[OpenSIPS-Users] Cleanup acc & mediaproxy when dialog bye_on_timeout triggers

Jeff Pyle jpyle at fidelityvoice.com
Mon Nov 16 14:01:05 CET 2009


Have you considered using set_dlg_flag() and is_dlg_flag_set()?

- Jeff

On 11/16/09 4:29 AM, "John Quick" <John.Quick at smartvox.co.uk> wrote:

> Hi Bogdan
> I have changed my script to only use engage_media_proxy() - I believe this has
> solved the problem with
> ending the media proxy sessions.
> However, local_route calling acc_aaa_request twice continues to be a problem.
> Thanks, I was aware of
> the risk that gflag will not always work if there are many calls all timing
> out together. I would
> prefer to use flags in the dialog or $DLG_dir, but practical tests showed that
> this was not possible.
> $DLG_dir and $DLG_status both return null, yet $DLG_lifetime returns a proper
> value. When I tried to
> use the function store_dlg_value() in local_route, opensips refused to run and
> reports an error in the
> script.
> John
>> -----Original Message-----
>> From: Bogdan-Andrei Iancu [mailto:bogdan at voice-system.ro]
>> Sent: 15 November 2009 19:32
>> To: John.Quick at smartvox.co.uk; OpenSIPS users mailling list
>> Subject: Re: [OpenSIPS-Users] Cleanup acc & mediaproxy when dialog
>> bye_on_timeout triggers
>> Hi John,
>> John Quick wrote:
>>> I am using the bye_on_timeout_flag within the Dialog module to limit the
>>> maximum duration for a
>> call.
>>> When it is triggered I am having problems cleaning up - i.e. accounting the
>>> end of call correctly
>> and
>>> ending the mediaproxy session. I am not using engage_media_proxy() function,
>>> instead preferring
>> to
>>> have more control with use_media_proxy() and end_media_session().
>>>> From a search of the archived questions on this forum I found the trick is
>>>> to create a
>> local_route
>>> section and use a check based on is_method("BYE") to determine when to
>>> execute acc_aaa_request().
>>> This works ok, but it is triggered twice at the end of the call - once for
>>> the BYE sent upstream
>> and
>>> again for the BYE sent downstream. The only solution I could find (I've
>>> spent many hours testing
>> ideas
>>> for this) is the use of a global flag that is toggled each time the
>>> local_route block gets a BYE
>>> method. Then the code only calls acc_aaa_request() when the global flag is
>>> set - i.e. only one in
>>> every two times.
>>>   if (is_gflag("1")) {
>>>       acc_aaa_request("Internal BYE");
>>>   }
>>>   # Use Global flag 1 to avoid double reporting/accounting of timeout BYE's
>>>   if (is_gflag("1")) {
>>>       reset_gflag("1");
>>>   } else {
>>>       set_gflag("1");
>>>   }
>>> This is not a very elegant or satisfactory solution but it should just about
>>> work and hopefully
>> not
>>> many timeouts will occur anyway. Any suggestions for a better solution would
>>> be welcome. I have
>> tried
>>> Dialog values and transaction flags - they don't work.
>> This solution is not correct (using the gflags) - triggering the acc
>> from local_route is the right thing to do, but not using the gflags -
>> these flags are global and shared by entire opensips which can run (in
>> the same time) multiple local_routes for different calls -> so you may
>> have a mixture between different calls.
>> Try using the "$DLG_dir" script variable
>> (http://www.opensips.org/html/docs/modules/devel/dialog.html#id273200)
>> to do the acc only for only one direction will be ideally, but reviewing
>> the code I think that variable is not populated in local_route - if you
>> could give it a try and let me know, it will be great.
>> Other idea will be to use the dialog flag or values - you said they do
>> not work, even if they should - what seams to be the problem here ?
>>> The other problem is how to end the media proxy session. The transaction
>>> flags set for the Invite
>> are
>>> not visible in the local_route block when it is handling the internally
>>> generated BYE's so I
>> cannot
>>> try the normal checks that would be used in the main route block for BYE.
>>> Perhaps I should just
>> call
>>> end_media_session(), but will it even work from local_route? If I change my
>>> script to start using
>>> engage_media_proxy() will I lose the flexibility of being able to check
>>> which calls need media
>> proxy
>>> and which don't? I don't want them all using it - only those with far-end
>>> NAT.
>> Shouldn't the media proxy automatically stop when the dialog is
>> destroyed ? if you use the enagage_mediaproxy(), I think you do not have
>> to explicitly terminate the relay session as the this op will be done
>> based on the "dialog destroyed" event.
>> If it does not work like this, you can use a dialog flag to remember
>> (from INVITE time) if the NAT was detected or not. And in local_route
>> you can do an explicit end_mediaproxy().
>> Regards,
>> Bogdan
>>> John Quick
>>> Smartvox Limited
>>> _______________________________________________
>>> Users mailing list
>>> Users at lists.opensips.org
>>> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
> _______________________________________________
> Users mailing list
> Users at lists.opensips.org
> http://lists.opensips.org/cgi-bin/mailman/listinfo/users

More information about the Users mailing list