element-hq
Latency can increase to ~10 seconds
Steps to reproduce
Reports of latency on a call getting up to 5 or 10 seconds, at which point obviously the call is unusable. It apparently starts off okay and then gets worse. That's all we have currently. Perhaps we have some buffers that are way too large in livekit?
Outcome
What did you expect?
What happened instead?
Operating system
No response
Browser information
No response
URL for webapp
No response
Will you send logs?
No
I have had this issue a few times, raising just in case you want someone who may be able replicate this
I've noticed while debugging device issues that unplugging a device and falling back to a different one seems to increase latency significantly.
Further to the above, the latency did jump up significantly when changing devices (maybe to ~500ms in some cases) but did recover again afterwards. The latency looked to be on the receiver side, looking at the graphs in WebRTC internals, I could see large spikes in the jitter buffer graphs when changing devices. This could imply the livekit server isn't quite doing timestamps correctly when a publisher stops sending packets, perhaps?
presumably this is going to be a livekit bug? (which is ironic given rate control was the main reason to use a preexisting SFU...)
I now keep getting this occasionally. I thought it was when I started using firefox but I'm now getting it on chrome, and the person I'm seeing delayed is on chrome too. The person who is delayed says his internet connection is bad, however that person's video is not delayed for anyone else in the call. This would suggest it's some kind of buffer edge case in Livekit perhaps?
A friend of mine is also having this issue on chrome with a bad internet connection.
closing due being stale for some time. Please reopen if the issue still occurs.
