Hi, i on my urbackupsrv host i am also running the urbackupclientbackend to allow it to backup itself. I noticed both services seem to bind to all ip addresses (both ipv4 and 6), but the moment i start hardening security by setting up a firewall (default policies: outgoing:accept, incoming:deny, fwd:drop) and adding rules for clients to connect and allow the broadcasts to come back in i see they cant reach each other, but other clients connect. Only when i add the same rules i had added on LAN clients to allow the server to broadcast to them and the other needed traffic, the server connects to its locally running client backend.
I’d expect them to bind to the loopback as well, so i checked if there is a config option for them to specify what addresses to bind to, but it seems that is missing.
Why am i doing this: i am hardening the security on the server, setting up fail2ban for the webinterface which i just allowed to be accessed from my home country, as you probably know that includes setting up a local firewall, so i configured that, which went great so far. But as said, i was expecting them to find each other through the loopback address, which would also make sense in terms of performance. I am not sure though how that works with the (server) broadcasts, if such a thing even exists on a loopback address. So i am calling in others to comment on that, possibly the ability to set a binding address could be feature request, depending on what is possible.