[linux-lvm] snapshots

[linux-lvm] snapshots

[Robert Dyas]

Snapshot questions:

1)Has anyone had any luck creating a snapshot (lvcreate -L100M -s -n backup
/dev/vg_master/lv_home) and then mounting it (mount -r /dev/vg_master/backup
/mnt)? If so, what version of the kernel, mount tools, filesystem, etc are
you using?

2) How can you use the snapshot feature to facilitate incremental or
differential backups? Wouldn't the dump or other backup program have to
twiddle the archive bit, but won't be able to since snapshots are readonly?
Any insight here would be appreciated.

Thanks!

Rob

Re: [linux-lvm] snapshots

[Andreas Dilger]

On Oct 18, 2001  16:35 -0400, Robert Dyas wrote:
> 1)Has anyone had any luck creating a snapshot (lvcreate -L100M -s -n backup
> /dev/vg_master/lv_home) and then mounting it (mount -r /dev/vg_master/backup
> /mnt)? If so, what version of the kernel, mount tools, filesystem, etc are
> you using?

Yes, but I don't know the exact versions.

> 2) How can you use the snapshot feature to facilitate incremental or
> differential backups? Wouldn't the dump or other backup program have to
> twiddle the archive bit, but won't be able to since snapshots are readonly?

Um, "archive bit" is a severe DOS-ism, which doesn't exist in Unix.  An
incremental backup on Unix is based on the file modification timestamp
which does not need to be changed when doing the backup.

Cheers, Andreas
--
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

[linux-lvm] Snapshots

[jonr]

Hello List,

I don't get it!

I have a xen system that has a DomU of FC11 with a 20GB LV for the  
disk. I created a snapshot using this command:

lvcreate -L +10G -s -n s_cgate-be1 /dev/xenvg/cgate-be1

This is the output of an 'lvdisplay' of the snapshot LV:

lvdisplay /dev/xenvg/s_cgate-be1
   --- Logical volume ---
   LV Name                /dev/xenvg/s_cgate-be1
   VG Name                xenvg
   LV UUID                7c1kEs-xYoO-Qxem-2E2M-hlee-pc4Q-zm6TZ5
   LV Write Access        read/write
   LV snapshot status     active destination for /dev/xenvg/cgate-be1
   LV Status              available
   # open                 2
   LV Size                20.00 GB
   Current LE             5120
   COW-table size         10.00 GB
   COW-table LE           2560
   Allocated to snapshot  0.11%
   Snapshot chunk size    4.00 KB
   Segments               1
   Allocation             inherit
   Read ahead sectors     auto
   - currently set to     256
   Block device           253:15

I then shutdown the DomU and pointed the 'Disk...' line in the config  
file for the DomU to the snapshot. It boots.

OK, I must be missing the simple answer because I cannot seem to grasp  
the concept of snapshots.


1. How can I have a 20GB LV as a disk and the snapshot be 10GB and  
boot the entire OS?

2. Can I create a DomU and then snapshot the LV and use the snapshot  
to create other DomU's?

3. If 2 is yes, would I want to continue using the snapshot as the  
disk or is there something else that should be done, i.e. dd the drive  
to a new LV?

In the 'lvdisplay' of the snapshot I have these extra lines:
   LV snapshot status     active destination for /dev/xenvg/cgate-be1
   # open                 2
   COW-table size         10.00 GB
   COW-table LE           2560
   Allocated to snapshot  0.11%
   Snapshot chunk size    4.00 KB

4. Does the 0.11% mean that the 10GB snapshot file is only using .11%  
of the 10GB LV?


If there is a good doc out there explaining all of this could someone  
just post a link. I don't mind reading but am not getting the answers  
to the above. I have found lots of docs explaining how to do it but  
nothing going much further than that.

Thanks for any help,

Jon

Re: [linux-lvm] Snapshots

[Peter Keller]

Hi Jon,

You seem to be right that there isn't much written down about how snapshots 
work under the bonnet in the LVM context, but the same ideas have been used 
for other snapshot implementations for years. You could try searching for 
filesystem snapshot technology in general, rather than anything 
LVM-specific. For example, this article explains some of the basic concepts:

   http://www.ibm.com/developerworks/tivoli/library/t-snaptsm1/index.html

LVM uses the copy-on-write method (this is what "COW" stands for).

