We’re using Jitsi Videobridge and JiCoFo, with our own client. We’re seeing Track Mute events frequently and randomly. These events are usually followed by an Track Unmute after some 2 - 50 seconds. In the meantime, the stream stops flowing to the client. If two clients are receiving the same stream, one may see this behaviour while the other sees an uninterrupted video stream.
Wireshark shows us that Mute coincides with a drop in received bandwidth. The client receiving the interrupted stream responds with PLIs/NACKs. We see this behaviour with H.264 and VP8 codecs. Streams from Jitsi Meet clients in the conference also behave in this way when our client is receiving them. We’ve compared our client against the Jitsi Meet client using WebRTC Internals and we do not see this behaviour when all clients are Jitsi Meet clients.
Our conclusion is that there is a problem with our stream setup, and that perhaps codec control messages are not being issued or interpreted as they should be. We’ve checked our SDPs very carefully, however, to make sure that PLI/NACK/FIR is signalled, and that the RTX payload is correctly associated with the active video codec payload.
We’re including the NPM package webrtc-adapter, and this is loading correctly. We’re transpiling TypeScript using ts-loader, and testing with Chrome V71. We’re using Jitsi-videobridge V1093-1.
Do you have any suggestions as to where we should be looking for the source of this issue?