[OpenAFS] Practical strategies for a backup in a (very) small cell

Jeffrey Hutzelman jhutz@cmu.edu
Tue, 10 Jan 2006 21:51:54 -0500

On Tuesday, January 10, 2006 12:19:52 PM -0500 Madhusudan Singh 
<singh.madhusudan@gmail.com> wrote:

> Hi
> With extensive help from Russ and others on this list, I setup our afs
> cell on  a debian server many months ago, and it has been working out for
> us really  well. I have defined backup volumes for every user and bos
> executes a backup  process every 24 hours at an unearthly hour. However,
> the backup volume is  located on the same physical disk as the actual
> user volumes, and thus is  really not a backup.
> The physical location of the cell server (there is only one machine in
> this  cell with no plans to add any more) is quite secure. I would like
> to migrate  the backup volumes onto removable storage all the same (such
> as a large USB  harddisk). I have a few questions :
> 1. Is this a wise choice given the constraints (I cannot add any more
> machines  to this cell, at least right now) ?
> 2. Depending on your opinion on 1, how should I migrate the backup
> volume,  nothing else, to the new location ?

You cannot "migrate" backup volumes to storage other than where the 
corresponding read/write volume resides.  An AFS backup volume is a 
read-only, copy-on-write snapshot of its parent; it is intended to allow 
users to recover from unintended changes without administrator intervention 
(if they notice soon enough), and to provide something you can take offline 
long enough to dump it without causing a significant service outage.

For disaster recovery, you will need to perform regular volume dumps to 
separate media such as disk or tape, using 'vos dump', the built-in backup 
system, or one of several third-party packages which are available for this 

-- Jeffrey T. Hutzelman (N3NHS) <jhutz+@cmu.edu>
   Sr. Research Systems Programmer
   School of Computer Science - Research Computing Facility
   Carnegie Mellon University - Pittsburgh, PA