On Thu, 10 Sep 2009, jonr@destar.net wrote:

> Hello List,
>
> I don't get it!
>
> I have a xen system that has a DomU of FC11 with a 20GB LV for the disk. I 
> created a snapshot using this command:
>
> lvcreate -L +10G -s -n s_cgate-be1 /dev/xenvg/cgate-be1
>
> This is the output of an 'lvdisplay' of the snapshot LV:
>
> lvdisplay /dev/xenvg/s_cgate-be1
> --- Logical volume ---
> LV Name                /dev/xenvg/s_cgate-be1
> VG Name                xenvg
> LV UUID                7c1kEs-xYoO-Qxem-2E2M-hlee-pc4Q-zm6TZ5
> LV Write Access        read/write
> LV snapshot status     active destination for /dev/xenvg/cgate-be1
> LV Status              available
> # open                 2
> LV Size                20.00 GB
> Current LE             5120
> COW-table size         10.00 GB
> COW-table LE           2560
> Allocated to snapshot  0.11%
> Snapshot chunk size    4.00 KB
> Segments               1
> Allocation             inherit
> Read ahead sectors     auto
> - currently set to     256
> Block device           253:15
>
> I then shutdown the DomU and pointed the 'Disk...' line in the config file 
> for the DomU to the snapshot. It boots.
>
> OK, I must be missing the simple answer because I cannot seem to grasp the 
> concept of snapshots.
>
>
> 1. How can I have a 20GB LV as a disk and the snapshot be 10GB and boot the 
> entire OS?

Because the snapshot only stores copies of the disk blocks that have been 
modified on the original volume since the snapshot was taken (the blocks 
that would otherwise be overwritten if you hadn't created the snapshot). 
When you read from the snapshot, LVM uses never-modified blocks from the 
original volume and pre-modification copies of blocks from the snapshot as 
needed.

If the original volume became corrupt, the snapshot would also be unusable.

> 2. Can I create a DomU and then snapshot the LV and use the snapshot to 
> create other DomU's?

Yes - there are articles around about how to do this. If you intend to run 
the DomU's on the same network, you should boot the copies into single-user 
mode initially to modify their network identities so that they don't 
conflict with each other. (Maybe you can find some other way of doing this 
without booting the DomU's: I know that you can do this with VMWare, at 
least for Linux guest OS's.)

> 3. If 2 is yes, would I want to continue using the snapshot as the disk or is 
> there something else that should be done, i.e. dd the drive to a new LV?
>
> In the 'lvdisplay' of the snapshot I have these extra lines:
> LV snapshot status     active destination for /dev/xenvg/cgate-be1
> # open                 2
> COW-table size         10.00 GB
> COW-table LE           2560
> Allocated to snapshot  0.11%
> Snapshot chunk size    4.00 KB

You can continue use the snapshot in principle, although you might like to 
make the snapshot the same size as the original volume in that case to make 
sure that it never fills up. In your example, when the accumulated changes 
on the original volume since the snapshot was taken reach 10Gb, the snapshot 
will be full and go invalid and you won't be able to use it.

If your aim is to have a single starting point to create more than one DomU, 
you should copy the data out and keep it safe somewhere. You can then use 
that copy to create further copies as needed. Personally, I would only boot 
up the first copy (or the snapshot) to keep the OS up to date (which will 
save you having to update every copy that you make from it).

> 4. Does the 0.11% mean that the 10GB snapshot file is only using .11% of the 
> 10GB LV?

Yes, that's right.

> If there is a good doc out there explaining all of this could someone just 
> post a link. I don't mind reading but am not getting the answers to the 
> above. I have found lots of docs explaining how to do it but nothing going 
> much further than that.
>
> Thanks for any help,

Have fun,
Peter.

