[OpenSIPS-Users] Mediaproxy conntrack timeout when only one endpoint is down

Giuseppe Roberti jnod at jnod.org
Mon Oct 27 14:42:30 CET 2008


I'm sorry.
My fault :)
There's no rtcp notification. I was wrong.

Regards.

Giuseppe Roberti wrote:
> Hi Dan.
> 
> I dont know python nor conntrack, but i have noticed that when one 
> endpoint have a network problem the relay got statistics about no rtcp, 
> and print out that.
> 
> Here the log:
> ...
> Got traffic information for stream: (audio) 192.168.1.66:16404 (RTP: 
> 192.168.1.66:16404, RTCP: Unknown) <-> 192.168.1.69:50052 <-> 
> 192.168.1.69:50054 <-> 192.168.1.100:10186 (RTP: 192.168.1.100:10186, 
> RTCP: Unknown)
> Got traffic information for stream: (audio) 192.168.1.66:16404 (RTP: 
> 192.168.1.66:16404, RTCP: 192.168.1.66:16405) <-> 192.168.1.69:50052 <-> 
> 192.168.1.69:50054 <-> 192.168.1.100:10186 (RTP: 192.168.1.100:10186, 
> RTCP: Unknown)
> ...
> 
> So, the question is, can you add a configuration variable in the config 
> file that tell media-relay to close a session when one rtcp change state 
> from Known to Unknown ?
> 
> Thank you.
> 
> Dan Pascu wrote:
>> On Thursday 23 October 2008, Giuseppe Roberti wrote:
>>> Hi.
>>>
>>> I have noticed that the timeout for conntrack udp its fired only when
>>> both endpoints have a network problem.
>>> How to timeout a stream when only one endpoint is down ?
>> I don't think you can do that. AFAIK, it is linux kernel policy to keep it 
>> up as long as there is data flowing in at least one direction. The kernel 
>> doesn't care much in which direction the data flows through the rule, as 
>> long as it hits the rule in either direction the timeout is reset.
>>
> 

-- 
Giuseppe Roberti
<jnod at jnod.org>



More information about the Users mailing list