really deleting

[Robert Edwards]

On Sun, 2 Mar 2003 06:52 pm, Brad Hards wrote:
> On Sun, 2 Mar 2003 18:36, Martin Pool wrote:
> > On  2 Mar 2003, Andrew Donehue <andrew at donehue.net> wrote:
> > >           I am after a program that will go through an ext2 based
> > > filesystem and look for any deleted files, and actually zero out the
> > > contents of that deleted file on the hdd..... The reason for this
> > > is.....
> >
> > I don't know of a program that'll do it, but you should be able to get
> > the information you need out of debugfs, and build something in a
> > script on that.
> >
> > Or perhaps you can use dump(8) rather than dd.
>
> Almost certainly, you don't want machines that are totally identical. Think
> about IP addresses, etc.
>
> http://www.linux-mag.com/2002-12/cloning_01.html provides some approaches
> that are a bit more sophisticated than dd. Kickstart will probably help,
> unless you are using a less flexible distribution (such as Debian :)
>
> Brad

Actually, all of our bunyip nodes and all of our lab machine use _identical_ 
root filesystems (they all use the same network boot image). It is quite 
practical to make clones of machines with _identical_ disk images. Use DHCP 
to get per-machine specific information (usually just the hostname and IP 
address).

Given Andrew's situation, this is what I would do (and have done):

 - rsync your working disk image to your backup area (you do have a backup of 
this, right?)
 - boot the machine from some rescue media (I still use the LinuxCare bootable 
business card)
 - zero out the hard drive with dd if=/dev/zero ...
 - partition and format etc.
 - rsync the whole business back again from the backup
 - don't forget to run lilo! (or grub etc.)
 - now dd and gzip your (clean) disk (or partition).

Always works for me (not that I clone hard disks that often - why bother when 
you can netboot?).

Cheers,

Bob Edwards.