[cisco-voip] 6608 XCODER vs 3945 XCODER and SIP calls (Nuance Speech Attendant)

[Paul]

Have you done a stare-n-compare of output from 'deb ccsip media' and 'show call active voice?'

and what about 'deb ccsip messages' or device-based SIP stack trace output to compare calls with 6608 vs 3945 or lastly media resource manager trace?

A lot of this output is challenging to parse through but if you can find any key differences between the calls (translatorX might help with the parsing?) but perhaps notepad++ or a 'diff' type tool can help you with this bit. 




________________________________
From: Lelio Fulgenzi <lelio at uoguelph.ca>
To: "cisco-voip at puck.nether.net" <cisco-voip at puck.nether.net> 
Sent: Friday, August 24, 2012 2:53 PM
Subject: [cisco-voip] 6608 XCODER vs 3945 XCODER and SIP calls (Nuance Speech Attendant)



We had a very interesting problem after I migrated from 6608 resources to 3945s. In particular, we have a Nuance Speech Attendant system integrated using third party advanced SIP phones (one for each server). Each SIP phone can handle 8 calls. The SIP phones have MTP required checked.

Here's what happens:

Problem: Callers reported hearing dead air when calling 
the speech attendant. Further research showed that the problem 
occurred on the third and subsequent calls, regardless of which server 
the call was on, so in effect, we could only process two calls at a 
time. 

After some 
troubleshooting and researching, we found that it was not the removal of
any OLD (transcoding) resources but the addition of NEW (transcoding) 
resources that caused the problem. This was difficult to catch because 
not only did Nuance NOT use the old transcoding resources, but only used
the NEW transcoding resources on the third and subsequent calls. This 
is very odd to say the least. In the old MRGL, we had two software MTP resources (two subscribers), in the new MRGL we had these two MTP resources, plus the new XCODERs on the 3945s. Part of the troubleshooting also included specifically assigning the old XCODER into the MRGL.

The solution was to create new 
media resource groups (and lists) so that the new transcoding resources 
were not accessible by the Nuance system.

I suspect, that the new XCODERs on the 3945s are advertising/negotiating MTP abilities when they really aren't.

Can anyone help me understand what just happened?