With two DCs, the Global Manager runs as two full clusters:
- GM-Active (DC1): 3 VMs behind a cluster VIP. This is where you log in and configure.
- GM-Standby (DC2): 3 VMs holding a full, constantly synced copy of the config. It takes over only when an admin activates it.
Why not spread one cluster across both DCs? A 3-VM cluster survives losing 1 VM. With two sites, one DC always holds 2 of the 3, so losing that DC takes the whole cluster down.
Who talks to whom:
1️⃣ GM-Active to GM-Standby: full config sync
2️⃣ GM to each LM: only the objects that DC needs
3️⃣ LM to LM: stretched group members and MAC/IP tables
4️⃣ Edge to Edge: data traffic over RTEP
Tip: register each LM with its cluster VIP, never a single VM IP.
Other parts of this series:
🔗 Part 1 (Architecture): Part 1: NSX Federation in VCF 9.1 – the architecture in 60 seconds
🔗 Part 3 (Stretched segment): Part 3: NSX Federation in VCF 9.1 – one subnet-two DCs in 60 seconds
🔗 Part 4 (Gateway options): Part 4: NSX Federation in VCF 9.1 – Gateway options in 60 seconds
