[Kamailio-Users] Failover from SRV lookups?

[Henning Westerholt]

On Wednesday 03 December 2008, G.O. Jensen wrote:
> Problem is as follows and I hope somebody can shed some light on it.
>
> Our GW supplier has two SBC's that they switch between... to find these I
> do a DNS lookup at their domain which gives me;
> [..]
>
> The "problem" is... OpenSER tries the first server (sbc2) first everytime.
> Natural since it's listed first. However, this one doesn't answer most of
> the time, and OpenSER then has to timeout before trying the next (with
> frequent resends from my UAs overloading the servers more in the
> meantime).
>
> With fr_timer out of the tm module I can set this timeout to 2 seconds,
> but it still takes nearly 10 seconds before I get a ringing signal (183)
> from the GW SBC.
>
> Is it possible to somehow get OpenSER to "cache" the active server? Or
> query it "beforehand" with an OPTIONS message or something to see which
> one is alive (that is quicker than waiting for timeouts and retries?)
>
> Thanks in advance for your prompt help since our GW changed this overnight
> on our production system and my users are not too happy about it :D

Hi Geir,

openser used indeed in the past always the first SRV record in this case. In 
the 1.4 branch this limitation is fixed, here is a random one choosen when 
records with the same priority are returned. Perhaps you can upgrade to the 
1.4 branch, if this is not possible now you could try to backport the 
respective changes to your version.

Cheers,

Henning