Re: /var/named

[chuck <Chuck_(_at_)_Yerkes_(_dot_)_com>]

It is claimed, but unverified, that Simon Dick wrote:
> 
> > On Wed, 3 Jun 1998, Todd T. Fries wrote:
> > 
> > > > Couldn't it read /etc/named.boot, and chroot to whatever directory
> > > > is specified there??
> > >
> > > .. and when you did a 'kill -HUP <pid>' it would be able to re-read it how?
> > 
> > I usually advice against sighup'ing named.  Thats what the ndc
> > restart/reload command is for, no?
> > 
> > As long as ndc knew where things were at, all would be well.
> > 
> > I haven't used bind under OpenBSD yet, so I don't know how big a problem
> > this is, though I routinely move db files around I am very used to
> > /etc/named.conf being there.
> 
> Stopping and restarting is OK as long as you don't serve too many
> domains, otherwise it saves a *lot* of time just HUPing named.

Except that it dumps the cache.  A HUP doesn't.

makes a difference on a heavily used server.

And ndc doesn't do magic, it either HUPs the process (reload) or
stops in and restarts it (restart).  It's simply a convenient
way to send signals to named (perhaps based on gated's gdc?).