[ On , December 4, 1999 at 17:20:00 (+0200), Kai Henningsen wrote: ]
> Subject: Re: [Exim] Delivery optimization tips for large lists ?
>
> pete@??? (Pete Naylor) wrote on 11.10.99 in <Pine.GSO.4.05.9910112128030.2099-100000@???>:
> >
> > On Mon, 11 Oct 1999, Peter Radcliffe wrote:
> > >
> > > The MX record for example.com points at mail.example.com.
> > > mail.example.com has two A records, 10.0.0.1 and 10.0.0.2, two
> > > distinct machines for failover. nscd caches mail.example.com as
> > > 10.0.0.1 when someone sends mail to example.com. 10.0.0.1 goes down.
> > > Mail never reaches 10.0.0.2 until the first A record times out from
> > > nscd, which is a long time since nscd's default timeouts for positive
> > > caching are quite high.
> > >
> > > This is broken.
> >
> > That's right - I hope nobody is silly enough to configure their DNS in the
> > manner that you've just described. Really bad practice.
>
> Um, if the described behaviour of nscd is real, then I strongly suspect it
> breaks a MUST in one of the DNS RFCs (by caching only one of those two A
> records).
It's very real. I don't have an RFC reference handy though....
> There can really be *no* excuse for this behaviour.
Indeed. I thought this was an FAQ long ago. One should *NEVER* run
nscd on a Solaris machine using Internet DNS services without adding:
enable-cache hosts no
to /etc/nscd.conf and rebooting. (either that or just kill nscd -- it
doesn't buy you much unless you also use NIS on a large/busy network)
Nscd's ability to do hostname caching was intended *ONLY* for the case
where hostnames were available only through NIS. Nscd does not
understand *anything* about DNS and it only gets in the way of a real
nameserver. Unfortunately Sun, in their infinitesimal wisdom, decided
that all the world should be using NIS and didn't give a second thought
to the danger of the default configuration of nscd.
--
Greg A. Woods
+1 416 218-0098 VE3TCP <gwoods@???> <robohack!woods>
Planix, Inc. <woods@???>; Secrets of the Weird <woods@???>
