruptive migrations without the requirement for Software or Hardware upgrades on the Classic Ethernet / FabricPath side.

Migration can now be performed with minimal traffic impact even when both DAG is functional on VXLAN network and HSRP gateway is functional on Classic Ethernet / FabricPath network for the same VLAN after the premigration step is performed on the Classic Ethernet / FabricPath HSRP gateway. For more information, see details for premigration step in Configuring Classic Ethernet / FabricPath to VXLAN Migration.

Coexistence of both DAG and HSRP gateway was not possible earlier for the same VLAN even after the premigration step was performed. This coexistence will enable optimal routing for the Layer 3 workloads that are migrated to VXLAN network during migration.

Layer 2 Interconnection

• Interconnecting the two networks via Layer 2 is crucial to facilitate seamless workload migration from Classic Ethernet / FabricPath to VXLAN.

• The border leaf on VXLAN network is connected via a Layer 2 interface to the Classic Ethernet / FabricPath network.

• The Layer 2 link can be a port channel trunk or a physical Ethernet trunk.

• The VXLAN border leaf switch can be a vPC or a NX-OS switch and the switch can be a TOR or an EOR. Similarly, the Classic Ethernet / FabricPath border-edge switch can be a vPC or a NX-OS switch. The switch could also host the HSRP gateway for the Classic Ethernet / FabricPath network.

For migration, you must configure the following on the VXLAN border leaf:

• The Layer 2 ports connecting the two network infrastructures must be configured as port-type external . These ports are referred as external interfaces.

• A unique Burned In Address (BIA) address for IPv4 and IPv6 must be configured on the SVI of each VXLAN border leaf during migration of the VLAN.

• If the VXLAN border leaf is in a vPC configuration, then the BIA address for the SVI must be different on both switches.

The following table provides few Layer 2 interconnection combinations:

• Ingress PACL region must be carved and made available before configuring the migration of workloads for EX/FX/FX2/FX3/GX/GX2B platforms deployed as VXLAN border leaf nodes.

  For example: You need to verify if the PACL region is carved before configuring the port-type external command on the ports connecting the VXLAN and Classic Ethernet / FabricPath networks. You can verify if the ingress PACL region is configured by using the show hardware access-list tcam region command. If the region is unavailable, configure the region using the hardware access-list tcam region ing-ifacl 512 command. Ensure that you reload the switch after the PACL region is configured.

• Verify that there is no ingress PACL policy configured on the external interfaces before migration. If they are configured, you must remove them before configuring the port-type external command.

• vPC Fabric Peering, Egress CNTACL, VRRP, and VXLAN Flood and Learn are not supported with this migration. Also, this migration does not support moving workloads that are multicast sources or receivers.

• It is recommend that you configure only up to six external interfaces.

• For migration, ensure that you do not have the Extended IFACL feature configured using the hardware access-list tcam label ing-ifacl 6 command.

• Migration of IPv4 and IPv6 applications must be performed sequentially as mentioned below:

  1. Premigration step must be performed on HSRP gateway for IPv4 gateway-IP for a particular VLAN. For more information, see details for premigration step in Configuring Classic Ethernet / FabricPath to VXLAN Migration.

  2. The migration procedure in terms of configuring SVIs with BIA address for IPv4 must be performed on each VXLAN border leaf node connecting to the Classic Ethernet / FabricPath network.

  3. Migrate all the IPv4 hosts from Classic Ethernet / FabricPath to VXLAN side.

  4. After all the IPv4 hosts for all VLANs are migrated from Classic Ethernet / FabricPath to VXLAN, the premigration step and migration procedure has to be repeated for IPv6.

  Note	It is recommended that you limit the migration of concurrent host to a maximum of 1000 hosts. Start the next migration only after the previous migration of hosts is complete.

• This feature is not supported on N9K-C92348GC.

• If we have a vPC VXLAN border leaf configured, Layer 3 peer-router needs to be enabled.

• If the Suppress ARP or Suppress ND feature is enabled on the VXLAN network during Classic Ethernet / FabricPath to VXLAN migration, the host must be learned in the respective ARP or ND tables on the VXLAN border leaf. You can send a GARP/ND before moving the host to VXLAN.

  If adjacency is not learned for the host that is moved to VXLAN, then traffic from the host behind Classic Ethernet / FabricPath network to this host can fail on the Classic Ethernet / FabricPath network.

  For example:

  • When host 10.10.1.8 is being moved to VXLAN, initially, it is not learned as shown:

    switch# sh ip arp 10.10.1.8 vrf vrf1501
    
    IP ARP Table
    Total number of entries: 0
    Address         Age       MAC Address     Interface       Flags
    switch# 
    
    switch(config)# sh ip route 10.10.1.8 vrf vrf1501
    
    10.10.1.0/24, ubest/mbest: 2/0, attached
        *via 10.10.1.1, Vlan1001, [0/0], 22:55:42, direct
        *via 10.10.1.4, Vlan1001, [0/0], 22:55:42, direct

  • After sending GARP from host 10.10.1.8, the ARP table output of the border leaf switch is as shown:

    switch# sh ip arp 10.10.1.8 vrf vrf1501
    
    Flags: * - Adjacencies learnt on non-active FHRP router
           + - Adjacencies synced via CFSoE
           # - Adjacencies Throttled for Glean
           CP - Added via L2RIB, Control plane Adjacencies
           PS - Added via L2RIB, Peer Sync
           RO - Re-Originated Peer Sync Entry
           D - Static Adjacencies attached to down interface
    
    IP ARP Table
    Total number of entries: 1
    Address         Age       MAC Address     Interface       Flags
    10.10.1.8       00:00:04  0000.8aa9.79d3  Vlan1001      
    
    switch(config)# sh ip route 10.10.1.8 vrf vrf1501
    
    10.10.1.8/32, ubest/mbest: 1/0, attached
        *via 10.10.1.8, Vlan1001, [190/0], 00:00:14, hmm

  • After GARP, the host is moved to leaf in the VXLAN network as shown:

    switch(config)# sh ip route 10.10.1.8 vrf vrf1501
    
    10.10.1.8/32, ubest/mbest: 1/0
        *via 2.2.2.5%default, [200/0], 00:00:23, bgp-200, internal, tag 200, segid: 
    11501 tunnelid: 0x2020205 encap: VXLAN

• Beginning with Cisco NX-OS Release 10.5(2)F, the following features are supported on Cisco Nexus 9500 Series switches with N9K-X9736C-FX3 line card:

  • HSRP

  • VRRP

  • VXLAN Flood and Learn (Static Ingress Replication)

  • FabricPath to VXLAN Migration