We have a self-hosted Mattermost instance at our workplace, and I’ve been trying to set up the Calls plugin. I’ve got it working on our internal networks, but not remotely.
Our instance is on Mattermost version 9.5.6, running in Docker using nginx, and we have another reverse proxy in front of the host the service runs on. In order to get around having to route UDP traffic through the reverse proxy, we have a public IP that forwards all TCP and UDP traffic on port 8443 straight to the host, where said ports are exposed from the Mattermost container. This public IP is set as the ICE host override in the Calls plugin settings.
Trying to start a call on a device connected to the workplace network or VPN works perfectly fine, and running a tcpdump shows constant requests back and forth between the Mattermost container and the client device. However, on a device outside the network, the call times out after a few seconds. Checking the container logs gives me a lot of this following error:
{"timestamp":"2024-06-25 15:06:27.585 +02:00","level":"info","msg":"Failed to send packet: write udp4 127.0.0.1:8443->x.x.x.x:40201: sendto: invalid argument","caller":"app/plugin_api.go:1000","plugin_id":"com.mattermost.calls","origin":"main.(*logger).Info log.go:104","origin":"ice/v2.(*candidateBase).writeTo github.com/pion/ice/v2@v2.3.24/candidate_base.go:340"}
{"timestamp":"2024-06-25 15:06:27.586 +02:00","level":"info","msg":"Failed to send packet: write udp4 127.0.0.1:8443->y.y.y.y:61186: sendto: invalid argument","caller":"app/plugin_api.go:1000","plugin_id":"com.mattermost.calls","origin":"main.(*logger).Info log.go:104","origin":"ice/v2.(*candidateBase).writeTo github.com/pion/ice/v2@v2.3.24/candidate_base.go:340"}
{"timestamp":"2024-06-25 15:06:27.788 +02:00","level":"info","msg":"Failed to send packet: write udp4 127.0.0.1:8443->x.x.x.x:40201: sendto: invalid argument","caller":"app/plugin_api.go:1000","plugin_id":"com.mattermost.calls","origin":"main.(*logger).Info log.go:104","origin":"ice/v2.(*candidateBase).writeTo github.com/pion/ice/v2@v2.3.24/candidate_base.go:340"}
{"timestamp":"2024-06-25 15:06:27.789 +02:00","level":"info","msg":"Failed to send packet: write udp4 127.0.0.1:8443->y.y.y.y:61186: sendto: invalid argument","caller":"app/plugin_api.go:1000","plugin_id":"com.mattermost.calls","origin":"main.(*logger).Info log.go:104","origin":"ice/v2.(*candidateBase).writeTo github.com/pion/ice/v2@v2.3.24/candidate_base.go:340"}
{"timestamp":"2024-06-25 15:06:27.991 +02:00","level":"info","msg":"Failed to send packet: write udp4 127.0.0.1:8443->x.x.x.x:40201: sendto: invalid argument","caller":"app/plugin_api.go:1000","plugin_id":"com.mattermost.calls","origin":"main.(*logger).Info log.go:104","origin":"ice/v2.(*candidateBase).writeTo github.com/pion/ice/v2@v2.3.24/candidate_base.go:340"}
{"timestamp":"2024-06-25 15:06:27.991 +02:00","level":"info","msg":"Failed to send packet: write udp4 127.0.0.1:8443->y.y.y.y:61186: sendto: invalid argument","caller":"app/plugin_api.go:1000","plugin_id":"com.mattermost.calls","origin":"main.(*logger).Info log.go:104","origin":"ice/v2.(*candidateBase).writeTo github.com/pion/ice/v2@v2.3.24/candidate_base.go:340"}
Running a tcpdump gave much the same results, repeated requests from the Mattermost container to both my local and public IP addresses (replaced with x.x.x.x and y.y.y.y here for anonymity).
From this I’ve gathered that, when starting a call, the Calls plugin tries to contact the client, instead of the other way around. What I’m really wondering here is if this is intended behavior or not. If so, this makes it unnecessarily difficult for our employees who work remotely to make use of this feature. We’d really like to not force them to use a VPN or change the settings on their private network setup to forward this kind of traffic to their devices.
Regards,
Thomas
