Hello! Before I get into the technical information, I want to give some specific background information.
I do not have any technical knowledge or experience when it comes to networking. I have this vague understanding of how everything works and I know how to google, so I can figure out most things fairly quickly. I couldn't explain it, and I'll definitely run into multiple issues on my way there, but I just continue troubleshooting until I learn something. Because of all this; I find it extremely difficult to ask for help or describe an issue when I run into one. As a result, I turned to ai for help.
Im not a big fan of ai but I use it for technical purposes and always give it a specific prompt for efficiency. Because of this, the most recent ai that was helping me (claude) kept running into the same issue and eventually told me that it cannot assist and I should turn to human experts on this forum. Thank god. That's really all I've been working toward. Really just working with the ai to better understand this stuff and how to ask for help. Below is what Claude typed up. I am happy to provide any other information once I am back at the work station. I am no longer working with ai for this. Thanks in advance!
Device: GL.iNet GL-MT6000 (Flint 2) Firmware: 4.8.3 Setup: Single WAN (AT&T Fiber via passthrough gateway) + Mullvad VPN (WireGuard, all-traffic routing) + AdGuard Home
Problem: I'm experiencing twice-daily network outages at predictable 12-hour intervals. Through kernel log analysis I've confirmed the cause: kmwan is tearing down and rebuilding the WAN interface during routine DHCP lease renewal.
AT&T issues 24-hour leases. At the 12-hour renewal mark, the brief connectivity interruption causes kmwan's ping health checks to fail. With track_mode='force', kmwan then deletes and recreates the WAN node, disrupting all connected clients.
Kernel log confirms:
kmwan: Delete node:wan
add node success. iface:wan, dev:eth1
Current kmwan config:
kmwan.global.enable='1'
kmwan.global.level='medium'
kmwan.global.sensitivity='3000'
kmwan.global.mode='failover'
kmwan.wan.track_mode='force'
kmwan.wan.addr_type='4'
kmwan.wan.check='1'
What I've tried:
-
Setting
kmwan.global.enable='0'— immediately killed WAN, had to re-enable -
Setting
kmwan.wan.track_mode='normal'+sensitivity='10000'— also killed WAN, track_mode change didn't persist
It's clear kmwan is actively managing WAN on this firmware, not just monitoring it. I cannot disable it or change track_mode without losing connectivity.
Questions:
-
Is there a supported way to make kmwan tolerate the brief interruption during DHCP renewal without triggering a WAN teardown?
-
Does Mullvad WireGuard (all-traffic routing) interfere with DHCP renewal on eth1, and if so, how is this handled correctly?
-
Is there a way to extend or adjust the DHCP lease renewal behavior on the WAN side to make it more resilient?
I'm on a single WAN connection with no failover needs, so kmwan's multi-WAN features aren't being used. I just need the connection to survive its own DHCP renewal every 12 hours.
Any help appreciated.
