Random video blackouts "to save bandwidth" after update to 1.0.4900

Hey everyone :slight_smile:

i’m back with a new problem and it’s a bit strange :wink:

So i’m hosting a Jitsi instance for my school, such the typical scenario is 1 video stream sent and then received by around 20 students and this in max. 8 conferences.
A few stats on my system:
Ubuntu Server 20.04 LTS
Jitsi-meet 2.0.5765-1
Jitsi-meet-web-config 1.0.4900
Jitsi-videobridge 2.1-478 (on the same system)

12 Cores
16 GB RAM
1 Gbit/s network up and down

The problem started occurring after i updated from 1.0.4628 to 1.0.4900.

A few students reported that they were experiencing blackouts in the video streams with the message “Video for XYZ has been turned off to save bandwidth”.
I immediately switched back to the previous version and the problem was gone so i guess there were some changes made to the bandwidth estimation or allocation logic.
Sadly these changes prevent my students (with sufficient bandwidth at hand (checked with internal speed test to the same server)) from participating in those meetings.

Do you have any idea what could cause this behavior or even better a solution for it?

Any help would be appreciated!

Thanks in advance and if you have any questions, please ask!

john2151

I’m experiencing the same error with my setup too after some recent upgrade. Any idea how to debug or fix this?

Can you try setting in your config.js useNewBandwidthAllocationStrategy: true, and report do you see a difference?

Thanks for the reply @damencho!

Sadly i have to report, that this doesn’t improve the situation, but just to be clear: It is just necessary to uncomment and set this value to true in /usr/share/jitsi-meet-web-config/config.js right?

Nope this is in /etc/jitsi/meet/yourdomain-config.js

Oh, ok sorry i got that wrong.
So i now added this to the main config file and as far as i can see with my synthetic test the problem is gone, but i will have to test it in a real world scenario.

Will report tomorrow with hopefully good news :slight_smile:

So sadly the problem is still not solved by this …

I already received a handful of complains in the first few minutes so …

Could this possibly be related to bridge websockets as commented at Flagship VP9 Support: Notes and Discussion - #214 by garloff

We suspect to be this Fix: fix bug calculating removed video endpoints in constraints. (#1643) · jitsi/jitsi-videobridge@80d06a6 · GitHub
But to try it out you need to update to latest from unstable.
We will push new stable with these changes, hopefully soon.

Ok, thanks for letting me know.
Since this is a productive environment i will wait until the next stable release and then report if it improved the problem.

Thanks again!

john2151

@damencho, if you have a binary build for Debian (even not packaged as a .deb), I’ll be happy to test this and provide feedback.

You can update from testing Index of /testing/, probably will release that to stable today.

I confirm that this new release has resolved the issue for us. Thank you for the quick fix!

1 Like