Bind 9.6.1: skipping zone transfer, but why ?

[Danny Mayer]

Chris Buxton wrote:
> On Jun 30, 2009, at 6:15 AM, bind9 wrote:
>> 1) "skipping zone transfer as master 213.173.250.146#53 (source
>> 0.0.0.0#0) is unreachable
>> (cached)" seem to indicate that the slave has cached a knowledge about
>> the master being
>> unreachable. It isn't. I can nslookup on the master from the slave
>> just fine. What is wrong?
> 
> The slave is caching, for some length of time set in the source code (an
> hour? something like that), that the master is unreachable for zone
> transfers.
> 
>> 2) what causes "transfer of '3yhta.dk/IN' from 213.173.250.146#53:
>> failed to connect:
>> connection refused" ? There is no evidence of "connection refused" in
>> the masters log, so where
>> could this come from?
> 

The connection refused error means that nothing is listening at that
port on that addresses. That means that either that address was not
configured to listen on that address or the server has gone down.

> 
> The master is unreachable over TCP. The port has gone deaf. We see this
> on some operating systems and not others. (We don't work much with BIND
> on Windows, so we hadn't seen the issue on that OS.) Basically, when the
> port is not used for a while, it looks like the OS shuts down the
> listener without telling the service.
> 

No, Windows doesn't do that. It is no different from a Unix O/S. I have
no idea what you mean by the listener here or the service, but on
Windows the service is only involved with getting the server running and
does not know or care about what IP addresses and ports get used if they
get used at all. This is no different from Unix.

Danny