For sake of information - the rtpengine_manage() function was added to simplify the config logic of using rtpengien_offer()/_answer()/_delete(). The manage function (for both rtpproxy and rtpengine) calls the other functions based on request method and response code. The current config file is using it in a subroute, executed in few other places, no longer under conditions related to method or reply code.
Also, the manage function handles properly when INVITE is without SDP and the ACK is carrying the SDP.
Cheers, Daniel
On 02/02/16 22:38, Alex Balashov wrote:
Ah, never mind. This was a stupid question.
In short-duration traffic land, most calls end with all branches failing. I didn't have rtpengine_delete() cleanups in the failure_route[] for that scenario.
The volume of calls is far too large for rtpengine's own RTP timeout ("garbage collector") to release the ports fast enough.
False alarm, and apologies for wasting time.
-- Alex