Title: Flint 4 Dual PPPoE Multi-WAN and WireGuard Design Review
Hello GL.iNet team,
I would appreciate guidance on the recommended design for my Flint 4 homelab. The network is operational and stable; this is not a troubleshooting request.
Current setup:
-
GL.iNet Flint 4 as the Internet edge router
-
Two independent PPPoE WAN connections
-
Multi-WAN enabled
-
Grandstream GWN7821P managed switch
-
Proxmox VE with an OPNsense VM
-
Synology NAS
-
Grandstream access points
-
Home Assistant and other internal services
-
WireGuard for remote access, including a GL.iNet travel router when away from home
My goal is to keep the benefits of Multi-WAN for normal LAN traffic while maintaining predictable egress behavior for WireGuard clients.
Desired behavior:
-
General LAN traffic, browsing, downloads, updates, and IoT traffic can use both WAN connections through Multi-WAN.
-
WireGuard clients using full tunnel should normally exit through one preferred residential WAN and therefore retain one public IP address during a session.
-
This is important for remote work and services that can be affected by an unexpected public-IP change, including streaming services.
-
If the preferred WAN fails, general Internet access should continue through the remaining WAN. I understand that an active VPN or streaming session may need to reconnect after a real WAN failure; I mainly want to avoid IP changes during normal healthy Multi-WAN operation.
The architecture I am considering is:
-
Keep the WireGuard endpoint and WireGuard full-tunnel egress associated with WAN 1.
-
Allow ordinary LAN traffic to remain eligible for Multi-WAN load balancing across WAN 1 and WAN 2.
-
Use WAN 2 as failover when WAN 1 is unavailable.
I would appreciate clarification on the following:
-
Is this the recommended design for Flint 4 and KMWAN?
-
What is the supported way to keep WireGuard server traffic and full-tunnel client egress on one preferred WAN: policy-based routing, WAN binding, or another mechanism?
-
Does KMWAN provide session persistence, sticky connections, or connection affinity relevant to long-lived WireGuard-originated sessions?
-
What should I expect from an active WireGuard session if the preferred WAN fails?
-
When WAN 1 returns, what is the recommended behavior for returning VPN traffic to the preferred WAN without unnecessarily disrupting users?
-
Would GL.iNet recommend keeping these routing decisions on the Flint 4, or moving any advanced policy-routing functions to my downstream OPNsense VM?
-
Is there an official reference architecture or documentation for dual PPPoE WAN, Multi-WAN, WireGuard server operation, and deterministic VPN egress?
I can provide sanitized screenshots of the Multi-WAN and WireGuard configuration if needed. I have removed credentials, public IP addresses, keys, and configuration exports from the attached topology.
Thank you for any guidance or documentation you can share.