[SR-Users] TM question

[Alex Balashov]

Hi,

I have a somewhat unusual log entry after some calls:

Sep  3 09:12:23 kamgw01 /usr/local/sbin/kamailio[542]: WARNING: tm 
[t_lookup.c:1551]: WARNING: script writer didn't release transaction

These calls fail in that they are processed instantly, but no time is 
ever given to receive a response from the upstream gateway.  For 
instance, the above call's initial INVITE was received at 09:12:23, a 
final request URI was resolved, and then this message, all in < 1 
second.   Trace on upstream gateway shows that the initial request was 
never received from Kamailio.

There is one additional aspect to how these calls are being processed 
in script that almost certainly has a bearing on this, though I don't 
know what it is:

They are handled through a gateway route route called 
route[OUTBOUND-ROUTE-TRY] that is, under normal circumstances 
reflecting its name, initially called from the request route, and 
subsequently from a failure route meant to advance it.

Normal operation mode is like this:

    route {
       ...

       route(OUTBOUND-ROUTE-TRY);
    }

    route[OUTBOUND-ROUTE-TRY] {
        if(dialog_count_of_vendor_gw < limit) {
            t_on_failure("OUTBOUND-ROUTE-FAILURE");

            if(!t_relay()) { sl_reply_error(); }


            exit;
        } else {
           ...
           # Advance index of AVP array to next gateway, etc.
           #

           if(!t_lookup_request())
              t_newtran();

           route(OUTBOUND-ROUTE-TRY);
           exit;
        }
    }

    ...

    # Armed in OUTBOUND-ROUTE-TRY with t_on_failure().

    failure_route[OUTBOUND-ROUTE-FAILURE] {
       ...

       # If we have more routes to retry, call
       # OUTBOUND-ROUTE-TRY again.

       route(OUTBOUND-ROUTE-TRY);
    }


However, what is happening to cause this error is a little different 
from the usual scenario.  On the initial attempt, the dialog limit 
check you see in OUTBOUND-ROUTE-TRY is failing, so the route adds a 
new transaction and calls itself again once.

Now, the reason I am creating a new transaction is because in certain 
failure modes when a transaction was never created (like, if the call 
also fails on the next attempt), it is otherwise not possible to send 
a stateful failure reply to the originating endpoint from the failure 
route, which is a requirement (one cannot issue stateless replies from 
failure routes).  So, because this routine is called both from the 
failure route and from the request route, I create a new transaction 
in order to make it possible to send stateful replies (t_reply()) from 
within this route regardless of the type of the parent route calling 
it, whether it is failure route or request route.

So, again, what is happening in the scenario in which I get the above 
log message is:

    1. INVITE is received
    2. Master request route runs OUTBOUND-ROUTE-TRY request route
    3. OUTBOUND-ROUTE-TRY checks gateway limit, finds it is exceeded
    4. OUTBOUND-ROUTE-TRY checks request against existing transaction,
       and if not found, creates a new one with t_newtran()
    5. OUTBOUND-ROUTE-TRY calls itself again.
    6. OUTBOUND-ROUTE-TRY succeeds this time, but evidently,
       t_relay() fails with the above message, or above message appears
       sometime before t_relay() is called.

Any insight is appreciated!

Thanks,

-- 
Alex Balashov - Principal
Evariste Systems LLC
1170 Peachtree Street
12th Floor, Suite 1200
Atlanta, GA 30309
Tel: +1-678-954-0670
Fax: +1-404-961-1892
Web: http://www.evaristesys.com/