Re: [Xen-users] How to Backup and Restore MBR within Logical Volumes?

On Thursday 12 November 2009, you wrote:
> > dd if=/dev/virtualmachines/windows7-x64 of=mbr.w7-x64 bs=512 count=1
>
> I think if you do this, you are only backing up the first 512 bytes of the
> logical volume, not the MBR.
>
> Someone correct me if I am wrong.

I did some tests just to be sure. As far as I can tell, dd interacts with lvm
in exactly the same way as with a physical disk or a loop device.

In the test I copied the first sector directly from the lvm partition or via
the loop device. It results in exactly the same sector being copied.

Also, if you try fdisk -l on the lvm disk or the loop device, it results in
the same output.

Below is the output from my tests:

[root@aragorn:~]# losetup /dev/loop1 /dev/base/kobaltwit_f11_disk

[root@aragorn:~]# fdisk -l /dev/base/kobaltwit_f11_disk

Disk /dev/base/kobaltwit_f11_disk: 10.7 GB, 10737418240 bytes
255 heads, 63 sectors/track, 1305 cylinders

Units = cylinders of 16065 * 512 = 8225280 bytes

                      Device Boot      Start         End      Blocks   Id
System

/dev/base/kobaltwit_f11_disk1   *           1        1305    10482381   83
Linux
[root@aragorn:~]# fdisk -l /dev/base/kobaltwit_f11_disk

Disk /dev/base/kobaltwit_f11_disk: 10.7 GB, 10737418240 bytes
255 heads, 63 sectors/track, 1305 cylinders

Units = cylinders of 16065 * 512 = 8225280 bytes

                      Device Boot      Start         End      Blocks   Id
System

/dev/base/kobaltwit_f11_disk1   *           1        1305    10482381   83
Linux

[root@aragorn:~]# dd if=/dev/base/kobaltwit_f11_disk of=mbr.lvm bs=512 count=1
1+0 records in
1+0 records out
512 bytes (512 B) copied, 7.6e-05 seconds, 6.7 MB/s

[root@aragorn:~]# dd if=/dev/loop1 of=mbr.loop bs=512 count=1
1+0 records in
1+0 records out
512 bytes (512 B) copied, 0.000151 seconds, 3.4 MB/s

[root@aragorn:~]# diff mbr.l*
[root@aragorn:~]#

These test seem to indicate to me that the lvm layer in completely transparent
to userland tools such as fdisk or dd.

So I still think the losetup step is superfluous and possibly causing
unnecessary overhead.

Geert

--
Kobalt W.I.T.
Web & Information Technology
Brusselsesteenweg 152
1850 Grimbergen

Tel  : +32 479 339 655
Email: info@xxxxxxxxxxxx