Paid support - Fix issue #11888 (Connection close after 40 seconds)

Hi everyone,

As last hope before thinking of drastic changes, we would like to ask for paid support for issue #11888 (Coturn fallback to JVB closes the connexion after 30-40 sec and restart · Issue #11888 · jitsi/jitsi-meet · GitHub).

This issue is currently the last impactful issue on out users experience and we probably won’t be able to solve it.

We can also give access to some of our internal ressources with deeper dig on the issue, the major part is described in the GitHub ticket.

Thank you in advance !

I’ve posted in the GitHub issue — I think this might be due to ICE restarts being disabled by default in Jitsi (enableIceRestart) due to a bug that causes them to not work properly. So when you change the network conditions during the call, the only option is to reconnect from scratch, rather than do the “seamless” ICE restart.

This assumes that you are closing direct access to udp/10000 from the client while the call is underway — it’s not clear to me from the GitHub issue if the problem also occurs if you start the call with access to udp/10000 from the client already closed, so that TURN is used from the beginning.

BTW, posting questions like this here on the community rather than the GitHub issues can get wider visibility for them from the start, until/unless they’re known to be a bug in Jitsi itself.

Hey @jbg,

Yes I already tried the enableIceRestart as I told you in GitHub but without success. Anyway in Jitsi SaaS I don’t think the connexion restart every 40 seconds. It that was the case, we would see micro cuts when it reconnects right ? Anyway the option doesn’t works and for slow connexions, may not be the solution.

For the conditions, I don’t assume changing the network conditions during the call, I only start fresh calls with 10000/UDP open or closed client side. I assume my customers don’t modify their firewall settings in production during daywork haha. So yes it is TURN from the beginning.

BTW, posting questions like this here on the community rather than the GitHub issues can get wider visibility for them from the start, until/unless they’re known to be a bug in Jitsi itself.

Yes I know I already got scolded for that :stuck_out_tongue:

Hi @Jerome_LEMAN_GARIN ,
You are only observing this reconnection behavior on your own platform or also on meet.ji.si ?
And what is your JItsi-Meet version ?

@Damien_FETIS

You are only observing this reconnection behavior on your own platform or also on meet.ji.si ?

No only on my platform with my particular environment (CentOS 7, nginx in front of everything, only 443, multiple applications)

And what is your JItsi-Meet version ?

[root@gofast-comm synapse]# rpm -qa | grep jitsi
jitsi-videobridge2-2.2_63_g252d14bc-2.noarch
jitsi-meet-web-config-1.0.6850-2.noarch
jitsi-meet-tokens-1.0.6850-2.noarch
jitsi-meet-web-1.0.6850-2.noarch

But we have this issue for more than a year and perform regular Jitsi update to keep up to date (as soon as our production constraints allows us to do it safely)