[vdo-devel] VDO over lvmthin?

[Sweet Tea Dorminy]

Greetings;

Unfortunately still true. While VDO not passing down discards is one
issue, it can be avoided by not thin provisioning the dm-thin device.
Additionally, though, with thin default block size of 256k or even
64k, VDO's 4k block size means snapshots will need to do a lot of IO
-- a 64k read and 64k write when breaking sharing between one snapshot
and another for 64k, etc. I'm not entirely sure whether this
eliminates the performance benefits of thin-based snapshots. I agree
having a snapshotable underlying block device is useful for backups,
and it would be great if there were a fast way to do so; I'm just not
clear which snapshot solution is faster in VDO's case (disregarding
the discard issue).

Apologies,

Sweet Tea


On Wed, Aug 4, 2021 at 4:04 PM Gionatan Danti <g.danti at assyoma.it> wrote:
>
> Hi all,
> does VDO play nicer with lvmthin nowadays, or its use over a thin volume
> is still discouraged?
>
> Full disclosure: I remember that the main issue with VDO over lvm thin
> is that VDO does not pass down discards, effectively forcing the
> underlying volume to always grow and never shrink. For that reason VDO
> should be used over LVM fat/classical volumes. On the other hand, having
> a snapshot-able underlying block device would be helpful in case of VDO
> itself going rouge (ie: due to some metadata corruption). Classical LVM
> snapshot are extremely slow and not appropriate for this task, hence the
> question about thin volumes as the base block device.
>
> Regards.
>
> --
> Danti Gionatan
> Supporto Tecnico
> Assyoma S.r.l. - www.assyoma.it
> email: g.danti at assyoma.it - info at assyoma.it
> GPG public key ID: FF5F32A8
>
> _______________________________________________
> vdo-devel mailing list
> vdo-devel at redhat.com
> https://listman.redhat.com/mailman/listinfo/vdo-devel
>