Frequent Disconnection when connected through prflx

General
1

Hi,

I am facing frequency ICE failure / disconnection. So, I do frequent ICE_RESTARTs.
when analysed, I found that peer connections connected through prflx is facing such issues.

So, I tried not adding prflx candidates for trickling as shown below

/* Render the candidate and add it to the remote_candidates cache for the admin API /
gboolean testRelay = TRUE;
if(candidate->type != NICE_CANDIDATE_TYPE_PEER_REFLEXIVE || testRelay ) {
if(candidate->type == NICE_CANDIDATE_TYPE_PEER_REFLEXIVE){
JANUS_LOG(LOG_VERB, “[%“SCNu64”] Force Priority: %d\n”, handle->handle_id, candidate->priority);
JANUS_LOG(LOG_VERB, “[%“SCNu64”] Force Foundation: %s\n”, handle->handle_id, candidate->foundation);
JANUS_LOG(LOG_ERR, “[%“SCNu64”] We are forcefully ignoring prflx candidates here\n”, handle->handle_id);
}
/ … but only if it’s ‘prflx’, the others we add ourselves */
goto candidatedone;
}

But still peer connections is connected through prflx.

What am I missing here?

@lorenzo

2

prflx are not advertised via trickle or SDP, they’re candidates discovered at runtime.

3

"i can see that peer rflx is found by libnice and it is giving the candidate as callback. But even if i am not setting this candidate in nice_agent_set_remote_candidates,libnice makes use of the peer rflx candidates. So peer rflx can’t be avoided for peer connection since it was found by libnice and it uses this internally ?

4

Yes, that’s how it works.