My internet domain is connected to my router’s IP address (dyndns), the router forwards ports 80, 443, 4443 (all tcp) and 10000 (udp) to my server.
The server runs a fresh jitsi-meet docker installation 6173 - edited .env to use letsencrypt, docker-host is my domain, url is a subdomain, generated passwords, docker-compose-up, added user with password, all is up and running.
The letsencrypt certificate was generated without a problem, both internal (LAN) and external (public internet) clients can access the server via https.
All can join a room and start a meeting when authenticated with user/password.
But as soon as an internal person joins a meeting with an external user the external video goes black.
Apparently the jvb does not process the external video signal correctly. The log shows no errors.
when I enable jvb tcp (JVB_TCP_HARVESTER_DISABLED=false) the external video works, so a wrong forward for UDP looked like a good explanation. But with netcat I could transfer UDP packages from the internet to port 10000 on my server, so udp forwarding as such seems to be ok.
What should I look for next?
If I just call it a day and enable TCP, what side effects should I expect? Server load, resource consumption, video quality …?
Additional info: I had working setups for 5076 and 5142 that I used a couple of times to host conferences. When I start these now there is the same problem. On the other hand a lot was changed in my infrastructure, new OS in both server and router, new docker version, new client os, client browser and so on.