Krister Jarl wrote:
> Can you please clarify for me why this is incorrect DNS configuration? I
> had the following setup in mind when writing the question. 
> 41520 IN   SRV     1 1 5060
> 41520 IN   SRV     2 1 5060
> What this essentially means is that should be used
> as long as it's responding. But this doesn't happen since it falls out
> of the blacklist.

Sure, server1 should be used whenever it is available and YXA should try 
to fall back to server2 when server1 is _not_ available.

If server1 is permanently unavailable, it is a misconfiguration of DNS 
to list it as the best candidate for This is how I 
understood the scenario that you wanted to "solve" with a longer 
blacklisting period or otherwise...

SIP, as specified in RFC3261 does have a problem where it will always 
fail (timeout) the first non-INVITE sent to when server1 
becomes unavailable (if the first request sent to after that 
is a non-INVITE). The reason for this is basically that the transaction 
layer will not ask the transport layer to try server2 before the 
transport layer has reported the timeout of server1, and when it does - 
well then it's too late because the (non-INVITE) transaction has timed out.

Yxa-devel mailing list

Reply via email to