[SR-Users] SIP and conntrackd experience?

[Daniel Tryba]

On Wednesday 30 September 2015 13:07:12 Grant Bagdasarian wrote:
> A is connected to C in say 10 seconds and they're exchanging RTP.
> Now, the firewalls at B are turned off, or crash or whatever, mid call.
> 
> Would this cause the session to be disconnected? The media is obviously
> gone, but will the session remain active? I'm looking for a way to restore
> almost instantly in case of any failure, be it firewall related.
> 
> My idea was to configure conntrackd on the firewalls which are setup in HA
> mode and where a crash of the primary would result in a failover to the
> backup with full connection restore.

As long as there is no additional signalling (DTMF over SIP INFO, session 
timers, onhold events....) or endpoints detect the lack of RTP (or RTCP) after 
a short timeout, the SIP session is still alive according to the endpoints.

So if RTP is only forwarded by the kernel (like the kernel forwarding 
capabilities of rtpengine) and you could replicate these forwards to a standby 
server, you could do a failover of both SIP and RTP. Theoretically... 

If you are running rtpproxy, which actively accepts and forwards the RTP 
traffic, you'd have to replicate the state of the proxy to a standby machine. 
Which is AFAIK not possible.