Re: backup strategy using rsync
On Thu, Jan 18, 2007 at 02:08:31PM -0600, Randall Smith wrote:
> I've been looking at various backup strategies in the case I need to
> quickly (within an hour or so) move to different hardware. I like
> Linux-VServer because it makes the system mostly hardware agnostic, but
> I only need a single instance. I got to thinking, how hard would it be
> to isolate the hardware dependent portions of a system, and simply
> backup and restore the hardware independent portions onto a new system
> using rsync? Can someone shed some light on the subject? A thinks I'm
> not clear on.
> I'm just trying to avoid a several day restore if someone walks out
> the door with the primary server or it blows up or whatever.
> I'm not using X, so that's not a problem.
> Thanks for the advice.
I think you have a couple of ways to go:
1. You need hot-redundancy so if one blows up the second
will take over immediatly with no downtime.
This is a high-availability solution and you should
focus on this.
However, unless these two servers are separate, the
person who walks out with one may walk out with both.
Remember 9/11. Data centers in one tower backed up to
mirrored centers _in_the_other_tower.
2. You need cold-redundancy. A production server, and a
spare server locked up ready to go. You can have debian
stable re-installed on it and when its needed, hook it
to the net, do an update/upgrade and tweak it.
3. You need prepared media to be able to restore from
bare-metal to a new box that you won't acquire until
You can look at mondo. There's also a package that
makes custom debian CDs that contain only the packages
that you have installed. Use your current box to make
these CDs, and also keep a set of debian-testing CDs in
case the new hardware doesn't boot with stable or the
CDs made on the old hardware. The package CDs will
still contain the packages to save on download time when
you need to install on new hardware.
Keep this bare-metal-recovery media in protected
storage, with a copy with your off-site backup.
I would suggest that whatever else you do, you implement number 3.
Hardware specific stuff is primarily in /etc although who knows what
each package puts in /var when it gets installed. Personlly, I back up
/etc but if/when I have to install to new hardware I do a real install.
If you have the packages locally (e.g. on CD), then it really doesn't
take very long at all (e.g. 30 minutes or so).
If you want to be able to recover from a stolen server within 1 hr, then
you need a spare server locked up somewhere, so that's option 2.
You could also set up a third box to do nothing but store backups. It
can also run one of the apt repositories like apt-proxy so you have
locally everything you need to restore already on-line. It can also
have all the documentation installed ready to go.