I have a RT-AC1900P running current version of Merlin firmware. Connected to an AT&T BGW 210 gateway in IP Passthrough mode. Has worked reliably for years. I also have an Ooma VOIP device with T-Mobile LTE backup, on which you're allowed to do some non-VOIP traffic. I found out that the limit on that non-VOIP traffic before you pay huge amounts of money is 1 GB/month. Before I found that out, I was toying with setting up Dual WAN (with automatic failover) and having the Ooma/T-Mobile connection as the second. As soon as I figured out the cost, I pulled the Ethernet cable that would allow the T-Mobile connection. I would occasionally get issues where the RT-AC1900P would appear to lose access to the BGW 210 for a few minutes, then resume access like nothing had happened, and I thought it might have to do with the failover and turned the dual wan feature off completely.
Soon after, I started to have disconnections that never self-resolved. I had to either reboot the BGW 210, the AC1900P, or disable and re-enable the WAN. Once I did any of these, it came back. Pulling the Ethernet cable from the AC1900P to the BGW 210 and then putting it back wouldn't resolve the situation. While the issue was going on, I often had an IPSEC VPN tunnel up from one PC that would completely unaffected by the issue, while nothing else could get a connection.
Eventually, I thought, maybe there's something in the Dual WAN feature still affecting this. I re-enabled Dual WAN to look at the settings and I found that the failover test settings were still there, with the change that auto-reconnect was now unchecked. I disabled failover testing completely (which I would naively have thought turning off Dual WAN should have done) and then turned Dual WAN back off. My theory is that the auto-failover feature did not get implicitly turned off with the Dual WAN feature, finding an issue merely disabled new connections on the primary WAN, and since the auto-reconnect was disabled, there was no way for it to recover.
I have run without issue for over a week now (and prior to removing the auto failover test, 2 days was the max I would go). This sounds like a bug, but I have no idea how/where I could/should report this. Clearly, should I want to activate auto-failover, I need to find a better test than what I was using before.
Soon after, I started to have disconnections that never self-resolved. I had to either reboot the BGW 210, the AC1900P, or disable and re-enable the WAN. Once I did any of these, it came back. Pulling the Ethernet cable from the AC1900P to the BGW 210 and then putting it back wouldn't resolve the situation. While the issue was going on, I often had an IPSEC VPN tunnel up from one PC that would completely unaffected by the issue, while nothing else could get a connection.
Eventually, I thought, maybe there's something in the Dual WAN feature still affecting this. I re-enabled Dual WAN to look at the settings and I found that the failover test settings were still there, with the change that auto-reconnect was now unchecked. I disabled failover testing completely (which I would naively have thought turning off Dual WAN should have done) and then turned Dual WAN back off. My theory is that the auto-failover feature did not get implicitly turned off with the Dual WAN feature, finding an issue merely disabled new connections on the primary WAN, and since the auto-reconnect was disabled, there was no way for it to recover.
I have run without issue for over a week now (and prior to removing the auto failover test, 2 days was the max I would go). This sounds like a bug, but I have no idea how/where I could/should report this. Clearly, should I want to activate auto-failover, I need to find a better test than what I was using before.
