Terminate state on gateway change

[jeliasson]

Hello everyone,

I'm currently dealing with a setup involving two gateways: the primary gateway (WAN1) is a wired unmetered connection, while the secondary gateway (WAN2) is a metered LTE connection. The failover to the LTE connection works seamlessly when the primary gateway (WAN1) goes down, but I'm facing a specific challenge.

The issue I'm encountering is that persistent connections (SSH, RDP, etc.) continue to use WAN2 even after the primary gateway is restored following a period of downtime. This results in unnecessary latency and, naturally, unwarranted traffic on the metered connection.

I'm wondering if there's a method to terminate states associated with a specific WAN interface when an active gateway changes. Essentially, I'm seeking a way to terminate WAN2 states when WAN1 becomes active. Alternatively, I'm open to discovering a more graceful approach to dropping secondary connections, prompting clients to reconnect/reestablish with the now active WAN1.

Any insights or suggestions on how to address this would be greatly appreciated.

Thanks!

[jeliasson]

Bump. If there's any part of my explanation that seems unclear, let me know, and I'll try to express it differently. Thanks again!

[firewall]

You should be able to smooth this over a bit by disabling sticky connections (advanced firewall settings). Research the possible implications of doing so, however. I ventured down that path and encountered more bullcorn than benefit.

As for how to flip it back to WAN1 with sticky conn enabled, you might have success with automating ifdown/ifup on WAN2 if a test for connectivity on WAN1 succeeds (shellscript+cron). Also you may need to account for the gateway interval period by inserting a sleep routine in the script.

These are the commands I use for another LTE connectivity need:

Code: [Select]

/usr/local/sbin/configctl interface linkup stop ue0
/usr/local/sbin/configctl interface reconfigure ue0
/usr/local/sbin/configctl interface linkup start ue0