The vSphere Web Client UI provides two major sections relevant to NSX routing.
These include the L2 and control-plane infrastructure dependencies and the routing subsystem configuration.
NSX distributed routing requires functions that are provided by the Controller Cluster. The following screen shot shows a Controller Cluster in a healthy state.
- There are three controllers deployed.
- The “Status” for all controllers is “Connected”.
- The software version for all controllers is the same.
- Each controller node has two peers.
Host kernel modules for distributed routing are installed and configured as part of VXLAN configuration on the host. This means distributed routing requires that ESXi hosts are prepared and VXLAN is configured on them.
- “Installation Status” is green.
- “VXLAN” is “Configured."
Make sure that VXLAN transport components are correctly configured.
- The VLAN ID must be correct for the VTEP transport VLAN. Note that in the screen shot above it is “0." In most real-world deployments this would not the case.
- MTU is configured to be 1600 or larger. Make sure that the MTU is not set to 9000 with the expectation that the MTU on VMs would be also set to 9000. The DVS maximum MTU is 9000, and if VMs are also at 9000, there is no space for VXLAN headers.
- VMKNics must have the correct addresses. Make sure that they are not set to 169.254.x.x addresses, indicating that nodes have failed to get addresses from DHCP.
- The teaming policy must be consistent for all cluster members of the same DVS.
- The number of VTEPs must be the same as the number of dvUplinks. Make sure that valid/expected IP addresses are listed.
Transport Zones have to be correctly aligned to DVS boundaries, to avoid the situation in which the DLR is missing on some clusters.