On this page
It is Saturday evening, the EVE-NG host is idle, and six Nexus 9000v switches are waiting for a controller. I want a Nexus Dashboard cluster with its data interface on the same wire as those switches, so the fabric controller can discover them without a single static route. This post gets the cluster up, from a qcow2 in a download folder to an empty Fabrics page. Part 2 builds the VXLAN EVPN fabric in the NDFC GUI, one screen at a time, and Part 3 does the same build with Ansible. The format is the one I like in a lab guide: an instruction, then the screen it produces.
Lab Topology
Everything runs on one EVE-NG Pro 7.2 host, and two Linux bridges do all the work. pnet0, the management bridge, is my home LAN. nat0, the NAT cloud bridge, carries the Nexus Dashboard data interface, every switch mgmt0 and the NDFC persistent service addresses. A small always-on KVM guest called dns01 serves DNS and NTP to both segments, because the Cluster Bringup wizard marks DNS and NTP as required fields and I wanted them answered locally.
The fabric NDFC will manage is two spines and four leaves as two vPC pairs, plus four servers, each dual-homed to a pair with an LACP bond. Server1 and Server3 live on NET-WEB, Server2 and Server4 on NET_APP, so Part 2 gets to prove same-subnet and routed reachability across the fabric with the same four hosts.
