Repository navigation
|
Summary Federated Talk calls (between two Nextcloud instances) fail with ICE FAILED because both Janus and eturnal advertise Docker-internal IPs (172.18.0.x) in their ICE candidates, which Environment
Problem The start.sh script in the talk container sets relay_ipv4_addr to the container's internal Docker IP (hostname -i → e.g. 172.18.0.5). Janus has no nat_1_1_mapping configured For local calls (same server), this works fine because all components (client → TURN → Janus) communicate within the same Docker network or through Docker port mapping. For federated calls, the remote client connects to the local Janus MCU via the signaling server. ICE negotiation fails because:
The remote client has no way to establish a media connection to the local Janus. Reproduction
Analysis In start.sh, the public IP is already resolved: But only the Docker IP is used for relay_ipv4_addr in the generated eturnal.yml. The public IP is only added to whitelist_peers. I confirmed that manually changing relay_ipv4_addr to the public IP and reloading eturnal (eturnalctl reload) fixes the TURN relay candidates — they correctly show the public IP. Suggested fix For federated calls to work, both need to be addressed:
Both values are already available in start.sh via $IPv4_ADDRESS_TALK. This could potentially be gated behind a configuration flag if there are concerns about breaking non-NAT setups, |
Replies: 8 comments 4 replies
|
Hi, thanks a lot for your report and analysis! Would you mind creating a PR that addresses this? Thanks in advance! |
|
Hi, any update here? |
|
I have the same issue when trying to set up an external coturn server (may be needed for certain firewalls that block all but 443 port). |
|
We're having the same issue for anybody trying to connect to our talk conferences, it appears Talk (the AIO image at least) wants host networking so that I can sort of enable host networking in a round-about way on swarm mode, but this can be problematic for multiple instances on the same server. |
|
Hi, I am not sure if it helps but we have the following fix scheduled for AIO v13: #7933 |
|
Still affected with the current aio-talk:latest (pulled today, Janus 1.4.1, eturnal 1.12.3, signaling 2.1.1). Setup: standalone aio-talk container via Docker Compose (bridge network, no host networking), signaling behind a reverse proxy (NPMplus). Media traffic (3478 + relay ports) goes directly to the Docker host. Findings: start.sh in the current image still sets relay_ipv4_addr from hostname -i, i.e. the Docker-internal IP. Workaround (works for me): a small wrapper entrypoint that runs after start.sh and rewrites the generated configs: bash (plus a fixed relay_min_port/relay_max_port range that I publish on the host) Question: Is it intended that federated calls require host networking for the talk container? If not, would a PR be welcome that sets relay_ipv4_addr and nat_1_1_mapping to $IPv4_ADDRESS_TALK when it differs from the container IP (the existing check on line 67 of start.sh already detects the non-NAT case)? |
|
Hello Nextcloud AIO team, I can confirm very similar behavior on a standard Nextcloud AIO v15.0.2 installation running on a UGREEN DXP2800 NAS. We are experiencing problems with federated Nextcloud Talk calls between two Nextcloud AIO instances. Local Talk calls work, and federated signaling/ringing works, but the audio/video connection fails with an ICE error. Our findings:
We have not modified the generated Janus or eturnal configuration. The findings described by @Vprimus seem to match our situation very closely. Could you please clarify whether this behavior is expected in a standard AIO installation and whether there is an officially supported fix that survives AIO updates? We would prefer not to modify generated configuration files or switch to a custom Talk container. Thank you for your work and support! |
|
Hi everyone, lets continue in nextcloud/spreed#20401 (comment) please |
Hi everyone, lets continue in nextcloud/spreed#20401 (comment) please