[OpenAFS] Re: Moving Magic Trio to another domain

Andrew Deason adeason@sinenomine.net
Tue, 24 Sep 2013 14:02:53 -0500


On Tue, 24 Sep 2013 14:45:57 -0400 (EDT)
stephen@physics.unc.edu wrote:

> > You can't run an "old" and "new" server on the same machine from a
> > single IP address, that's true. But you _can_ just run the "old"
> > server, and point the old and new CellServDB entries at it, and it
> > looks like two different cells and two different servers that serve
> > the same data.
> >
> > To maybe help illustrate, it's like in HTTP/1.0 (without the 'host:
> > ' header) having two different DNS A records for the same server. If
> > you had the hostname newwww.example.com and oldwww.example.com both
> > pointing to 198.51.100.5, they would both serve the same contents,
> > but they sort of "look" like two different hosts.
> 
> Interesting. In order to avoid hilarity in such a scenario, would it
> be necessary to ensure matching of users and group (and id numbers of
> both) in both cells' PTS database?

In such a scenario, there's no "both databases". There's only one
server, one ptdb, one vldb, one /vicepa partition, etc etc. I get the
impression I'm not being clear enough, but I'm not sure how to be more
clear. In such a scenario, a client's CellServDB could look like, for
example:

> old.example.com # old domain
198.51.100.5 # foo.old.example.com
> new.example.com # new domain
198.51.100.5 # foo.new.example.com

-- 
Andrew Deason
adeason@sinenomine.net