Re: [linux-lvm] Snapshots

[André Gillibert]

jonr@destar.net wrote:
> Hello List,
> 
> I don't get it!
> 
> I have a xen system that has a DomU of FC11 with a 20GB LV for the  
> disk. I created a snapshot using this command:
> 
> lvcreate -L +10G -s -n s_cgate-be1 /dev/xenvg/cgate-be1
> 
> This is the output of an 'lvdisplay' of the snapshot LV:
> 
> lvdisplay /dev/xenvg/s_cgate-be1
>    --- Logical volume ---
>    LV Name                /dev/xenvg/s_cgate-be1
>    VG Name                xenvg
>    LV UUID                7c1kEs-xYoO-Qxem-2E2M-hlee-pc4Q-zm6TZ5
>    LV Write Access        read/write
>    LV snapshot status     active destination for /dev/xenvg/cgate-be1
>    LV Status              available
>    # open                 2
>    LV Size                20.00 GB
>    Current LE             5120
>    COW-table size         10.00 GB
>    COW-table LE           2560
>    Allocated to snapshot  0.11%
>    Snapshot chunk size    4.00 KB
>    Segments               1
>    Allocation             inherit
>    Read ahead sectors     auto
>    - currently set to     256
>    Block device           253:15
> 
> I then shutdown the DomU and pointed the 'Disk...' line in the config  
> file for the DomU to the snapshot. It boots.
> 
> OK, I must be missing the simple answer because I cannot seem to grasp  
> the concept of snapshots.
> 
> 
> 1. How can I have a 20GB LV as a disk and the snapshot be 10GB and  
> boot the entire OS?
> 
That's because the snapshot uses these 10GB as a "COW-table". i.e. A "Copy On Write table".
When the snapshot is first created, no copy is performed. All the blocks of the snapshot are identical to the blocks of the original LV. Accessing the snapshot or original LV access the same block.
Then, if you modify some data in either the original LV or the snapshot LV, an "exception list" listing data chunks that are different in the snapshot LV than in the original LV, is created.
When reading a snapshot LV block, LVM first looks at the "exception list". If the chunk accessed is the in the "exception list", the chunk data in the exception list is returned. If the chunk isn't in the exception list, the chunk data is extracted from the original volume.
When writing a snapshot LV block, it's written in a chunk of the exception list, either existing, or created at this time.
When writing a chunk in the original volume, it's first copied in the "exception list", and then, modified in the original volume.

The chunk size is specified with the --chunksize parameter (4k to 512k) to lvcreate(8), when creating a snapshot.

If ever you've so many exceptions that the snapshot volume overflows, it's dropped.
Consequently, you must take care of allocating enough space in snapshots. This space is the size of the data that may be modified in the snapshot or original volume.
Snapshots are fine for short lived snapshots (e.g. when backing up) but aren't designed to cope with heavy modifications of the entire file system.

> 2. Can I create a DomU and then snapshot the LV and use the snapshot  
> to create other DomU's?

I'm not a Xen user. Sorry, I don't know.
Here, Google may help:
<http://www.google.com/search?rls=en&hl=en&source=hp&q=xen+device+mapper>

> 
> 3. If 2 is yes, would I want to continue using the snapshot as the  
> disk or is there something else that should be done, i.e. dd the drive  
> to a new LV?
> 
> In the 'lvdisplay' of the snapshot I have these extra lines:
>    LV snapshot status     active destination for /dev/xenvg/cgate-be1
>    # open                 2
>    COW-table size         10.00 GB
>    COW-table LE           2560
>    Allocated to snapshot  0.11%
>    Snapshot chunk size    4.00 KB
> 
> 4. Does the 0.11% mean that the 10GB snapshot file is only using .11%  
> of the 10GB LV?
> 
It means that the internal LVM data structure + the exception chunks are using 0.11% of the 10GB "COW-table partition".

Create snapshot, dd(1), delete snapshot is a way to create a "forked" copy of the original volume.
That may, or may not be what you want. If you intend to share much data with the original volume, a simple snapshot may be the way to go. If a small or near-zero amount of data is to be shared, in that case, snapshotting won't save data space, but may incur performance penalties (see <http://www.nikhef.nl/~dennisvd/lvmcrap.html>), and so, my advice is to create snapshot, dd(1), and delete snapshot.

