Tailscale launches MagicRoute beta to replace region-based routing

Tailscale has released MagicRoute in beta for Premium and Enterprise customers, replacing its previous regional routing system. The new feature scores connectors from each client's perspective using location signals rather than DERP region proximity, which had caused connectors to flap between regions during minor latency jitter and break active connections. MagicRoute keeps clients attached to their preferred connector as long as it remains healthy, and supports tailnets with custom DERP servers that were previously excluded.

How MagicRoute differs from regional routing

• Old system: Regional routing grouped connectors by home DERP server; clients matched to same/nearest DERP region
• Flapping problem: Connectors on fast backbones had nearly identical latency to multiple DERP regions, causing constant switching that broke SSH sessions and file transfers
• New scoring: MagicRoute ranks connectors per-client using coarse location (approximate distance), which is stable unlike latency measurements
• Failover behavior: When preferred connector fails, client works down ranked list to next healthy option
• Console options: Four modes now available—active-passive failover, regional routing, regional with in-region failover (support contact required), and MagicRoute
• Client support: All versions handle MagicRoute for subnet routers, app connectors, and Tailscale Service hosts; exit node recommendations require client 1.86.0+
• No granular opt-out: MagicRoute is tailnet-wide by design to prevent unexpected routing interactions

Why it matters

For organizations running multiple Tailscale subnet routers or app connectors, MagicRoute reduces operational fragility by eliminating a common source of cascading connection failures. The change also removes a barrier for users with custom DERP deployments.