[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [lvm-devel] lvcreate -s - why specifying size for snapshot ?



-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Sandeep K Sinha wrote:
> On Fri, Mar 20, 2009 at 9:48 PM, Jonathan Brassow <jbrassow redhat com> wrote:
>> On Mar 20, 2009, at 11:10 AM, Bryn M. Reeves wrote:
>>
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA1
>>>
>>> Sandeep K Sinha wrote:
>>>> Hi all,
>>>> Disk
>>>> After hunting for relevant documents and failing to find one.
>>>> I would like to know if its not true that the size of the snapshot
>>>> should be equal to the original volume?
>>>> If so, then why do we accept size for a snapshot?
>>>>
>>>> Or Am I missing something somewhere else?
>>>>
>>> The size of the snapshot governs the amount of space set aside for
>>> storing the changes to the origin volume. E.g. if you made a snapshot
>>> and then completely overwrote the origin the snapshot would have to be
>>> at least as big as the origin volume to hold the changes.
>>>
> Agreed to all the explainations, but then I would like to know why am
> I able to succeed in creating a snapshot > original volume.
> LVM should puke an error for it right.

Not at all - going back to the original example: suppose you created a
snapshot and then completely overwrite the origin volume *twice*. Now
the snapshot must be twice the size of the origin - one to hold the data
that was originally on the origin and one to hold the data from the
first overwrite.

There is no real or apparent relationship between the size of the origin
and the size of the snapshot: it only has to do with the quantity of
change you need to record.

> Or is our snapshot like a CDP? I dont thinks so.

Continuous data protection? A snapshot is very similar to CDP but it is
just tracking the state of the origin at one point in time (the time at
which it was created). You don't have the ability to roll back to
arbitrary intermediate states but the state of all blocks on the device
(snapshot chunks) at the time the snapshot was created is tracked so
long as there is enough space in the exception store.

> IMO, extending lvm to have a block level CDP as a target should be a
> good option.

The replicator target addresses some of the use-cases for CDP and aims
to provide write-order fidelity for the replicated copies. See the
project pages here:

http://people.redhat.com/heinzm/sw/dm/dm-replicator/

Regards,
Bryn.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iEYEARECAAYFAknDzT0ACgkQ6YSQoMYUY94P4wCgt8Efxd+ebZGEfHTkEZVDd7Qk
mRoAoLbDvatbIMZpZlAb2XehkerQadwE
=n+/Y
-----END PGP SIGNATURE-----


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]