OOM after ~1.5 days on aws instance

I’m currently running a jitsi node on an aws t2-micro instance.

The odd issue I’m having right now is that the node will run for ~1.5 days before becomming unresponsive. This is without anyone connecting to it during that period.

The exact error I’m seeing is:

meet login: [206420.705830] Out of memory: Kill process 880 (java) score 579 or sacrifice child
[206420.714299] Killed process 880 (java) total-vm:5330284kB, anon-rss:582436kB, file-rss:0kB, shmem-rss:0kB

This was set up via the quick start guide for ubuntu.

video bridge expect 3gb of RAM and will run out of memory on 2gb systems.

you can lower the VIDEOBRIDGE_MAX_MEMORY=3072m setting in
/usr/share/jitsi-videobridge/jvb.sh
to make it work on systems with only 2gb.

Is there any way to make it work on 0.5Gb systems (see t2.micro free tier aws)?

On a system with only 512mb i would first try set
* soft stack 256
in /etc/security/limits.conf
it makes pthread’s use 256kb stack by default instead of 8Mb for each thread and that would free up a lot of RAM to begin with.

After that set
VIDEOBRIDGE_MAX_MEMORY=400m

And optionally try change the garbage collector for the JVM that run videobridge to use shenandoah-gc

1 Like

Awesome thanks!! I’ll let you know how I get on.

I have 1 GB Ram and having same issue so this
VIDEOBRIDGE_MAX_MEMORY=1024 setting in
/usr/share/jitsi-videobridge/jvb.sh

will work ?