RTP statistics correlation

[petekelly]

Hi

I am using Homer 7 in a situation where a call will proxy though Opensips multiple times with the same Call-ID and fromtag. Because of this I am generating alternative call-ids for rtpengine to avoid sessions clashing with each other.

The problem I am having is then that the correlation information for rtp contains a different callid to the Call-ID used in the SIP, so there are no rtpstats being presented in the Homer UI.

Is there a way to tell Homer to look for these alternate callids when generating the rtp statistics in the UI?

Pete

[github-actions]

Please star this repository to motivate the developers and to get higher priority! ⭐

[lmangani]

Hey @petekelly good to see you around! Where there's a will there's a way and correlation is pretty flexible as long as there's a vector to use but to get started, could you attach a full example? it's hard to tell without seeing a full set and without knowing what statistics you are referring to.

[petekelly]

Hey @lmangani I'm always about hovering :)

So I have a scenario where the same SIP proxy receives a call 3 times. Same Call-ID and fromtag. All is fine but each call needs its own rtpengine session. So I give each rtpengine session it's own "call-id" which differs from the SIP Call-ID.

e.g.

SIP Call-ID = XXX

rtpengine first proxy of the call: XXXa
rtpengine second proxy of the call XXXb
rtpengine third proxy of the call XXXc

All this is fine, and OpenSIPS sends HEP to homer and the call traces join up just fine when searching for call-id XXX.

However when rtpengine sends its data for rtpstats to homer, it only knows to use callid (or correlation ID?) XXXa/b/c - so there are no rtp stats when viewing the call in Homer..

I assume when the call is loaded in Homer, it searches the DB for the rtpstats based on Call-ID=XXX - can it also be told to query Call-ID=XXXa or XXXb or XXXc ?

I tried sending a second correlation SIP header in with the call (so XXXa/b/c) but it hasn't seemed to work.

Hope that makes sense...

Pete