Part 4 of a series on making VPP an eVPN/VxLAN speaker, focusing on automated failover and health checking. The author introduces vpp-evpn, a control/management plane built on VPP and Bird, and details a new health subsystem (evpnh) that sends authenticated IPv6 link-local multicast heartbeats between nodes sharing a bridge domain. The system uses HAProxy-style rise/fall counters to prevent flapping, a self-fence mechanism for stuck gateways, and a central registry (vpp-evpnr) that autonomously promotes healthy candidates when a primary fails. Implementation details cover the 124-byte wire format (HMAC-SHA256 authenticated, similar to CARP/VRRP), Go concurrency patterns, Prometheus observability, a SolidJS WebUI with SSE event streaming and TOFU admin setup, and CLI extensions for health and traffic visibility. The system is running in production on four nodes with non-orchestrated failover taking ~8 seconds.
