The selection of rtpengine instance is done when the call is initiated, after that should be a sticky behaviour, because the replies have to go to the same instance, as well as BYE or other messages within early/established dialog.

What it seems you are looking for is high availability, so in case a rtpengine goes does another instance can take over the rtp relay sessions. This should be possible with rtpengine and redis, see the docs of rtpengine application.

Considering the above, specially the stickiness needed between request and reply, plus for BYE/CANCEL, changing the current behaviour makes no sense, but being open source, one can contribute such alternative controlled by an modparam.

It is not a bug right now, but how it is expected to work.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or unsubscribe.