[asterisk-users] Bridged ZAP calls do not release

[Mitch Thompson]

Hello, All.

I am currently running a test configuration for the telephony engineer here.
I have two Dell PE servers, each with 2 A108 Sangoma T1/E1 cards.  Here is a
rough drawing:


Ameritec Call Generator
   8 T1 ISDN/PRI lines    ---> Asterisk        ---> Asterisk         --->
Ameritec Call Generator
                               Sangoma A108          Sangoma A108
Terminating 8 ISDN Lines

The left Asterisk is set up to be the Master clock in both directions, as
well as pri_net in both directions.
The right Asterisk is Slave clock and pri_cpe from the left, Master and
pri_net to the right.
Both Ameritecs are, of course, receiving clock from Span one and are User.

In the left Asterisk, the dial plan is:

10101. Answer the call, wait 20 seconds, hang up.
1010[2-3]. Tandem the call through to the right Asterisk.

In the right Asterisk, the dial plan is:

10102. Answer the call, wait 20 seconds, hang up.
10103, Tandem the call through to the Ameritec.

>From the originating Ameritec, dialing 10101 results in normal calls, call
clearing, etc.

However, when dialing 10102, the right Asterisk will send a DISCONNECT after
the 20 seconds, but the left Asterisk is never releasing the call, on either
leg of the bridge.

Essentially the same thing is happening when dialing 10103, except that the
right Asterisk is properly sending a RELEASE to the Ameritec in response to
a DISCONNECT.  In the span between Asterisks, the right Asterisk is sending
a DISCONNECT to the left one, but the left Asterisk is never sending a
RELEASE, even though a debug of the span indicates that it received the
DISCONNECT, and entered state 12(Disconnect Indication).  Eventually, a
timer is reached, and the right Asterisk sends RELEASE COMPLETE, but the
left Asterisk never drops the call.

Sorry to be so detailed, but I couldn't find anything with Google.  In both
Asterisks, /etc/zaptel.conf, /etc/asterisk/zapata.conf are essentially
identical, except where noted above.  I can't find anything in the Sangoma
configuration program (wancfg) which would lead me to believe the problem
lies there.

Any clues?

Thanks in advance.

Mitch Thompson