And I managed to make it work (P2P, TURN and JVB works depending on the situation).
In P2P mode with 2 participants, everything works as expected.
In TURN or JVB mode (depending if ports are opened or not), the video nearly instantly cuts when switching from P2P mode to another mode. I say instantly because I still have a single frame displayed in the new participant window joining the room, then video cuts off for everyone.
It should be something related to performance settings because I still have audio so it still communicates in the channel but I can’t manage to find what.
In server logs, JVB says the usual Selected pair, ICE connected, Pair succeeded etc… No errors on this side
In browser console it is the same, I have nothing related to stopping the video transmission because of some quality settings, it says adding and creating the remote track, ICE connected JVB (in JVB mode), send ICE candidate, data channel opened, sending constrainsts and nothing more like it usually does.
Do you have any idea on how to troubleshoot this issue ?
Yes it goes through JVB now with 2 participants, I do have audio and I don’t have video. Except sometimes some frames at the begining so the video is able to go through the channel but something prevents it
But this communication is still to send messages between the client and the bridge to establish the communication right ? Not the audio/video stream itself.
And I can see in videobridge that the communication is established, I have the audio and as I said before sometimes we see some frames goes through the video channel so I’m pretty sure this is not related
But that communication is very important requesting the videos to be delivered, the resolution to be sent etc. For example, when that communication is broken there is no screensharing.
Sure and that’s why we removed it so (I don’t remember the details) it uses prosody to communicate with the bridge and it works well like that. It’s maybe less performant I saw some entries about that in forums but it works in our environment (which is a bit specific I have to admit).
Anyway if the .js conf doesn’t work I will try to enable again websockets
Yes we are using subfolders and already resolved the related issues, there is still 2 404/CORS errors which are related to language and manifest.json, I can solve it easily but it always has been like that.
Well I’m running into an infinite “pending” wss request with an nginx 499 error with websockets.
I won’t have time to inspect more unfortunatly so I will stay on the old version in production and wait to see if future updates of videobridge fixes the issue or try to inspect more on this later in january.