I hope it helps.
-- 
André Gillibert

[linux-lvm] Snapshots

[Jon Hardcastle]

Hi,

Do I need some free extents to create a READ ONLY snapshot? I dont have free extents but require a read-only snapshot before doing an upgrade.

Cheers.


-----------------------
N: Jon Hardcastle
E: Jon@eHardcastle.com
'Do not worry about tomorrow, for tomorrow will bring worries of its own.'
-----------------------


      

Re: [linux-lvm] Snapshots

[Peter Keller]

Dear John,

On Thu, 17 Sep 2009, Jon Hardcastle wrote:

> Hi,
>
> Do I need some free extents to create a READ ONLY snapshot? I dont have 
> free extents but require a read-only snapshot before doing an upgrade.

Extents are written to the snapshot when changes take place on the volume 
that the snapshot is based on. If you have no free extents, trying to create 
a snapshot strikes me as pointless: you may as well mount the origin volume 
read-only and back up its contents directly (or unmount it and back it up at 
the level of the volume). Or do you have some special circumstances?

If you make the snapshot read-only, as soon as changes occur on the origin 
volume the snapshot will become useless: some of the data that are required 
to make the snapshot usable (i.e. unchanged copies of altered blocks on the 
origin volume)  will no longer exist.

Regards,
Peter.


-- 
Peter Keller                                     Tel.: +44 (0)1223 353033
Global Phasing Ltd.,                             Fax.: +44 (0)1223 366889
Sheraton House,
Castle Park,
Cambridge CB3 0AX
United Kingdom

[linux-lvm] Snapshots

[Mark Woodward]

I have been looking into LVM2 for a while now and while I think it is 
useful for a range of applications, its seems pretty limited in overall 
scope.

Is there any active development happening? Most of the code seems like 
bug-fixes or minor tweaks. Is the snapshot system being improved? 
Specifically, snapshots of snapshots? thin provisioned snapshots or 
auto-expand (i.e. snapshots don't run out of space).  If you look at 
technologies like ZFS snapshots are far better supported, but ZFS will 
probably never be real under Linux, and it really is far more than is 
really needed.

So, I guess my questions are these: Is LVM in maintenance mode or is 
there active development? If it is being actively developed, is there a 
road map and is there a group or site specifically dedicated to the 
development?

Re: [linux-lvm] Snapshots

[Mike Snitzer]

On Thu, Feb 16 2012 at  8:25am -0500,
Mark Woodward <markw@mohawksoft.com> wrote:

> I have been looking into LVM2 for a while now and while I think it
> is useful for a range of applications, its seems pretty limited in
> overall scope.
> 
> Is there any active development happening? Most of the code seems
> like bug-fixes or minor tweaks. Is the snapshot system being
> improved? Specifically, snapshots of snapshots? thin provisioned
> snapshots or auto-expand (i.e. snapshots don't run out of space).
> If you look at technologies like ZFS snapshots are far better
> supported, but ZFS will probably never be real under Linux, and it
> really is far more than is really needed.
> 
> So, I guess my questions are these: Is LVM in maintenance mode or is
> there active development? If it is being actively developed, is
> there a road map and is there a group or site specifically dedicated
> to the development?

Um, where are you even getting this idea that LVM2 is in maintenance
mode?  Or that snapshots haven't improved?

Sorry to come off defensive but your entire post is founded on incorrect
understanding.

Anyway, if you look at the change history of the lvm2 repository (be it
cvs or git, cvs commits are mirrored to git) you'll see there have been
regular changes flowing in and most recently a very extensive evolution
of the code to add support for thin provisioning with highly efficient
snapshots (ala btrfs or ZFS).

try: git clone git://sources.redhat.com/git/lvm2

take a look at the WHATS_NEW file and tell me what you think.

