> There was a post made about a year ago regarding this rndc reconfig issue
> (shown below).
> >From my understanding it seems that if you compile named with

> "--enable-threads" and if you start the name server with "-n #" (where the
> number is larger than 1) then you should be able to do a "rndc reconfig"
> without a good deal of downtime.
> Is this correct or not?
> >From trying this on a good amount of zones there is still the same amount of

> downtime as doing "rndc reload".
> I have tried this on RHEL, FreeBSD 6.1, NetBSD, Gentoo, Debian, and
> Mandrake. All function the exact same way.
>
> Can this timeout be prevented in anyway?
> I've seen this question asked numerous times without a definate answer, so I
> thought I would just ask it one more time.
>
> If anydone does have this working, what were the compile options and the
> startup options that you used?


You have 100000 zones. It takes time to read the config
file(s). It takes time to find if the old zone exists so
it can be re-used or if it doesn't build a new. To
{re-}establish all the acls etc. for each zone. This is
all done as a single operation.

> On 11/09/05, Paul Vixie wrote:
>
> > Tavis Gustafson writes:
> >
> > > Hello all. We currently run bind9 on 3 auth dns servers, one primary,
> > > two secondaries. With over 100k domains, any time we do rndc reconfig
> > > on master or rndc reload on secondaries, we get lookup timeouts of up to

> >
> > > 30 seconds. Even with staggering the updates on primary and secondaries

> >
> > > it's getting to be a real problem.

> >
> > 9.3.1, built with threads on anything but freebsd<6.0, should make this
> > better.
> >
> > 9.4.0, built with threads on anything but freebsd<6.0, *will* make this
> > better.
> >

>
> Thanks for any help.
> -jm
>
>
>

--
ISC Training! October 16-20, 2006, in the San Francisco Bay Area,
covering topics from DNS to DHCP. Email training@isc.org.
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: Mark_Andrews@isc.org