I see, so your case it is mostly one to many conference. got it. Do you deploy a pool (more than one) Jibri for recording too?
It is a gorgeous monitoring stat you have. May we have your grafana/prometheus script we can apply . Awesome
Thank you, Bro.
Yes, pool of 10 jibri VMs on KVM.
This is from munin, I found somewhere basic plugin and updated/adopted it for my needs…
Hi @Damencho. You said that if there are more than 35 people in a conference, the experience probably won’t be good. However, I don’t need video in my conference, so I would only use audio. In that case, do you think it would apply that “more than 35 people” problem, or could I work with the limit of 75 people?
Welcome to the forum, @NicolasChernavsky!
If you’re not using video, you should be fine.
Well many things had changed since March, so the numbers mentioned there should be now different. There were number of performance improvements on the client side and on the server side and we continue with improvements.
For example on meet.jit.si we now run with channel lastN so you will not get more than 20 videos … and I would say that 75 in a conference should be perfectly fine.
Check my testing here, I see the door to the room 1000 users now
Sir, i Setup jvb autoscale in azure when adding new machines some participants audio video stopped, Please find the attached autoscaled jvb logs. And guide jvb.log (2.8 MB) me. how to resolve this issue.
I have a self-hosting jitsi (version 2.0.5076-1), and I want to know if is it possible to do a videoconference with 200 participants without video and audio. What hardware requirements should it have?
If we wanted to do videoconference with audio and video, what options do we have?
I read that adding bridges can increase limits, is there any instructions to do that?
Hi, i think this guide will help you:
From my experience 2 JVBs (8CPU,8G RAM) with Octo setup will do 200 receive participants with ease. One caution here, all participants need powerful clients PC, today I’ve to solve problem that one participant was kicked out of 115 participants conference after few seconds in. Cause was his 100% CPU load and weak 2core-U celeron. Nothing we can do about this.
We’ve been able to support more than 300 simulated participants with jitsi-meet-torture and we’ve hosted 2 events, one with 204 and another one with more than 150 for Barcelona city council.
You can host 200 with 2x8cpu jvb, but we’d suggest having always +1.
PS: there’s some settings that can help you spread load better than default octo settings.
PS: Ui might be tweaked to decrease participant’s load.
Thanks for share this!
Could you elaborate further on what settings in addition to the octo you used?
For both the server and the client…
Hi @damencho is there a mainstream answer on that question ? Is a change on the configuration file enough (is there a tuto or valid example somewhere) ?
As @hatschand @chaosgrid Iwould like tu use another rtmp server than youtube’s (my own nginx on wich I have access to configuration) to broadcast a meeting to larger audiences. At the moment we re-encode the HDMI output of a screen which is not the best solution…
On the same vein is it possible in the same way to play a local video (or a video from an online server (rawfile or webpage with a JS videoplayer)instead of the webcam or in another window.
thanks again for all this work.
This is currently possible.
Thanks @Freddie for the answer.
Please be lenient with a dummy, where can I find the doc or an example or a tuto.
Streaming to any rtmp - [How to] Live stream to other services using Jitsi
Sharing a local video - just use the screen sharing option, it will bring up a window with 3 tabs from where you’ll be able to select the application you want to share.
The channelLastN feature is based on the list of dominant speakers, so I would like to know, if channelLastN=10 option will work if I set disableAudiolLevels=true as recommended elsewhere. If no, which one of the two is recommended?
disableAudiolLevels is a client-side setting, I’m not sure it improves performance as since this was discussed we fixed some issues and improved performance of the audio levels.
On the other question, the dominant speaker is detected by the bridge and signaled by it so, client settings do not affect that.
disableAudioLevels would relieve server-side load. Is this not accurate anymore?