This issue is seen in openWRT 22 (kernel 5.10) whereas the transport socket would not properly bind to the interface of the profile's IP address, almost at random, in a multi-homed system with a failover SIP profiles/gateways.
In this case, there are two sets of Sofia profiles/GW: One set for the primary IP/network interface and another one for the failover interface. The latter profiles are usually not active, but when the primary fails, the failover profiles take over.
We noted that often times, the failover traffic would always be directed towards the primary interface, therefore on failover, its profile would be of no use. This was flagged as a failover failure by QA.
No amount of Sofia configuration seems to make any difference. See this post for more details: https://forum.signalwire.community/t/network-layer-binding/481
This issue is seen in openWRT 22 (kernel 5.10) whereas the transport socket would not properly bind to the interface of the profile's IP address, almost at random, in a multi-homed system with a failover SIP profiles/gateways. In this case, there are two sets of Sofia profiles/GW: One set for the primary IP/network interface and another one for the failover interface. The latter profiles are usually not active, but when the primary fails, the failover profiles take over. We noted that often times, the failover traffic would always be directed towards the primary interface, therefore on failover, its profile would be of no use. This was flagged as a failover failure by QA. No amount of Sofia configuration seems to make any difference. See this post for more details: https://forum.signalwire.community/t/network-layer-binding/481