asterisk users - Sep 2004 - Maximum tollerable lag/jitter for IAX2 w/oji tterbuffer enabled?

> -----Original Message-----
> From: Chris Shaw [mailto:chriss@watertech.com]
> Sent: September 7, 2004 4:40 PM
> To: Asterisk Users Mailing List - Non-Commercial Discussion
> Subject: Re: [Asterisk-Users] Maximum tollerable lag/jitter for IAX2
> w/ojitterbuffer enabled?
> 
{clip}
> 
> If you can reproduce it, this smells like a bug... IAX runs  over TCP and
TCP
> doesn't just disconnect sockets unless it recieves a RESET or 
> a FINISHED or there's a timeout (usually like 5 minutes or more
depending
> on your TCP/IP stack). Needless to say that to disconnect a TCP
connection, 
> that would have to be one hell of a lag spike... * must be actively
disconnecting the
> connection....
Reproducing it is part of the problem. I've been getting user reports on and
off for some time but I can't find anything out of the ordinary - initially
this was looking like a Voicemail bug as many people were getting cut off
while leaving messages. How would I debug the precise drop condition? I've
Googled for more information on 'iax2 debug' but come up naught.
> I've heard the jitter buffer is a bit buggy, have you tried 
> turning it off
> completely?
Is 'jitterbuffer=no' not sufficent to clobber that function?
 
Kris Boutilier
Information Systems Coordinator
Sunshine Coast Regional District

> Is 'jitterbuffer=no' not sufficent to clobber that function?
Yes, sorry I didn't see that part... Hmm....

And IAX2 debug on doesn't give you any hints at all? like who hung up first,
etc... ?

    -Chris

> -----Original Message-----
> From: Chris Shaw [mailto:chriss@watertech.com]
> Sent: September 7, 2004 5:14 PM
> To: Asterisk Users Mailing List - Non-Commercial Discussion
> Subject: Re: [Asterisk-Users] Maximum tollerable lag/jitter for IAX2
> w/ojitterbuffer enabled?
> 
> 
> > Is 'jitterbuffer=no' not sufficent to clobber that function?
> 
> Yes, sorry I didn't see that part... Hmm....
> 
> And IAX2 debug on doesn't give you any hints at all? like who 
> hung up first,
> etc... ?
> 
Alas, I haven't caught it in the act yet - there have been 420 IAX2 to Zap
and another 500+ Zap to Zap calls on that server so far today. I've had
reports of about 30 dropped calls.

What I really need is to find a better way to post-mortem debugging for
large numbers of concurrent IAX2 calls...	

Kris Boutilier
Information Systems Coordinator
Sunshine Coast Regional District

On Tue, 7 Sep 2004, Kris Boutilier wrote:
> Reproducing it is part of the problem. I've been getting user reports
on and
> off for some time but I can't find anything out of the ordinary -
initially
> this was looking like a Voicemail bug as many people were getting cut off
> while leaving messages. How would I debug the precise drop condition?
I've
> Googled for more information on 'iax2 debug' but come up naught.

Run Asterisk with debugging turned on - see my various posts here 
explaining how to capture it all in /var/log/asterisk/debug.  That will 
reveal all.

But I suspect you got some NAT between the end points, and that NAT is 
messing up and breaking the communications?

Steve

Collecting debug today - got 8mb so far. Just turned up IAX2 debug so
expecting it to balloon.

There is no NAT involved, nor stateful firewalling etc. - this is a flat
10.0.0.0/24 subnet with one 10base-t hub and two 10base-t cable modems
(operating peer to peer) in between the endpoints. Bitrate is proven out at
not less than 512kbps using Iperf (http://dast.nlanr.net/Projects/Iperf/).
Jitter buffer, trunking and now native bridging (notransfer) code has been
intentionally disabled to reduce possible culprits. Hosts are running a load
average of between 0.00 and 0.02.

The call drop issue has been very sporadic and effectively unreproducable
from a testing perspective. At the moment I'm leaning in the direction of an
IAX2 ACK packet being dropped off the network - I've noticed about 0.5%
collisions on the wire (this being a half-duplex network) which seems to be
contributing to audible pops and clicks with the jitter buffer disabled. 

I've also been getting occasional reports of far end echo on long distance
calls, but I suspect the ongoing thread about 'Question on echo's for
Canadian Asterisk users...' will get to the bottom of that one.

Does anyone know how robust/agressive the UDP retry code is in Asterisk?
> -----Original Message-----
> From: steve@daviesfam.org [mailto:steve@daviesfam.org]
> Sent: September 8, 2004 3:02 AM
> To: Asterisk Users Mailing List - Non-Commercial Discussion
> Cc: 'Chris Shaw'
> Subject: RE: [Asterisk-Users] Maximum tollerable lag/jitter for IAX2
> w/oji tterbuffer enabled?
> 
> On Tue, 7 Sep 2004, Kris Boutilier wrote:
> 
{clip}
> 
> 
> Run Asterisk with debugging turned on - see my various posts here 
> explaining how to capture it all in /var/log/asterisk/debug.  
> That will reveal all.
> 
> But I suspect you got some NAT between the end points, and 
> that NAT is messing up and breaking the communications?
> 
> Steve
>