In these three parts, I will speak about the configuration of the interfaces and uplinks respectively for physical ESXi and Edge VM.
- Part 1: ESXi interfaces
- Part 2: Edge VM for T0 and T1 with one uplink North-South (physical world)
- Part 3: Edge VM for T0 and T1 with redundant uplinks North-South (physical world)
Note: In our infrastructure the MTU for the tunnel transport (GENEVE) is set to 9000, it has to be greater or equal than 1600.
I have a vlan for the TEP IP (VLAN 825) and an IP pool. Like this during deployment of a new ESXi or Edge, I don’t have to set and manage manually the IP for the TEP.
For ESXi, we can add the number of physical interfaces we want.
In my case, I separate the compute and the storage:
- vmnic0 and vmnic2 for IaaS (NSX)
- vmnic1 and vmnic3 for ScaleIO storage.
I have connected the different transport zones (Overlay, Overlay-PKS and VLAN-Legacy) on the same N-VDS (NVDS-T-01 for Test).
With this configuration, when I add an ESXi in NSX, I only need to specify which of these transport zones I want and the NVDS.
We must then choose the Uplink Profile, which in my case puts the vmnic0 and vmnic2 in active-active mode (Load Balance Source – depending on the Source Port ID of the VM linked to one of the interfaces (will always go through the same interface for the same port ).
In active-active mode, we need two IP TEP, one for each interface.
At the end, we have to map the vmnic0 and vmnic2 with the physical interfaces (vCenter).