Hi Andy,
Am 11.05.2026 um 19:49 schrieb Andy Smith:
Hi,
On Mon, May 11, 2026 at 02:13:42PM +0200, Arno Lehmann wrote:
an LVM snapshot is *not* an exact copy of the original volume, and of
course much less so of the partition, disk, or anything else. So a
block level dump of an LVM snapshot is not something you can just dump
back as a disaster recovery step.
Can you expand upon this?
You might have seen my earlier reply to that same question.
However, re-thinking things here, I would say I fell into that same trap
I tried to navigate around in mixing different levels or layers of
functionality.
Let me try to clarify.
It's not my experience. I use LVM snapshots
pretty much every week and in every case they are bit for bit copies of
the block device they are a snapshot of.
I'm not doubting that -- and that is because in this situation we talk
about the block device as provided by the lvm system. Which is not quite
the same thing as the blocks on disk without the lvm view around it.
Any time I have reverted the
block device to the snapshot, it has been as if that data was put back
on there any other way.
That, I think, we are all quite sure of these days.
Do you mean, for example, the bits that live
outside of block devices? Such as bootloaders, partition tables and the
like?
No, for those I wouldn't claim such things. I was actually thinking
about a different thing; my actual experience was with things such as
block size detection and alignment the file system "inherits" from the
LV properties and which may not necessarily be identical to what can be
found in a different LVM environment or when dumping a snapshot's block
dump to a physical disk. Or partition.
And there we are again back to the original question -- because there's
some ambiguity involved when you talk about "imaging a drive in a server
for DR purposes". The block storage backing a file system is quite a
different thing than an actual partition, a complete disk (and what do
we do if the disk is a member of a RAID array?) and to be successful
when restoring that way, you also need at least a boot loader, if not an
(U)EFI partition as well.
Well. Easy to get lost in details, and I suspect this reminds me why I
prefer my backups done differently.
Anyway -- let me rephrase my statement to be more correct but probably
also more vague: The dump of the snapshot of an LV is surely going to be
bit-by-bit identical to the origin LV at the time the snapshot was
taken, but as a means for recovery, it may depend on a bit of
infrastructure that is not part of phase one of a bare-metal recovery,
i.e. server with block storage only.
(I've already spoken about how a block device snapshot is limited and
agree that it doesn't solve the whole problem here. I'm not able to
agree that you "can't just dump a snapshot back". I do it all the time!)
I only do it occasionally, but it needs more than just an empty disk to
be useful.
Which allows me to circle back to the original question: Besides the
backup, which could be a dump of /something/, you'll always need a bit
of infrastructure, tools and a plan and when you start solving those,
the question of how the original backup is done is usually of secondary
importance.
Is that something you could agree with? ;-)
Cheers,
Arno
Thanks,
Andy
--
Arno Lehmann
IT-Service Lehmann
Sandstr. 6, 49080 Osnabrück
