[asterisk-users] IAX jitterbufer oddity

[Matt Riddell]

On 27/10/09 2:07 AM, Steve Davies wrote:
> Hi,
>
> First a confession - The box in question is a 1.2.35 box, so this may
> be solved in a newer version as I know the JB code is all hugely
> changed, but... It may be worth checking into.
>
> Scenario:
>
> - IAX outbound call from Asterisk, which rings okay.
> - Remote end sends ANSWER, which we immediately ACK.
> - The ANSWER control packet gets put into the JB (that's how I read the code)
> - The remote end is clustered, and we receive a TXREQ within 1ms of our ACK
> - chan_iax2 starts to process the TXREQ correctly.
>
> What I think happens at this point is that the ANSWER control frame
> now leaves the JB in order, but is not processed because the channel
> state has moved into the new transferring state, so ANSWER has no
> meaning, app_dial never forwards the ANSWER control event to the
> calling channel, and the bridge is never fully completed, so it all
> eventually times out.
>
> Disabling the JB in IAX does resolve the issue, but is not ideal.
>
> I have tried to follow the code in the various versions 1.2, 1.4 and
> 1.6, but it is just too complicated. Does anyone know if this was
> addressed since 1.2, or can it still happen in 1.4 or 1.6?

Just a shot - all boxes using NTP?

-- 
Cheers,

Matt Riddell
Director
_______________________________________________

http://www.venturevoip.com/news.php (Daily Asterisk News)
http://www.venturevoip.com/st.php (SmoothTorque Predictive Dialer)
http://www.venturevoip.com/c3.php (ConduIT3 PABX Systems)