NSX Installation, Configuration, and Upgrade
Planning and lifecycle
Installation starts with compatibility: check compute manager, host, Edge, and Manager versions; DNS/NTP; IP pools; management network; and MTU. Register the compute manager, deploy the Manager cluster, configure IP pools and transport zones, prepare hosts as transport nodes (N-VDS/VDS), and deploy Edge nodes in an Edge cluster. Uplink profiles bind physical NICs, VLANs, teaming, and MTU. Overlay transport requires TEP IPs and GENEVE connectivity. A transport node must reach other TEPs; Edge nodes need uplink connectivity and an Edge cluster. Host preparation may use a vSphere Distributed Switch or an NSX-managed virtual switch depending on design; confirm support.
Upgrade is lifecycle work: back up configuration, verify interoperability, upgrade management components before transport components, then validate health. Use upgrade coordinator where available and do not skip readiness checks. Configuration drift between desired and realized state is diagnosed with alarms and transport node status.
Traps: confusing management IP with TEP IP; forgetting MTU on the physical underlay; assuming the Manager cluster provides data-plane redundancy; upgrading hosts before Manager; treating a transport zone as a security boundary.
Rule of thumb: underlay first, management next, transport last, then test. Worked example: if overlay VMs on different hosts can ping but large packets fail, check underlay and TEP MTU; GENEVE adds overhead, so the physical path must support the larger frame.