[asterisk-dev] Possible bug in either RTP handling or AGI handling - depends on how you look at it

[Nir Simionovich]

Hi All,

   I've been trying to investigate the scenario I'm going to describe 
for the past 2 week, however,
I am unable to pinpoint the exact issue - or to which component it 
relates directly. I will describe
what happens, to as much details as I can.

   I have a customer that operates several A2Billing servers (yes, I 
don't like a2b, but people do).
The scenario that happens looks like this, as a call is being made on 
the system and the two ends
are talking, for no apparent reason, media will suddenly go away to be 
replaced by a screeching
high pitch sound, forcing you to disconnect the call.

   Now here is the funny part, once the sound is heard and you 
disconnect the call, two things
happen:

1. The actual inbound channel remains active, although the channel had 
been disconnected.
2. The channel will remain active for a period of exactly (and I do mean 
exactly) 16 minutes!
3. The AGI script dies once the channel dies, however, will not 
terminate properly causing a miss
     of billing.
4. And this is the odd part, if the call had lasted 45 minutes till the 
sound was heard, it's counters
     in Asterisk, when looking at "core show channels" would show 
00:00:00 - in other words, something
     had caused a complete channel reset or dead-lock somewhere.
5. When doing a "sip show channels", the problematic channel will show 
up as a "BYE" state, and
     will remain in that state for the 16 minutes - how odd is that?

   The above had been observed on SIP channels - and I've regressed both 
1.4.X, 1.6.2.X and 1.8.X,
all exhibit the same behavior. This can be seen on both development 
versions and stable versions.
I hadn't yet tested this with IAX2 channels, however, as the media 
handling in IAX2 is totally different
than with SIP, and it would appear that this has some relation to SIP 
RTP handling - I can assume I
won't see the issue there. Non the less, the problem is still something 
that needs to be investigated.

   I've got a system setup that can be used in order to debug the 
situation.

   Any thoughts anybody? I hadn't posted this on the users list, simply 
due to the fact that I've ascertained
this is not a config or a2billing issue.

Regards,
   Nir Simionovich