[OpenSIPS-Users] next_gw still does append_branch in failure_route?

[Jeff Pyle]

Hello,

I¹m using the LCR module in Opensips 1.5.  (I realize I should probably be
using drouting if I¹m starting from scratch, but I need the gateway group
and flags functionality that only appear to be in the LCR module.  More on
that at the end.)

Bogdan emailed the list about not needing append_branch any longer in a
failure_route 
(http://www.mail-archive.com/devel@lists.opensips.org/msg00663.html).  All
the documentation seems to indicate that next_gw still does the
append_branch automatically.

The behavior I¹m seeing, along with Bogdan¹s email from January, seem to
indicate it¹s not necessary anymore.  In my configuration, next_gw is called
from the request_route.  The request is sent out with t_relay.  The request
fails with a 503, and is caught in the armed failure_route.  next_gw is
called again, then t_relay.  There appear to be two branches present, the
old one to the first (failed) gateway and the new one to the second newly
loaded gateway.  t_relay does a parallel fork to both of them.  The first
gateway fails again 503, and in my test setup, so does the second, also with
a 503.  One of these 503s is properly processed by the armed failure_route,
the other one is converted to a 500 and relayed to the UAC.