Slave nameserver question

[Barry Margolin]

In article <gc2l63$1ul8$1 at sf1.isc.org>,
 "Cherney John-CJC030" <John.Cherney at motorola.com> wrote:

> You have all been very helpful in answering my crazy questions. It would
> be wrong of me to leave you hanging. 
> 
> Barry, you're on the right track. After the nameserver comes up, it will
> be told who the master is. In the meantime, though, while it is coming
> up, it must be a slave, so it has to have a masters line. Putting an

So you're going to modify the named.conf to replace the self-master with 
the real master?  Why don't you just make it a master server in the 
initial configuration, and then rewite it to slave after it comes up?

> unreachable address there didn't seem like a good idea. Putting its own
> IP address seems like it should work, but I didn't know if there was
> something in the RFCs that disallowed that, or something in the code
> that would protect itself from a situation like that. 
> 
> Kevin, thank you for testing that. I wasn't going to get a chance to
> test that until next week sometime. I will take an extra long coffee
> break next week in your honor. If you start to get the jitters next
> Wednesday afternoon, that will be why. :)
> 
> Thanks!
> jwc
> 
> -----Original Message-----
> From: bind-users-bounce at isc.org [mailto:bind-users-bounce at isc.org] On
> Behalf Of Barry Margolin
> Sent: Wednesday, October 01, 2008 10:50 PM
> To: comp-protocols-dns-bind at moderators.individual.net
> Subject: Re: Slave nameserver question
> 
> In article <gc1c35$24e2$1 at sf1.isc.org>, Kevin Darcy <kcd at chrysler.com>
> wrote:
> 
> > Barry Margolin wrote:
> > > In article <gc0udg$1k6p$1 at sf1.isc.org>, Kevin Darcy 
> > > <kcd at chrysler.com>
> > > wrote:
> > >
> > >   
> > >> Cherney John-CJC030 wrote:
> > >>     
> > >>> Besides being a bad idea from a general design perspective, is it 
> > >>> possible to set up a nameserver as a slave for a domain, but have 
> > >>> the masters field point to itself? ("I am a slave for this 
> > >>> information, and the master is myself.") In thinking about it, it 
> > >>> seems like it should be OK. The slave will always be able to 
> > >>> contact the master, so the data should never go stale. The serial 
> > >>> number is always up to date, so there won't be any bandwidth used 
> > >>> in zone transfers. Is there something somewhere that would make 
> > >>> this not work? (Something in the code for executing refreshes or 
> > >>> parsing the named.conf file?)
> > >>>   
> > >>>       
> > >> Easy enough to test...
> > >>
> > >> (Tick tock, tick tock...)
> > >>
> > >> Yeah, it works.
> > >>
> > >> But... why? Just define it as a master.
> > >>     
> > >
> > > Maybe what he's really planning on doing is listing two masters: the
> 
> > > real master and itself.  Pointing to the real master causes updates 
> > > to propagate, pointing to itself prevents expiration.
> > >   
> > "the master", singular.
> > 
> > "... there won't be any bandwidth used in zone transfers".
> > 
> > Seems like he's setting up a master zone, but for whatever reason 
> > wants to call it a slave.
> 
> For the purposes of his question, he was only asking about this master.
> 
> That doesn't mean he doesn't plan to do something different in actual
> practice.
> 
> Well, maybe he'll come back and tell us.
> 
> --
> Barry Margolin, barmar at alum.mit.edu
> Arlington, MA
> *** PLEASE don't copy me on replies, I'll read them in the group ***

-- 
Barry Margolin, barmar at alum.mit.edu
Arlington, MA
*** PLEASE don't copy me on replies, I'll read them in the group ***