Firefox falls back to TCP w/o reason


I’ve set up jitsi-meet under Debian 10, standard package manager install. nginx on port 80 and 443, jvb on UDP 10000 and TCP 4443.

For Chrome users, everything works as expected, they talk TCP 443 to nginx, then video flows over UDP 10000 or a randomly allocated port to videobridge.

For Firefox users (specifically tested with Firefox 74 on Ubuntu and Mac OS), the video connection constantly drops back to the TCP 4443 fallback port. Same machines on the same network, so definitely no firewall / routing / etc. issue. Also, when disabling the Harvester port completely, only allowing UDP, Firefox happily works with the UDP connection.

Is that a configuration issue? Something easily fixed? While I don’t expect people to be behind restrictive firewalls, I would like to keep the TCP fallback in place.

1 Like

What do you mean by “constantly” here? Does it happen 100% of the time, or just annoyingly frequently, in situations where it’s unnecessary?

Sorry, my mistake, I had to take a closer look:

Chrome only connects through UDP.

Firefox connects through TCP, and also generated some low-bandwidth traffic there, but it also makes a UDP connection, and typically the bulk of the traffic seems to goes there. More complex, but probably also fine.

I missed this the first time, as I only saw the TCP connections being established and used, and didn’t check for the additional UDP connections coming from the same instance. All through NAT from the same IP, so I had to do some digging to trace those back.

This solves it, thanks!

1 Like