How to troubleshoot general quality issues? (participant dropping video, NON OPTIMAL connection warnings, etc)

i’ve got my server set up and i’ve successfully done a few calls with my family (3-5 participants) so i know everything is set up correctly. (latest debian packages on a buster server, 443/tcp/udp & 10000/udp are open. ssl certs are set up properly)

i did a 2 hour call yesterday and in general, everything works. there were 3 or 4 instances where participant’s video feed went blank/black for a few seconds, and sometimes i could see their network indicator as POOR or NON-OPTIMAL, even though round trip time is less than 100ms and no dropped packets. video bitrate is also kept very low for some reason. i don’t think anyone was sending more than 640x360.

everyone is on at least a 20Mbps line, and my server is a VPS with gigabit connection. i was monitoring my server throughout the call and i never saw either CPU or RAM being maxed out (1vCPU, 1GB RAM). everyone using the latest chrome and one person using jitsi-android

i’m sorry these issues are vague, but there really isn’t any other way to describe it other than intermittent and temporary performance issues. is there anything i need to look at on my server? or is it just the nature of the endpoint computer (bad wifi, CPU maxed, etc)?

overall, jitsi has worked very well. thanks to everyone who contributes!

I am getting a similar experience. Our VM is 6 core’s, 8GB RAM, and testing is between two user from remote locations.

No error messages in jvb.log except a healthcheck will fail from time to time, on timeout. PT3

I also turned down the audio monitor that would identify the active mic per other online recommendations.

Not sure where I should take my troubleshooting at this point.

Thank you for your thoughts team.

when you only have 2 users, i don’t think anything goes through the JVB. only peer-ro-peer connection when 2 users. you need to test with at least 3

So are you saying that the issue would then be only with the end users in this case given p2p?

Thank you,

that is my understanding of the system. during your session you can hover your mouse over the connection icon and see if it’s p2p or connected to your server on 10000/udp

The bridge does make available a bunch of stats via REST. We query those and push the data to datadog so we can monitor things at a glance. If you increase the log levels, the Endpoints will also dump all of the stats of their pipeline when they expire (this data is also available at any time via REST).

That’s ADSL yes ? with the most advanced version it means at best 25Mbits/download and 4Mbits/upload. In more common versions it means 20Mbits/download and 1Mbits/upload. If the client computer is far from the DSLAM, it can be even much less, that is 5 Mbits/download and 512kbits/upload. For video conferencing the upload speed is very much the limiting factor.
Given that the Jitsi encoding is not the most advanced, and that quite often computers can decide to access the network on their own (think software updates), it don’t get you far, certainly not high resolution video in most common cases.

i think only one participant is on 20mbit ADSL. the rest of us are at least 50/50 fiber to the home or at least to the neighborhood

i’ll double check what the upload speeds are

thanks

I was able to do some more testing and it shows that I am using port 10000 when in chat with only one other user. So this would indicate that JVB is being used even when only between two users?

Still having intermittent quality issues making the product unusable.

So I am back to my original question.

Not sure where I should take my troubleshooting at this point.

Thank you for your thoughts team.

I managed to see some statistics at “example.com/colibri/stats”. But you need to enable rest and colibri at Videobridge config. And monitor cpu/ram/bandwidth usage. May give you an idea about the problem.