[Asterisk-Users] 7960 fails to ack INVITE when multiple lines configured to different * servers

[Chris A. Icide]

Recently I've been doing some testing of a Asterisk installation for a 
client and I've re-assigned two lines on my 7960 to use the test asterisk 
box that has the developed installation for the client.  So I have 4 lines 
configured for my asterisk installation and two lines for the test bed.

What I've noticed is that when I leave the test bed down (asterisk not 
running, but the linux box itself on-line), my 7960 will fail to ack 
INVITES sent by my asterisk box.  If I dial an extension that is assigned 
to the 7960 under my asterisk box, after an indeterminate amount of time 
(say somewhere between 5 and 30 minutes) the 7960 will fail to ack the invite.

The console on the active asterisk server shows that it attempts to dial 
the 7960, and timeouts according to the time specified in the dial timeout 
statement.  The 7960 shows no indication of receiving the call.  sip show 
peers on the asterisk box shows the associated 7960 channels registered, 
and a qualify statement in the sip config results in showing good 
connectivity during this problem.  Calls can be placed from the phone on 
the same extension that fails to receive the call.  Asterisk still seems to 
think the phone is registered.

Now here is the interesting part.  Rebooting the cisco phone fixes the 
problem for some indeterminate amount of time (5-30 mins), but then the 
failure mode appears.  Also, firing up asterisk on the second box fixes the 
problem as well, even if the cisco phone is in the failure mode I 
described, firing up of the test bed asterisk server immediate fixes the 
failure mode from the first asterisk box.

The only thing I can think that might be an issue is that the cisco phone 
is on an RFC1918 address space which it shares with the test bed box, and 
the phone interacts with the test bed asterisk server without going through 
a NAT, whereas the main asterisk server is on the other side of a NAT, and 
has nat=yes configured in the sip.conf as well as NAT enabled on the cisco 
phone.

The problem I see is that the cisco phone uses the NAT enabled for all 
lines (you cannot configure it for each line).  So in this situation when 
the second asterisk server is in operation, everything works fine, but when 
it's shut down, after some amount of time, the cisco phone ACTS as if it's 
not registered with the active asterisk server.  But turn on the test bed 
server and all is fixed.  Also disabling or re-configuring the lines on the 
cisco phone that were configured for the test bed server back to the main 
server also fixes the issue.

Has anyone else ran into this issue?


-Chris