Mike

Re: [linux-lvm] Snapshots

[Marco Pizzoli]

[-- Attachment #1: Type: text/plain, Size: 1715 bytes --]

Hi Mike

On Thu, Feb 16, 2012 at 2:57 PM, Mike Snitzer <snitzer@redhat.com> wrote:

> On Thu, Feb 16 2012 at  8:25am -0500,
> Mark Woodward <markw@mohawksoft.com> wrote:
>
> > I have been looking into LVM2 for a while now and while I think it
> > is useful for a range of applications, its seems pretty limited in
> > overall scope.
> >
> > Is there any active development happening? Most of the code seems
> > like bug-fixes or minor tweaks. Is the snapshot system being
> > improved? Specifically, snapshots of snapshots? thin provisioned
> > snapshots or auto-expand (i.e. snapshots don't run out of space).
> > If you look at technologies like ZFS snapshots are far better
> > supported, but ZFS will probably never be real under Linux, and it
> > really is far more than is really needed.
> >
> > So, I guess my questions are these: Is LVM in maintenance mode or is
> > there active development? If it is being actively developed, is
> > there a road map and is there a group or site specifically dedicated
> > to the development?
>
> Um, where are you even getting this idea that LVM2 is in maintenance
> mode?  Or that snapshots haven't improved?
>
> Sorry to come off defensive but your entire post is founded on incorrect
> understanding.
>
> Anyway, if you look at the change history of the lvm2 repository (be it
> cvs or git, cvs commits are mirrored to git) you'll see there have been
> regular changes flowing in and most recently a very extensive evolution
> of the code to add support for thin provisioning with highly efficient
> snapshots (ala btrfs or ZFS).
>
> try: git clone git://sources.redhat.com/git/lvm2
>

Please, could you tell me if is there a gitweb interface somewhere?

Thanks
Marco

[-- Attachment #2: Type: text/html, Size: 2273 bytes --]

Re: [linux-lvm] Snapshots

[Mike Snitzer]

On Thu, Feb 16 2012 at  9:05am -0500,
Marco Pizzoli <marco.pizzoli@gmail.com> wrote:

> Hi Mike
> 
> On Thu, Feb 16, 2012 at 2:57 PM, Mike Snitzer <snitzer@redhat.com> wrote:
> 
> > On Thu, Feb 16 2012 at  8:25am -0500,
> > Mark Woodward <markw@mohawksoft.com> wrote:
> >
> > > I have been looking into LVM2 for a while now and while I think it
> > > is useful for a range of applications, its seems pretty limited in
> > > overall scope.
> > >
> > > Is there any active development happening? Most of the code seems
> > > like bug-fixes or minor tweaks. Is the snapshot system being
> > > improved? Specifically, snapshots of snapshots? thin provisioned
> > > snapshots or auto-expand (i.e. snapshots don't run out of space).
> > > If you look at technologies like ZFS snapshots are far better
> > > supported, but ZFS will probably never be real under Linux, and it
> > > really is far more than is really needed.
> > >
> > > So, I guess my questions are these: Is LVM in maintenance mode or is
> > > there active development? If it is being actively developed, is
> > > there a road map and is there a group or site specifically dedicated
> > > to the development?
> >
> > Um, where are you even getting this idea that LVM2 is in maintenance
> > mode?  Or that snapshots haven't improved?
> >
> > Sorry to come off defensive but your entire post is founded on incorrect
> > understanding.
> >
> > Anyway, if you look at the change history of the lvm2 repository (be it
> > cvs or git, cvs commits are mirrored to git) you'll see there have been
> > regular changes flowing in and most recently a very extensive evolution
> > of the code to add support for thin provisioning with highly efficient
> > snapshots (ala btrfs or ZFS).
> >
> > try: git clone git://sources.redhat.com/git/lvm2
> >
> 
> Please, could you tell me if is there a gitweb interface somewhere?

http://sourceware.org/git/?p=lvm2.git

Re: [linux-lvm] Snapshots

[Mark Woodward]

On 02/16/2012 08:57 AM, Mike Snitzer wrote:
> On Thu, Feb 16 2012 at  8:25am -0500,
> Mark Woodward<markw@mohawksoft.com>  wrote:
>
>> I have been looking into LVM2 for a while now and while I think it
>> is useful for a range of applications, its seems pretty limited in
>> overall scope.
>>
>> Is there any active development happening? Most of the code seems
>> like bug-fixes or minor tweaks. Is the snapshot system being
>> improved? Specifically, snapshots of snapshots? thin provisioned
>> snapshots or auto-expand (i.e. snapshots don't run out of space).
>> If you look at technologies like ZFS snapshots are far better
>> supported, but ZFS will probably never be real under Linux, and it
>> really is far more than is really needed.
>>
>> So, I guess my questions are these: Is LVM in maintenance mode or is
>> there active development? If it is being actively developed, is
>> there a road map and is there a group or site specifically dedicated
>> to the development?
> Um, where are you even getting this idea that LVM2 is in maintenance
> mode?  Or that snapshots haven't improved?
>
> Sorry to come off defensive but your entire post is founded on incorrect
> understanding.

