Multi network suppor?

[airsay]

Does this support multiple networks on the host system? I am thinking of rolling tis out on a system with a VPN interface and a regular/Public internet interface

[dariusstefan]

This is not supported out of the box, but it should be achievable with some changes. The following is a high-level idea of what would be needed, not a tested solution.

SIP plane

OpenSIPS would need two socket declarations, one per network, each advertising the correct external IP. Since Docker NATs all traffic to a single container IP, the two networks can be differentiated by mapping both to different internal ports while keeping port 5060 on both external interfaces:

docker-compose.yml

ports:
  - "${HOST_IP_PUBLIC}:5060:5060/udp"
  - "${HOST_IP_VPN}:5060:5061/udp"

opensips.cfg

socket=udp:${OPENSIPS_IP}:5060 as ${HOST_IP_PUBLIC}
socket=udp:${OPENSIPS_IP}:5061 as ${HOST_IP_VPN}

Registration socket per contact is already stored in the location table by usrloc, so OpenSIPS would automatically use the correct socket when reaching a registered client — no additional routing logic needed here.

RTP plane

RTPProxy can stay as-is on the existing Docker network. Docker already forwards the RTP port range from all host interfaces to the container, so RTP from both networks reaches RTPProxy regardless.

The only change needed is in branch_route — RTPProxy needs to advertise the correct external IP in the SDP for each leg, driven by $socket_in(advertised_ip) and $socket_out(advertised_ip):

if ($socket_in(advertised_ip) == "${HOST_IP_VPN}")
    $rtp_relay_peer(ip) = "${HOST_IP_VPN}";

if ($socket_out(advertised_ip) == "${HOST_IP_VPN}")
    $rtp_relay(ip) = "${HOST_IP_VPN}";

RTPProxy's -A startup flag covers the public side as a default; only VPN-side legs need an explicit override. Outbound RTP from RTPProxy is handled automatically by the host routing table. Cross-network calls (public caller → VPN callee and vice versa) are handled correctly.