High and Negative RTT values

Hi, community!

Now we have all Jitsi-meet components in docker containers in single CentOS7 VM with

  • 12 core CPU (Intel® Xeon® Gold 6132 CPU @ 2.60GHz),
  • 16Gb of RAM and
  • 10Gbit Network card.

We have a lot of such log entry’s

WARNING: Negative rtt value: -1.0003660000000991 ms, remote processing delay was PT3.541000366S (232063), srSentTime was 2020-06-05T09:06:07.890Z, received time was 2020-06-05T09:06:11.430Z

WARNING: Suspiciously high rtt value: 15643.804688 ms, remote processing delay was PT0.450195312S (29504), srSentTime was 2020-06-05T09:06:15.158Z, received time was 2020-06-05T09:06:31.252Z

WARNING: RTT suspiciously high (1459ms), capping to 1000ms.

It’s only Warnings but tons of them!
So my questions is:

  1. What could the cause of negative RTT?
  2. What could the cause of such high RTT? (10000 and even more)
  3. What can we do to minimize amount of that lot events?

Here is a conferences graph of that week for our deployment.

And CPU graph

Starting today we are on stable-4627-1
Before we was on stable-4548-1

PS. Multiple JVB with OCTO in plans for now.

1 Like

I can recommend the following two client side optimizations that reduces high RTT values:



1 Like

Thanks a lot for your tweeks, I forgot to mention, that most of your recommendations are implemented, these two definitely set.

HI @Anton_Karlan
I also have a docker jitsi server running in centOS with 6 cores,16GB RAM and 2.2 GZ

but I am facing unusual issue with jvb that it is consuming nearly 100% cpu within 6-7 users and also consuming 1-2.5 GB network only for these. can you plz share your opinions that what was in your case?
I also tried docker latest version locally and inspected same issue. here is the thread HIgh CPU usage by JVB in Docker Server