[lug] Backup system ideas

[Sean Reifschneider]

On Fri, May 20, 2005 at 03:24:27PM -0600, Nate Duehr wrote:
>I decided to give dirvish a whirl (pun intended!) on a couple of home 
>machines just to see how it did, and I'm impressed. 

Yeah, I hadn't seen it before either.  I'm using rsync to back up a bunch
of systems using a few simple scripts.  I basically keep incremental data
using "--backup-dir ../incrementals/daily-YYYY-MM-DD --backup", and use
some simple scripts to merge dailies into weeklies and monthlies.  I also
use a massive "find" command after every backup to write out a list of
files that were a part of the current backup including inode numbers in
case a historic copy has to be recovered later.

This works really well for recovering the current copy (in the event of a
disaster), or recovering a particular historic version of a file.
Recovering the full system to a few weeks ago is more difficult, as it
requires corelating the "find" information to get the historic data from
across possibly multiple directories.  However, that is a very rare
occurance in my environment.  I can't think of a time in the last decade
I've wanted to do that.

The "cp -al" of the full image before running the backup is an interesting
idea, but it really hammers the file-system.  That way to have a full copy
of the file-system, without using a fully copy worth of space.

Sean
-- 
 The most important thing in communication is to hear what isn't being said.
                 -- Peter Drucker
Sean Reifschneider, Member of Technical Staff <jafo at tummy.com>
tummy.com, ltd. - Linux Consulting since 1995: Ask me about High Availability