45
Dell EMC VxRail Multirack Deployment Guide
A
Routing in a Layer 2 VXLAN overlay
In this document, the VxRail nodes are deployed using
sfo01w01vc01.sfo01.rainpole.local
as an
external vCenter server. The vCenter server is part of an existing management cluster that is encapsulated in
a separate VXLAN tunnel, virtual-network 1611. The VxRail uses virtual-network 1631 for external
management which connects the hosts to the vCenter server. Figure 22 shows the two virtual networks. Both
virtual-networks (shown as dashed lines) are statically configured on switch
sfo-edge01
. In this diagram
sfo01w01
represents all eight nodes in VLAN 1631.
OSPF Area 0
Spine 1
Z9264-ON
Spine 2
Z9264-ON
sfo01w01
VxRail Nodes
sfo01-leaf04
10.222.222.3
sfo01-leaf03
10.222.222.3
sfo01-leaf02
10.222.222.1
sfo01-leaf01
10.222.222.1
sfo01m01
Management cluster
sfo-edge03
sfo-edge01
10.222.222.64
External Management
Existing Management
ID VLAN Name
vCenter
IP: 172.16.11.62/24
GW: 172.16.11.253
sfo01w01vxrail01:
IP: 172.16.31.101/24
GW: 172.16.31.253
VLAN 1611:
IP: 172.16.11.253/24
VLAN 1631:
IP: 172.16.31.253/24
Routing between L2 VXLAN virtual networks
The vCenter server
sfo01w01vc01.sfo01.rainpole.local
has the IP address
172.16.11.62/24
and
the first VxRail node,
sfo01w01vxrail01
has the IP address
172.16.31.101/24
. In the current
implementation of a static L2 VXLAN, routing in and out of tunnels (RIOT) is not supported. Additionally, a
VLAN participating in a virtual network cannot be configured as an Layer 3 (L3) interface.
To allow two hosts in two separate VXLANs to communicate, the VXLAN header must be removed so that
inter-VLAN routing can take place. This process allows the payload to be encapsulated using the destination
VXLAN. To achieve this, the network for both IP addresses is defined on
sfo-edge03
, which in this
example, is a Dell Networking S4048-ON. This switch has an interface on each VNI and performs the inter-
VLAN routing. Traffic is then re-encapsulated by
sfo-edge01
to reach the target destination
Note:
Router redundancy is not shown in this example for simplicity. Best practices for router redundancy
should be considered during deployment.