Turn server stop replying to binding requests and cut call after 10 min 4.5.2

[Greg]

Good day my current setup is following

WEBRTC OverTCP  <> Turn1 <UDP> Turn2 <UDP>WebrtcServerGateway

Turn 1 = 192.168.224.38

Turn 2 = 192.168.224.142

The session establishes fine, and is going ok for 10 minutes

1: Start of the session:

One time in 3 segonds i see turn 1 send binding request to turn 2 and gets a reply:

But after 10 minutes turn2 stop replying and the voice cuts, however turn1 keep sending binding requests az crazy every 100ms or so

Turn 2 receives them, but do not answer, if i enable option 'channel-lifetime' for more than 600, the call will live, however as this is not recommended i wil not.

In the logs i see nothing at all, the only idea i have is that the session is not refreshing, 

Also turn2 have option server-relay enabled.

The only other idea i have is that until the end of the call before being cut, my web application is sending constantly on the turn2 server the turn1 XOR-Peer-address

Any idea how to make in sort that the session keeps alive without using channel-lifetime ?

What is the main possible cause that channel-lifetime fixes ?

Thank you

Greg

[Gustavo Garcia]

Is the WebrtcServerGateway maintaining the TURN connection to Turn2 open?   What logs/errors do you get in that side?

[Greg]

Hello, thank you for your reply the  WebrtcServerGateway is a Janus which is configured to use the TURN2 as turn (as he is behind nat)

Turn2 has two IP interfaces

1: 192.168.224.142

2: 10.233.12.183

Janus has one interface:

1: 10.233.116.218

It receives the traffic from TURN1 to 192.168.224.142, it relays traffic to Janus from 10.233.116.218

Janus is at 10.233.116.218 so the communication is between turn2 and janus is between 10.233.12.183 <> 10.233.116.218

Turn2 gives the candidates which are the ones he has, his different IP's and turn1 IP, at the end good candidate is found at 10.233.116.218:23839 (janus) and 192.168.224.142 (first ip of the turn2 server) however the traffic goes to a correct ip 10.233.12.183 (second ip of the turn2 server)

In exactly 9 minutes there is a refresh coming from Janus to Turn2

Then after 1 minute, around time which is call is going to be cut, i see the binding request reply which is basically a result of the spam from turn1 <> to turn2 server binding requests.

After that the call is cut, in logs there is no error, i have a "stale nonce" errors but i have removed that by adding the corresponding option to make in sort that nonce lives 2 hours.

I have also monitored the candidates of the each session:

So the full chain of the interaction would be

CLIENT <PUBLIC IP .200> TCP <>TURN1 <PUBLIC IP .133>TURN1<PRIVATEIP 192.168.224.38> TURN2 <PRIVATEIP 192.168.224.142> TURN2 <PRIVATEIP 10.233.12.183>JANUS <PRIVATEIP 10.233.116.218>

пн, 12 сент. 2022 г. в 19:33, Gustavo Garcia <gust...@gmail.com>:

--
You received this message because you are subscribed to a topic in the Google Groups "TURN Server (Open-Source project)" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/turn-server-project-rfc5766-turn-server/cgiTsP72zU0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to turn-server-project-rfc57...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/turn-server-project-rfc5766-turn-server/7aee4c3c-81ab-480b-9a0d-90b71bfd3bccn%40googlegroups.com.