[nSLUG] bare metal recovery

D G Teed donald.teed at gmail.com
Fri Jul 17 13:12:28 ADT 2009


On Fri, Jul 17, 2009 at 11:56 AM, Hatem Nassrat<hnassrat at gmail.com> wrote:
> On Fri, Jul 17, 2009 at 10:56:30AM -0300, D G Teed wrote:
>> I've heard a suggested recovery technique is to install a bare minimal
>> install to the swap partition, boot that, install the backup/recover
>
> I thought swap partition is volatile, I wouldn't think you ca use it for
> this purpose. Ofcourse my experience is quite limited, but to me this is
> new stuff.

The partition which was normally swap is borrowed for a fresh install.
Later it can be made back into swap.

>> software, and restore to the original partitions, run the boot loader
>> tool, fix fstab, etc.
>
> I think if you run a live CD or USB it may be easier.

Yes, or network boot.  If you have backup software which supports
this type of recovery environment.

> [...]
>> on a live system.  Has anyone else managed to recover the
>> root partition from their backup software while running on the
>> same root?
>
> If you have a copy of "/" backed up soemwhere, the recomended strategy
> is to load up a live cd, or plug in the drive that the data will be
> pushed to on another running machine (ofcourse you need a way to get to
> the backup data, netwrok (rsync) or a burnt DVD maybe options). Mount
> the drive, e.g.
>
>    sudo -s
>    mkdir /mnt/brokendisk
>    mkdir /mnt/backupdisk
>    mount /dev/sda2 /mnt/brokendisk
>    mount /dev/sdb2 /mnt/backupdisk
>    rsync -av --progress /mnt/backupdisk/ /mnt/brokendisk/
>
> ofcourse changing the above to suit your layout.
>

That would work in some situations, but for enterprise backup
solutions your backup is on a tape library, and you have some
hardware with no USB support, etc.

I know there are ways to image and restore disks as snapshots or
templates, but I'm wondering if anyone has used backup/recovery
software to overwrite the system they are running on.  i.e. was
I just lucky to have this work on my Solaris 7 experience?
Assuming things like /dev and /proc are not touched by a
recovery, why should a machine go down when files in /
are updated?  Perhaps /etc/mnttab could be sensitive.
When I have time I'll experiment with this, but I thought someone
here might know.

--Donald



More information about the nSLUG mailing list