timestamp for shared snapshot

timestamp for shared snapshot

[devicemapper niko]

[-- Attachment #1.1: Type: text/plain, Size: 152 bytes --]

Hi,guys.

I tried shared snapshot recently,I found there's no timestamp for snapshot.

Is it possible to add timestamp and comment for each snapshots ?

[-- Attachment #1.2: Type: text/html, Size: 218 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: timestamp for shared snapshot

[Mike Snitzer]

On Wed, Feb 16 2011 at  4:42am -0500,
devicemapper niko <devicemapper.niko@gmail.com> wrote:

> Hi,guys.
> 
> I tried shared snapshot recently,I found there's no timestamp for snapshot.
> 
> Is it possible to add timestamp and comment for each snapshots ?

I assume you're using Mikulas' shared snapshot kernel patches? E.g.:
http://people.redhat.com/mpatocka/patches/kernel/new-snapshots/r23/

And a patched lvm2 with shared snapshot support?  E.g.:
http://people.redhat.com/mpatocka/patches/userspace/new-snapshots/lvm-2.02.73/

It should be noted that the above shared snapshot work is unlikely to
make it upstream (portions of it likely will, for instance the lvm2
userspace changes).  A new shared snapshot implementation is in
development.

In any case, the easiest thing to do is use descriptive names for each
snapshot (include your own timestamp, etc).

Mike

Re: timestamp for shared snapshot

[Alasdair G Kergon]

> > Is it possible to add timestamp and comment for each snapshots ?

Good idea - as a userspace lvm2 feature, yes.

We currently store creation details for each VG.
I agree we should also store such details for each LV.
This should be optional, but default to enabled.
(People short of space can perhaps turn it off, discarding any existing
details.)

Storing a comment is perhaps an extension to tag support that's been discussed,
to allow longer tags with a wider character set.

> It should be noted that the above shared snapshot work is unlikely to
> make it upstream (portions of it likely will, for instance the lvm2
> userspace changes).  A new shared snapshot implementation is in
> development.

Different snapshot implementations are optimised for different types of usage.
Each one involves choosing between different trade-offs.  We're unlikely ever
to have one solution that is optimal for everyone.

In that spirit, I hope to be submitting more than one shared snapshot
implementation upstream, including *both* of the ones referred to in this
message.  But I want the interfaces used by these implementations to be
basically compatible so that the same userspace tools can make use of them.

Alasdair

Re: timestamp for shared snapshot

[devicemapper niko]

[-- Attachment #1.1: Type: text/plain, Size: 1576 bytes --]

On Thu, Feb 17, 2011 at 1:55 AM, Alasdair G Kergon <agk@redhat.com> wrote:

> > > Is it possible to add timestamp and comment for each snapshots ?
>
> Good idea - as a userspace lvm2 feature, yes.
>
> We currently store creation details for each VG.
> I agree we should also store such details for each LV.
> This should be optional, but default to enabled.
> (People short of space can perhaps turn it off, discarding any existing
> details.)
>
> Storing a comment is perhaps an extension to tag support that's been
> discussed,
> to allow longer tags with a wider character set.
>
U mean saving the snapshots  timestamp infomation in VGDA?
It's not good i think. hundreds of snaps may be tagged but not attached
using dmsetup create command.
saving attached snaps infomation in VGDA is more reasonable


> > It should be noted that the above shared snapshot work is unlikely to
> > make it upstream (portions of it likely will, for instance the lvm2
> > userspace changes).  A new shared snapshot implementation is in
> > development.
>
> Different snapshot implementations are optimised for different types of
> usage.
> Each one involves choosing between different trade-offs.  We're unlikely
> ever
> to have one solution that is optimal for everyone.
>
> In that spirit, I hope to be submitting more than one shared snapshot
> implementation upstream, including *both* of the ones referred to in this
> message.  But I want the interfaces used by these implementations to be
> basically compatible so that the same userspace tools can make use of them.
>
> Alasdair
>
>

[-- Attachment #1.2: Type: text/html, Size: 2114 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]