Summary:ASTERISK-10575: Record on sip trunk does not maintian voice codec data rate
Reporter:Jim Fainer (praeter)Labels:
Date Opened:2007-10-19 06:50:18Date Closed:2011-06-07 14:07:25
Versions:Frequency of
Environment:Attachments:( 0) calltest.txt
Description:Scenario: Two asterisk systems; one is sip source(asterisk 1.2.16) the other is unit under test(asterisk Unit under test is setup to answer all incoming calls and immediately begins recording for the duration of the SIP session. If a call is placed to the unit under test, spoken voice is first sent (test 1 2 3), then the call is placed on hold for three seconds, and then a final  spoken voice (end 1 2 3) for a total call duration of 19 seconds (verified by cdr record). The subsequent  recording is 4 minutes(verified by listening to entire recording; no skips or silence). If a call is placed to the unit under test and only spoken voice is done, recording = call duration. This is telling me that the data rate for pure system to system audio is different than live voice. The same condition exists if the milliwatt signal is sent instead of MoH. Recorded file exceeds call duration by x5.


Sip Source:
OS - CentOS 4.4
Kernel - 2.6.9-42.0.10.ELsmp

Unit under test:
OS - CentOS 5
Kernel - 2.6.18-8.1.14.el5 (SMP)

Both on 100M FD ethernet. Same local network. Same switch.
Comments:By: Jason Parker (jparker) 2007-10-19 11:31:24

Is the recording just going incredibly slow when you play it back?  How are you playing it back?

By: Jim Fainer (praeter) 2007-10-19 11:37:16

Wishful thinking. The audio is perfectly fine just longer. If I could upload a larger file I'd give you one of the test recordings. 19 sec ~ 4Meg

By: Jason Parker (jparker) 2007-10-19 11:40:33

So it just isn't stopping the recording when you expect it to?

Can you post the relevant configs for the machine doing the recording?

By: Jim Fainer (praeter) 2007-10-19 12:01:32

No. As described above, if I bookend MoH with spoken voice it is just that <spoken Voice><MoH almost like it file transfered sound><Spoken Voice>
I know this does not make any sense. Everything about the recorded file is fine except any non-spoken voice moves between systems like a file transfer at a higher data rate.


exten => s,1,NoOp('Custom-rcdr application start.')
exten => s,n,Answer
exten => s,n,Record(test/TestRecord%d.wav|0|0|x)
exten => s,n,Hangup
exten => h,1,NoOp('Custom-rcdr application stop.')

I use FreePBX for base configuration so the context is from-trunk which will do a goto custom-rcdr. There is only one trunk with a default inbound route to custom-rcdr. No extensions. No outbound routes.



By: Jason Parker (jparker) 2007-10-25 11:15:50

If it's longer and there is no silence, and the audio isn't distorted...  Please explain.

By: Jim Fainer (praeter) 2007-10-25 12:47:59

I am not certain how else to explain the issue. I have tried to attach a recorded file so you can see for yourself but can't get it to load without error. Is there a ftp site? What part of all the descriptions I have supplied thus far is not clear? The issue is strange, that is why I posted a bug report.

By: Jim Fainer (praeter) 2007-10-25 14:19:44

So I did a tcpdump of from the server under test. I just pulled the basic header information with the most important information being the packet size and the absolute time. The file calltest.txt has the results. I used the same scenario as earlier <spoken Voice><MoH><Spoken Voice>. If you graph the data over time you will see two different data rates. The packet size remains constant yet it transfers many more packets over time when MoH is enabled and it throttles back when it returns to spoken voice. Is rtp set as VBR or CBR?

By: Jim Fainer (praeter) 2007-11-05 15:34:18.000-0600

Did this die? I did mark this as miner issue based on the community as a whole, but it is more important for the work I was doing.

By: Jason Parker (jparker) 2007-11-06 13:44:28.000-0600

I am still not clear what is in the 3:41 "gap".

start|8 seconds of voice|3 seconds of MoH|8 seconds of voice|end

Where is the gap?  What is in the gap?

By: Jim Fainer (praeter) 2007-11-06 14:06:11.000-0600

It is all MoH. Three seconds elapsed actual time in MoH results in 3:44 of recorded MoH. As I have said, the data rate changes as soon as it is an asterisk to asterisk communication. As soon as an endpoint(telephone) is reinserted in the process the data rate throttles back to 64K. There has got to be a better way to communicate with you. A telephone call...or something. This communication is not working because I can't seem to get across to you as hard as I try.

Have you plotted the attached file? Is there a way I can send you a copy of the recorded file? Is there someone else who can review the issue? Have you attempted to replicate the issue? I can do it every time. Help me help you.

By: Jason Parker (jparker) 2007-11-06 14:19:22.000-0600

This is an issue with Asterisk 1.2.  Please try 1.4 (on both sides), and reopen if you can still reproduce.

By: Jim Fainer (praeter) 2007-11-06 15:53:35.000-0600

Changed Sip source to asterisk version as advised from qwell. Although, the time is closer 16 seconds of actual time = 60 seconds of recording. Issue still exists. New sip source kernel 2.6.18-8.1.15.

By: Jason Parker (jparker) 2007-11-06 15:59:34.000-0600

We need to see the full config and full CLI output from both sides of the call.

By: Tilghman Lesher (tilghman) 2007-12-07 14:21:04.000-0600

No response from reporter.  Please, reopen ONLY if you can provide the requested debugging information.