[SR-Users] Parallel forking and rtpengine handling

Sebastian Damm damm at sipgate.de
Thu Jan 23 10:04:58 CET 2020


On Wed, Jan 22, 2020 at 8:29 PM Daniel-Constantin Mierla
<miconda at gmail.com> wrote:
> A remark for kamailio transaction states point of view: in failure route the entire transaction is in failed state, so there is not active branch, so at that point it should delete all (previous) rtpengine sessions/branches.

This is what I had in mind. I don't think I end up in failure route if
one branch is answered. My Question was, how to handle the canceled

> As I got it, the use of via-branch flag for deleting a session makes sense in event route for branch-failure, when other branches can still be active or one was answered.

Do branches that get cancelled due to another branch answering the
call go through the branch failure route?

> Also in the case of parallel forking, if via-branch is not give to rtpengine offer command, does the 2nd (and the next) rtpengine offer command overwrite the previous one, so the rtpengine keeps only the data from the last one?

We stumbled upon this problem only because we used rtpengine without
the branch parameter. And what we saw was that the second
rtpengine_offer overwrote the first one, making the first branch
impossible to be answered in certain scenarios.

Regards and thanks for all the answers so far.

Sebastian Damm
Voice Engineer
sipgate GmbH
Gladbacher Straße 74 | 40219 Düsseldorf

More information about the sr-users mailing list