[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: System Reinstall Setup

On Sat, Jun 12, 2004 at 09:56:12PM -0700, Kenward Vaughan wrote:
> On Sat, Jun 12, 2004 at 02:53:24PM -0500, David wrote:
> I think a whole system backup is asking for trouble if there are
> glitches in the SW which lead to catastrophe.  Probably not very
> likely, but... 

Although a distinct possibility, it isn't too likely.  I'd rather expect
hardware failure more often.

With what I have in mind, though, you'd have periodic snapshots.  If the
latest snapshot were troublesome, you could step backward until you got
a good one.

> Additionally, a whole system backup is pretty heavy, period.  Why not
> hit the important directories, then install fresh and overwrite the
> conf files, etc?  

In a large system, yes, it would be rather heavy. However, in my case,
it only entailed two CD's.  In between the full backups - I'm thinking
that semi-annual full backups would be adequate in my case, and then
monthly incrementals.

A fresh install would, indeed, be most desirable.  However, in my case,
I'm on dialup, and downloading what I now have would take an incredibly
long time.  I have thought, though, that perhaps it might be good to
simply keep a backup of my /var/cache/apt/archives and then do the most
minimal install possible, restore this directory and then proceed from
there.  right now, this directory is about 400 Meg, which, in my case,
would represent some 40 hours to download.

> I've been playing with this myself, having tried backup2l with partial
> success (backed up well, but didn't restore properly).  In further
> reading, the Linux-Complete-Backup-and-Recovery-HOWTO gives apparently
> good stuff for getting all the info about your drives/etc for
> catastrophic failures, which could be coupled to the dump package to
> give you all the info and archives needed to get it all back.
> I believe the dump package contains what you want for achieving a full
> backup if that remains your goal.

I don't have much trouble doing the backups if what I have envisioned is
feasible.  It did take me a while getting the script workable, partly
due to a goof I made in the scripting, but I now have tarfiles (gzipped,
so I'll need gzip and tar on my recovery disk).

> I haven't done much searching for
> other simple scripts which do the same on a smaller scale, but expect
> that a search of Karsten Self's site will do the trick.  :-)
> > In the event of a total HD loss, boot some linux rescue system, create
> > fresh, clean partitions, extract the full backup and then extract the
> > incrementals in order.
> Again, you may find the info in the above HOWTO to be useful in this
> regard.

I've studied that HOWTO briefly.  He has a good strategy, but from what
I envisioned, he is trying to make it totally automatic - possibly a
system that several administrators could use.  Basically here's my setup

Filesystem            Size  Used Avail Use% Mounted on
/dev/hdb2             282M  103M  164M  39% /
/dev/hdb1              16M  5.0M  9.5M  35% /boot
/dev/hdb5             3.7G  629M  2.9G  18% /var
/dev/hdb6             4.6G  2.0G  2.5G  45% /usr
/dev/hdb7             6.8G  2.0G  4.5G  31% /home

Simplifying, it's a matter of for i in filesystem, cd $filesystem tar -cf
... "." ;  (a little more than that, of course)

To restore, mount some linux system, build the partitions, mount them,
cd into that partition and tar -x ... using the latest full backup.  Do
the same with any incremental backups in order and you have your system
back just as it was.  Of course, as you said at the begin of your post,
if the system itself caused the crash, you would not be in good shape.

Reply to: