[SR-Users] Dialog module timeout behaviours

[Alex Balashov]

Hello,

I'm running 4.3.5:1b0c0a and have run into two distinct, bizarre issues
with the dialog module that I am unable to explain. 

In all cases, I am using $dlg_ctx(send_bye) = 1, and passing in a dialog
timeout of 7200 sec via AVP (timeout_avp modparam). 

1) Local BYEs are sent after either ~60 or ~120 sec --

This happens on a very small number of answered calls for no apparent
reason. I have spent days poring through the traces trying to figure out
if there's anything eccentric about how the messages, but can't find
anything out of the ordinary except (a) 1 retransmission of 200 OK to
INVITE due to not receiving ACK fast enough due to end-to-end latency,
and (b) consequently, retransmission of ACK.

My understanding is that the dialog module does not prophylactically end
calls as a response to strange or improper messaging, but only (a) when
the dialog timeout is reached or (b) keepalive OPTION pings (if enabled)
fail.

Keepalive OPTIONS pings are not enabled in these scenarios, and are not
sent in the traces.

(2) Local BYEs sent in response to race conditions that result in
strange dialog state transitions --

There are some calls that look something like:

A. Caller INVITEs callee;
B. Callee sends progress;
C. Callee sends 200 OK answer and immediately sends BYE (e.g. 1 ms
later);
D. Kamailio receives these nearly simultaneously and forwards the BYE
first;
E. Caller sends 487 Request Terminated response (processing BYE as if it
were CANCEL);
F. Kamailio forwards 200 OK back to caller, who ignores it.
G. Callee retransmits 200 OK through Kamailio.

A few minutes later, Kamailio sends BYE in both directions even though
dialog should have been removed from state tracking long ago because of
the dispositive 487 reply.

...

In particular, I'm curious as to whether there have been any fixes to
deal with scenarios like #2 since 4.3.5. I wasn't able to spot any
obvious ones in the commit history for the module. I'm also at a
complete loss to explain #1. 

I have traces for all of these scenarios, but would prefer to send
privately.

Many thanks!

-- Alex

-- 
Alex Balashov | Principal | Evariste Systems LLC

Tel: +1-706-510-6800 / +1-800-250-5920 (toll-free) 
Web: http://www.evaristesys.com/, http://www.csrpswitch.com/