Jicofo&Octo : max particpant to one conf very small

We have a setup with Octo enabled, and jicofo is setup with
org.jitsi.jicofo.BridgeSelector.BRIDGE_SELECTION_STRATEGY=RegionBasedBridgeSelectionStrategy

We have only 1 region, with multiple JVB’s (that auto scale).

With such a conf we want to know the max participant in ONE room.
Around 50 users the service start falling, bad sound quality, freeze …
We saw that only 1 JVB was used, and Jicofo stress value is around 0.5.
The load (cpu/ram/network) is OK on all component (including the jvb), so I can’t explain such a result.
I wonder why not balance customers to another JVB (stress value to few I guess), and also JVB can handle more users (test more 200 users in multiple conf on same jvb is ok), but here we can’t reach such value.

On jicofo I have many

2021-02-05T14:11:44.734Z Jicofo 2021-02-05 15:11:44.734 WARNING: [29] org.jitsi.jicofo.JitsiMeetConferenceImpl.log() No jingle session yet for roomtest@muc.jitsi-dev/116430f9

On JVB I have many :

2021-02-05T14:43:58.811Z WARNING: TCC packet contained received sequence numbers: 5403-5529, 6034-6040. Couldn’t find packet detail for the seq nums: 5403-5525. Latest seqNum was 6525, size is 1000. Latest RTT is 1848.006592 ms.

Nothing else on other component.

If someone can help

1 Like

This RTT is huge. The network maybe overloaded …