Maximum number of participants on a meeting on meet.jit.si server

A so the green was indeed only prosody … Are you running some modifications on top of jitsi-meet?

I cannot stop being puzzled by this. One thing I see different, is that we load “smacks” under the virtual host and you load it globally.
We load it under the main virtual host and the recorder one. And for the recorder we just put smacks_max_hibernated_sessions = 1000; as there are many recorders using the same jid.

Maybe that’s a difference and as we dropped the component connection and now traffic to jicofo goes through the client connection on auth.domain
I’m totally shooting in the dark …

One another small thing you can use is load from the prosody community modules “secure_interfaces”, load it in the global part of the prosody config, after coping it in /usr/lib/prosody/modules. This will disable the TLS encryption between prosody and jicofo, as it is on localhost.

I think it’s 4… or maybe I’m missing something?

Hi. Yes, but there are 2 CPUx4Core=8Core/16 threads. Here is overall CPU usage:

Thank you!

@damencho thank you for this hint, I’ll try to change it accordingly you suggested and will report back.

Thank you very much!

Only logo exchange and css tweak to show it correctly (increase 74px to 120px). That should not harm it?

Yep.

Hi @damencho, I’ve updated our configurations. I’ve sent you updated configs to email. I would appreciate If you find the time to take a look at it, if it is OK that way.

Thank you for support!

Hi @damencho we had strange issue today. All jibri servers stops recording. We had important presentation at 13 00 and wanted to record it. There was 101 people in room (Octo split room onto 2 JVB) moderator wanted to start recording but after she hit “record” button whole UI becomes unresponsive and she was disconnected from room. Prosody was 100% on CPU and she was unable to get into room again. ( I saved my dev console log if interested) I was in room too I saw she is stalled so I started recording on my PC it looked like it started OK but after cca. 25 minutes of fake recording error message came out that recording can’t be started. I looked into it. There is this error on Jibri VM:

2021-05-04 13:10:00.017 INFO: [17] org.jitsi.jibri.selenium.status_checks.MediaReceivedStatusCheck.log() Jibri client receive bitrates: {audio={download=43, upload=0}, download=1335, upload=0, video={download=1169, upload=0}}, num participants: 96, numMutedParticipants: 92, numJigasis: 0, all clients muted? false
2021-05-04 13:10:14.912 FINE: [17] org.jitsi.jibri.util.ProcessStatePublisher.log() Process ffmpeg hasn’t written in 2 seconds, publishing periodic update
2021-05-04 13:10:14.914 INFO: [3190] org.jitsi.jibri.capture.ffmpeg.FfmpegCapturer.log() Ffmpeg quit abruptly. Last output line: frame=10702 fps= 28 q=22.0 size= 85760kB time=00:05:56.70 bitrate=1969.5kbits/s speed=0.926x
2021-05-04 13:10:14.919 INFO: [3190] org.jitsi.jibri.capture.ffmpeg.FfmpegCapturer.log() Ffmpeg capturer transitioning from state Running to Error: QuitUnexpectedly SESSION frame=10702 fps= 28 q=22.0 size= 85760kB time=00:05:56.70 bitrate=1969.5kbits/s speed=0.926x
2021-05-04 13:10:14.920 INFO: [3190] org.jitsi.jibri.service.impl.FileRecordingJibriService.log() File recording service transitioning from state Running to Error: QuitUnexpectedly SESSION frame=10702 fps= 28 q=22.0 size= 85760kB time=00:05:56.70 bitrate=1969.5kbits/s speed=0.926x
2021-05-04 13:10:14.921 INFO: [3190] org.jitsi.jibri.api.xmpp.XmppApi.log() Current service had an error Error: QuitUnexpectedly SESSION frame=10702 fps= 28 q=22.0 size= 85760kB time=00:05:56.70 bitrate=1969.5kbits/s speed=0.926x , sending error iq

Then recording reties on second Jibri VM with the same result and so on 5 times. In dropbox are uploaded files, but it is only garbage in.

Recording was working fine this morning, we recorded one presentation at 9 00 but after this problem at 13 pm are all Jibri VM unusable and recording cant be started.

Are you interested in Jibri logs or Jicofo log?

Thank you for your help an patience!

How many people were with video? Sounds to me like a resource problem and ffmpeg crashed…

Hi @damencho, there were one presentation with video and 4 other participants with video enabled, but maybe there was more, but last-n is set to 5 from 90 participants in room.

Strange is that from this moment on I was not able to record any presentation even with two participants.

What resources should I check?
EDIT: you are right, MEM was running out, but 16GB should be OK? Or?

Thank you very much!

16GB on what - JMS or Jibri? You’re not running Jibri in containers, are you?

I’m running Jibris in VMs with 8CPU/16G RAM and now is eating all memory during recording and crash :frowning:

Is jibri VM on the same machine with JMS?

No, Jibris are only VM on that host.

So are you saying you have multiple Jibri VMs running on that one host machine with 8CPU/16GB or EACH Jibri is on a separate server with 8CPU/16GB?

I meant are Jibris and JMS on the same physical machine? Do they share the same CPU?

:slight_smile: no, signaling host is separate HW +13JVBs + 25 Jibri VMs.

You’re not being clear with your answers; it’s a really simple question - are all the Jibri VMs on the same machine? Yes or No?

Sorry for that. :frowning: I didn’t get it. Yes it is KVM virtualization host with 640G RAM and 128C/256T (2xEPYC7002) system. So there should be not the problem, or? It is happening with only one Jibri running on this server alone!