Re: LVM vs. Ext4 snapshots (was: [PATCH v1 00/30] Ext4 snapshots)

[Amir G.]

On Sat, Jun 11, 2011 at 10:35 AM, Joe Thornber <thornber@xxxxxxxxxx> wrote:
> On Sat, Jun 11, 2011 at 08:41:38AM +0300, Amir G. wrote:
>> On Fri, Jun 10, 2011 at 1:11 PM, Joe Thornber <thornber@xxxxxxxxxx> wrote:
>> > On Fri, Jun 10, 2011 at 11:01:41AM +0200, Lukas Czerner wrote:
>> >> On Fri, 10 Jun 2011, Amir G. wrote:
>> >>
>> >> > CC'ing lvm-devel and fsdevel
>> >> >
>> >> >
>> >> > On Wed, Jun 8, 2011 at 9:26 PM, Amir G. <amir73il@xxxxxxxxxxxxxxxxxxxxx> wrote:
>> >> > For the sake of letting everyone understand the differences and trade
>> >> > offs between
>> >> > LVM and ext4 snapshots, so ext4 snapshots can get a fair trial, I need
>> >> > to ask you
>> >> > some questions about the implementation, which I could not figure out by myself
>> >> > from reading the documents.
>> >
>> > First up let me say that I'm not intending to support writeable
>> > _external_ origins with multisnap.  This will come as a suprise to
>> > many people, but I don't think we can resolve the dual requirements to
>> > efficiently update many, many snapshots when a write occurs _and_ make
>> > those snapshots quick to delete (when you're encouraging people to
>> > take lots of snapshots performance of delete becomes a real issue).
>> >
>>
>> If I understand this article correctly:
>> http://people.redhat.com/mpatocka/papers/shared-snapshots.pdf
>> It says that _external_ origin write updates can be efficient to readonly
>> (or not written) snapshots.
>>
>> Could you not support readonly snapshots of an _external_ origin?
>
> Yes, that is the intention, and very little work to add.  We just do
> something different if the metadata lookup returns -ENODATA.  Above I
> said I didn't intend to support _writeable_ external snaps.  Readable
> ones are a must, for instance for supporting virtual machine base
> images.
>
>> You could even support writable snapshots, that will degrade write
>> performance to origin temporarily.
>> It can be useful, if one wants to "try-out" mounting a temporary
>> writable snapshot, when the origin is not even mounted.
>> After the "try-out", the temporary snapshot can be deleted
>> and origin write performance would go back to normal.
>
> Not sure what you're getting at here.  All snapshots are writeable.
>

I meant _readonly_ snapshots of a _writable_ _external_ origin,
which is what ext4 snapshots provides.
All snapshots are chained on a list that points to the origin and
only the latest (active) snapshot metadata get updated on origin writes.
When older snapshots lookup return -ENODATA, you go up the list
to the newer snapshot and up to the origin.

Those _incremental_ snapshots cannot be _writable_, because older
snapshots may implicitly share blocks with newer snapshots, but it should
be possible to make _writable_ clones of these snapshots.
Not sure what the implications are for deleting snapshots, because I am
not familiar with all the implementation details of multisnap.


> Of course you can take a snapshot of an external origin and then use
> this as your temporary origin for experiments.  If the origin is
> itself a dm device then LVM can shuffle tables around to make this
> transparent.
>
> The user may want to commit to their experiment at a later time by
> merging back to the external origin.  This involves copying, but no
> more than a copy-on-write scheme.  Arguably it's better to do the copy
> only once we know they want to commit to it.
>
> - Joe
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/