[Asterisk-Users] RC1 problem? (Conversation over two IAX2 streams = nasty, gappy audio) (fwd)

[steve at daviesfam.org]

Hi,

With reference to Andrew Kohlsmith's problem with calls over IAX2 going
dead after 1minute 5 seconds:

This is caused by a bug in the "optimized bridging" code in chan_iax2.c 
interacting badly with the IAX2 jitter buffer.

This problem only affect calls where:

1) There are 2 or more servers doing optimised bridging between the 
end-points of the call.

ie a call like A-end -IAX2-> *serverX -IAX2-> *serverY -IAX2-> B-end.

To put it another way:
  Server A-end does Dial(IAX2/serverX/something)
  Server X in turn does exten => something,1,Dial(IAX2/serverY/something)
  Server Y in turn does exten => something,1,Dial(IAX2/bend/endpoint)
  Server B-end then does Dial(ZAP/g1/...) or whatever

(Note that some IAX service providers (eg Nufone) do forward calls within
their own network.  Those hops count too)

AND

2) where one or both end-points are using the IAX2 jitter buffer

This problem is in 1.0-RC1 and has been around in CVS HEAD since the last 
month or so.

We've committed a fix to the Asterisk HEAD CVS to fix the bug.

If you are affected, you can either disable the jitter buffers in iax.conf
on your end-point machines, or (and preferably I think) you can upgrade
your machine(s) to CVS HEAD.  Or look out for the inevitable 1.0-RC2 and 
upgrade to that.

Jeremy has already upgraded Nufone's servers; other service providers
running 1.0-RC1 or CVS HEAD - you'll need to upgrade too - unfortunately
the problem can't be fixed just by upgrading the end-points.

Regards,
Steve Davies