Jitsi Meet Load Balance - performance and resources

Install & Config
1

Hi all,
I am a newbie of Jitsi Meet and this is the first time I install and use it.
I want to use it with load balance, so I watched clip : How to Load Balance Jitsi Meet
Link : https://www.youtube.com/watch?v=LyGV4uW8km8&t=493s
Now, I have two servers, named : server1 (install jitsimeet, prosody, jitcofo, videobridge), and server2 (install only videobridge) as video guideline. Two server both have 4GB Ram, 2 vCPUs (I use for testing load balance).
When I create call with 6 conccurent users, I checked in both server and have following parametters:
I saw that server1, jvb process used very few resources (ram, cpu), but server2, jvb used lots.
Surely many people have used jitsi with load balance, so can tell me if it makes sense and can give me advice when using load balance. Thank you very much!
=========server1===========
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3913 jicofo 20 0 5941560 188116 28608 S 0.7 4.7 0:09.32 java
4161 jvb 20 0 5859104 183520 29372 S 0.7 4.5 0:12.78 java
2495 prosody 20 0 64840 20548 5160 S 0.3 0.5 0:03.46 lua5.1
1 root 20 0 160064 9280 6700 S 0.0 0.2 0:03.20 systemd

=========server2===========
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
17295 jvb 20 0 6148992 403812 31108 S 92.4 10.0 9:57.22 java

Then, I create more calls, 4 calls with 2 or 3 concurrent users per call. but only one server receiving traffic and jvb progressively increasing the percentage of CPU processing, I call that is server 2.
But when I stop jvb process in server2, all traffic and process of these 4 calls move to server 1, and it still working stably. similar if kill jvb on server 1 and turn on jvb on server 2.
So I have one question about load balance of jitsi:
Are every call equally divided among multiple pins? Or is this only available when the server has problems and is no longer usable?

2

1 call is served by one jvb by default.

3

Hi, thanks for your answer.

I created 3 call (2 concurrent users per call) and result is same.
I tail log of job process in server1, it only:
"OpenJDK 64-Bit Server VM warning: Option UseConcMarkSweepGC was deprecated in version 9.0 and will likely be removed in a future release.

SCTP JNI load: Linux OS detected

SCTP lib loaded

=====>: org_jitsi_modified_sctp4j_SctpJni.c calling init

=====>: org_jitsi_modified_sctp4j_SctpJni.c about to set SCTP_DEBUG_ALL"
and no more new logs when there are new calls.

4

Hi, I fixed it and know it seem to be work. But I still have a question

  1. I create more calls, 4 calls with 2 or 3 concurrent users per call. but only one server receiving traffic and jvb progressively increasing the percentage of CPU processing, I call that is server 2.
    But when I stop jvb process in server2, all traffic and process of these 4 calls move to server 1, and it still working stably. similar if kill jvb on server 1 and turn on jvb on server 2.
    So I have one question about load balance of jitsi:
    Are every call equally divided among multiple pins? Or is this only available when the server has problems and is no longer usable?

  2. In all of jvb process in 2 server, It always show log “Error received when publishing to the node: sharedStatsNode. Message: null. Condition: forbidden” when I start videobridge. I think it is an error, but I see the system still working and calling normally. Do you know what error I encountered when configuring load balancing?

Thanks and best regard!