-*-text-*-
$Header: /CVSROOT/public/scylla-charybdis/md5backup/doc/linked_file_store.txt,v 1.1 2004/05/09 21:12:52 tino Exp $

Linked file store
=================

Version 0.3.1 now has support of linked file store.

There are some helper scripts which are not completely mature for
today.  These scripts are based on scylla-charybdis (see
http://www.scylla-charybdis.com/, this is probably where you
downloaded md5backup from, too).

For more information please have a look into doc/sc-backup.txt


Purpose:
========

I have a lot of leased servers.  These servers must be backed up.  For
two purposes:

a) Be able to restore a previous point in time in case of a failure.

b) Be able to restore the machine if it is destroyed

This now is done as follows:

a) The machine makes a md5backup at a regular interval from the only
   local hard drive to the only local hard drive.

b) The new MD5 files are hardlinked into a secondary file store.

c) These files in the secondary file store are transported to an
   external other system with a huge hard drive using Scylla+Charybdis
   (see sc-move.sh), thus they are removed(!)  after successful
   transfer.

d) A regular shell script shall run each day and sends eMail if
   everything is OK, so all computers have send some backup data.
   Note that this script is not written yet.

e) These external hard drive then is backed up at regular intervals
   onto my local (intranet) backup pool, via Scylla+Charybdis, too.

Yes, I know, this is a little bit redundant.  However this is the
easiest way to do it - for me.


Why do I want to do it this way?
================================

I like to be sure that everything is OK in case something breaks.
Usually you have to make it sure *after* a crash.  How can you tell
that the backup system is OK in case the backup system is broken?

Well, you can, if you do it as above.  Even if something really
strange happens, all the nifty stuff in S+C and md5backup make sure,
it's nearly impossible that you loose any of the valued data.  And as
well, as there is a 1:1 association to the files, it's really easy to
access all the data without any other utility.  All you have to do is
to identify which data is which.

However this usually can be done by context:

- use `file` command

- look into the contents (view pictures, archive contents, grep text)

As you like this is even possible without unpacking anything by
accessing the file store directly.  Quick and dirty, as I like it.


Additional note:
================

Many machines backup themself.  The interesting part is, if two
nearly similar machines store data to the backup, S+C ensures,
that the data is not stored twice as the MD5 sum matches!
Therefor you only transfer each file once even if it is stored
several times on several servers.

Yes, I need this, as else I would have information explosion.

-Tino
$Log: linked_file_store.txt,v $
Revision 1.1  2004/05/09 21:12:52  tino
added