Hi Andy,
yes, right...it's a matter of choice...
what I was trying to say is that the consistency should be kept and do record and loose route on same device with proper IP: either do RR on LB with LB's IP and loose also on LB, either do them on proxy with proxies's IP.
regards, bogdan
Andreas Granig wrote:
Bogdan-Andrei Iancu wrote:
just a short note...on the sip proxy you should do rr with the proxy's IP and not with the one of the LB. Once the LB decides which proxy handles the call, all sequential request must go through that proxy and ideally to skip the LB.
Well, I personally prefer to do rr on the LB, because you can then perform a failover to another proxy on the LB if the specific proxy fails. Think of a scenario where LB routes the INVITE transaction to proxy P1 selected by Call-ID hashing, and after successful call-establishment P1 fails. If the BYE then hits LB, it will try to route it to P1 again because of the same Call-ID, and after hitting a timeout, a fallback to P2 can be performed. So no failover has to be implemented in the UACs.
For accounting purposes, the acc-records can be collected from all of the proxies to create the CDRs.
Just my 2 cents, Andy