Self hosted JITSI takes 50% of CPU usage after a while


I got a problem, where jitsi takes about 50-100% CPU usage after some time idling around. I searched in this forum and on github for some solutions and found threads with simular description of the problem. Sadly nothing worked suggested in these posts.

I am hosting jitsi on diffrent VM’s with the latest ubuntu LTS and installed all according to the quick installation on github. The problem occures after some time on all my VM’s.

Here a screenshot:

It seems that this is not suposed to happend. After I restart the services it works for around 30-40 minitues and starts to use 50-100% CPU again without any sessions active. Even if we have active sessions it rises. The frontend does not give any erros, as well as in the logs I cannot see anything strange - maybe I’m looking in the wrong logs.

We use this at the moment for our lectures and oral examens at my university. Me and my team are working since last friday on setting up a infrastructure for the next weeks. Maybe someone had a simular issues or can help.

Thanks in advance!

1 Like

I’m afraid that VM is not going to cut it. I’d suggest 4 cores and 8 GB of RAM.

One of the maschines has 8 Cores and 32 GB Ram, and has the exact same issue

Maybe what I have done/found

org.jitsi.videobridge.DISABLE_TCP_HARVESTER=true and see how it goes.

I added thsi in the videobridge config file, but it did not help.
At the same time I tried to disable the healthcheck to see if it is the healthcheck issue (an other post suggested that) - also with no effect (I couldnt disable the healthcheck completly, but the healthcheck isnt the problem).

Okay this behaivior only happens on VM’s with more than one Core. All our one Core Maschines are fine, all multicore maschines freak out.

Any updates on this how to deal with this ?

We just use 1 Core VM’s otherwise they bug out every single time.

I also had the same issue. I have one jitsi-meet server and multiple videobridges. The main meet server was using 100% CPU (8 cores). As your suggestion I scaled it down to 1 core, and its fine now.

I am running the stable Jitsi on a single Debian machine (not multiple videobridges) on 8 cores without experiencing this problem.


I’ve a dedicated machine for testing purposes and worked perfect with Ubuntu 19. Yesterday I’ve upgraded to to Ubuntu 20.04 and happens the same; 100% utilization of 1 of the 2 CPUs.

Tried to disable “org.jitsi.videobridge.DISABLE_TCP_HARVESTER=true” as metioned in another post but still the same.

Any Ideas -besides of reinstalling everything- ?

Thank you for your help and for this AWESOME software!!

Kind regards

Interestingly, I see this behavior after upgrading to videobridge2. Currently, the CPU usage over time for our videobridge servers looks like this:

(not allowed to insert >1 image because of new account, will add with next post)

Before the upgrade, the VMs were using the full CPU:

I don’t think that “org.jitsi.videobridge.DISABLE_TCP_HARVESTER=true” is a solution here. This was mentioned as way to save CPU usage, not to “unlock” it.

Can someone with 100% cpu load on one core perform create a java flame graph of that PID and upload the SVG. Such a graph will tell what your core is doing and where to fix it.

Also try disable the timeseries logging of jvb
as it did consume quite a lot of CPU.

I am seeing the same result on a 2 core system. I’ve upgraded a week ago from a one core system which did not have this issue. I tried lowering the logging, but this seem to make no difference.

Below was de usage before the restart of JVB at 14:12. After that it keeps at a 20% load, still too much for no users.

dofl: Can you try profile your server using async-profiler and upload the generated .svg file

async-profiler makes it easy to setup and generate a flame-graph of a runnig java application such as the JVB server.

which version of Java?
Is it dockerized?

I have same problem in AWS ECS dockerized Ubuntu 20.04 and Java 14…
Now I switched to Java 8, I will see few days how it goes