Install Help - 90% Cpu, 4 core 8Gb Ubuntu

Hoping the group would be able to offer some insight to what I am seeing. I just did a test session with ~30 users and 50% of them using webcams.

I was watching the CPU usage on my Hyper-V Ubuntu install and it hit 90% and started to drop the connections.

I am not a Linux admin so I suspect this is all my config.

The guest server has 4 cpus with 8Gb ram. My public IP is a full 1GB U/D (session only used 50 Mbps at most). I also checked that AES is on the virtual CPU.

The install is straight from the guide with just one change in that I disabled the option to show the audio indicators of people speaking.

Am I expecting too much from this fantastic product?

Any ideas would be greatly appreciated.

1 Like

Well, how did you find out that the Cpu use was 90% then ? was it with top ?

Thank you for replying. I am a total newb on Linux.
I used a SSH program called MobaXterm that shows a summary of CPU, ram ect.
I am going to do another test using HTOP. However, I do think the CPU usage reported was somewhat accurate as I could see it increase as users joined the session. Then, as it got to the 90% range, we would hear the audio start to kick out and then people start to drop. The CPU would then go down to about 10% and start to rejoin people and slowly increase back up.

Perhaps I am asking to much from this program? The VM is hosted on a Microsoft Hyper-V. Again, I just hoped that 4 cores and 8 GB of ram would have been enough.

I am going to check the logs and see if I can see anything but any advice on troubleshooting or just saying that it is too much for Jitsi, let me know please.

Unfortunately I’m not familiar with your software. The usual trouble with top is that it is reporting use for one CPU so people are not reading the output correctly. You could try to set channelLastN to 3 as recommended to limit the load for the system. Check also that your Vps is not suffering from noisy neighbours problem, many cheap vps have shared Cpus. When you factor this, dedicated hardware seem less expensive.

Ok…last question…where is that setting for channelLastN? and…would you happen to have a link to a doc on that so I could read.

Thanks so much

this file should be in your install under /etc/jitsi/jitsi-meet (it’s renamed by prefixing it with your site name)

Thanks…I actually found it before you replied and made the change. However, no real difference.

It appears that there are numerous users struggling with high CPU in JVB. I suspect that it just needs a little more work and documentation to help users get it setup.

But thank you so much for trying to help.

Hi there, I am experiencing a similar problem. I’m running Jitsi on a virtual root server (2 cores, 4 GB RAM, Ubuntu 18.04 LTS) hosted by linevast. The server was shut down yesterday by the user support due to repeatedly exceeding the maximum CPU load. I can’t see the logs right now, because of the shutdown, but the server was probably not experiencing a big number of parallel calls (only about 20 people know about the server).
Does anyone have any idea how to change the config to prevent this behaviour from happening?

A recent update may be the cause.

I don’t know if this has been solved yet.

Please see :

Stumbled recently upon this thread, have a look at it, if this is maybe the problem?

I don’t know but if it’s the case the remedy is real easy, just update to the current version.

Can you try tune or disable swap on the server? By analysing flame graphs of one server with high CPU i notice that much time was spent swapping and waiting for disk writes when writing the log files.