Pat Regan on 6 Sep 2004 19:28:03 -0000

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

Re: [PLUG] Disaster Recovery Strategies

On Thu, 2 Sep 2004, David Richardson wrote:

> I was thinking about just using dd to copy the entire hard drive on
> the currently working machine to a file and then if I ever need to put
> in a new drive just using dd to write the contents of the old drive
> onto the new drive.  Is it a problem if the new drive is a different
> size than the old?  Or does anyone have a suggest of a better way?

I have a feeling this will be a long answer, and it might be a little more
manual than you like.  It also may be a little overkill, but here is what
I am doing now.  It could be done better, but I am making do with the
hardware I have available.

My primary workstation is also my backup server.  It has a 3ware raid
controller with three 160 gig drives attached.  Two of those drives are
mirrored, and are 3 year warrantee Western Digital drives.  The third is
just an independant drive.  It is a Maxtor 1 year warrantee drive that I
already had to RMA, and had a bunch of trouble with the original drive.

The mirror is where my os and home directories reside, the independant
drive is my scratch space and backup area.  I use the backup drive to
store backups for the local workstation, my laptop, and two remote
desktops at different locations.

I have been using a very nice piece of software, rdiff-backup:

It is very bandwidth and storage efficient.  The website does a better job
explaining what it does than I will, so I won't try :).

I run rdiff-backup against my local machine and laptop once per day (the
laptop only gets hit when it happens to be connected at the right time, or
I manually run it once in a while).  The remove machines (connected by
cable modem) are backed up either once per week, or once per month
(depending on machine).

There are a lot of ways you can do this, but I break the rdiff backup jobs
into one job for the home filesystem, one job for everything else (my
laptop also has a seperate job for backing up my Wine C: drive).  I
exclude a lot of log files, and most of my mp3s and video files to keep
the backup size manageable.

So far, this isn't much of a disaster recovery plan...  So, once a month I
have a script that runs cpio against the rdiff directories that excludes
the increments.  I use bzip2 with the highest compression available as
well (time doesn't matter to me, and I nice the heck out of the jobs).

I have a 10 pack of dvd+rw discs that I use to rotate backups.  My laptop
and desktop combined take up 3 discs, and the 2 remote machines fit on
one.  I keep two months of cpio backups on dvd+rw at any given time.
Every three months or so (depending on how important I feel the changes to
my disks are), I burn regular dvd+r backups and store them offsite.  I
used to burn weekly differentials as well, but I was always too busy to
keep it up.

This gives me 30 or so days of "Oops!  I deleted a file!" on the cheap 160
gig.  I'm never more than one month behind if all 3 drives died at the
same time.  In the event of a fire, I won't be more than 3 months behind.

Three months is fine, because I keep my revision control repositories on
my webhost and keychain as well :).

For recovery, I have a small custom Morphix boot cd (about 180 meg) that I
append to the dvd discs.  So I just boot off the dvd, partition the new
drive, un-cpio the archives, run lilo on the new drive, and I am back up
and running.

I have tested a restore in VMWare, and it worked like a charm.  I just
changed the drive X was using, and it came right up.

I'm very impressed with the speed of rdiff-backup over the internet.  I
did both initial backups locally to my laptop, and moved them to my
desktop.  Now they probably take 1-2 hours on average to backup each

I'd be happy to share my (rather shoddy) scripts with anyone who wants

I hope this was helpful,


Philadelphia Linux Users Group         --
Announcements -
General Discussion  --