Understood, I apologize.
> Anyway, if you look at the change history of the lvm2 repository (be it
> cvs or git, cvs commits are mirrored to git) you'll see there have been
> regular changes flowing in and most recently a very extensive evolution
> of the code to add support for thin provisioning with highly efficient
> snapshots (ala btrfs or ZFS).
I guess I missed that. I need to check out the changes, looks like you 
have auto-grow with monitoring? I was looking more towards the 
"snapshots of snapshots" ability. Is that on the radar?

> try: git clone git://sources.redhat.com/git/lvm2
>
> take a look at the WHATS_NEW file and tell me what you think.
>
> Mike

Re: [linux-lvm] Snapshots

[Stuart D. Gathman]

Long ago, Nostradamus foresaw that on Feb 16, Mark Woodward would write:

> I guess I missed that. I need to check out the changes, looks like you have 
> auto-grow with monitoring? I was looking more towards the "snapshots of 
> snapshots" ability. Is that on the radar?

The ZumaStor 3rd party userland tools support that.

An elegant part of the LVM system is that the device mapper kernel support is
very general, and new data structures can be experimented with entirely in user
code - with a script language even.  Metadata for experimental structures does
not have to stored with the main metadata.

--
 	      Stuart D. Gathman <stuart@bmsi.com>
     Business Management Systems Inc.  Phone: 703 591-0911 Fax: 703 591-6154
"Confutatis maledictis, flammis acribus addictis" - background song for
a Microsoft sponsored "Where do you want to go from here?" commercial.

Re: [linux-lvm] Snapshots

[Mike Snitzer]

On Thu, Feb 16 2012 at 10:18am -0500,
Stuart D. Gathman <stuart@bmsi.com> wrote:

> Long ago, Nostradamus foresaw that on Feb 16, Mark Woodward would write:
> 
> >I guess I missed that. I need to check out the changes, looks like
> >you have auto-grow with monitoring? I was looking more towards the
> >"snapshots of snapshots" ability. Is that on the radar?
> 
> The ZumaStor 3rd party userland tools support that.

Now AFAIK ZumaStor is a dead technology... nobody is actually
maintaining the code upstream.  The shared snapshot store that
dm-thin-pool provides is functionally equivalent to what Zumastor
provides -- dm-thinp provides a single storage pool for all snapshots.
This avoids excessive copy-out that the old dm-snapshot suffered from
due to it requiring a distinct snapshot store for each snapshot.

I'd really like to encourage you and others to evaluate and test
snapshots with dm-thinp -- now that lvm2 provides support it _should_ be
more approachable.

> An elegant part of the LVM system is that the device mapper kernel support is
> very general, and new data structures can be experimented with entirely in user
> code - with a script language even.  Metadata for experimental structures does
> not have to stored with the main metadata.

Please note that the dm-thinp code has metadata in the kernel (on-disk
format for btrees, etc) much like a filesystem would have.  So there is
both kernel and userspace (lvm2) metadata for dm-thinp.

Mike

Re: [linux-lvm] Snapshots

[Stuart D. Gathman]

Long ago, Nostradamus foresaw that on Feb 16, Mike Snitzer would write:

>> An elegant part of the LVM system is that the device mapper kernel support is
>> very general, and new data structures can be experimented with entirely in user
>> code - with a script language even.  Metadata for experimental structures does
>> not have to stored with the main metadata.
>
> Please note that the dm-thinp code has metadata in the kernel (on-disk
> format for btrees, etc) much like a filesystem would have.  So there is
> both kernel and userspace (lvm2) metadata for dm-thinp.

Yes, but isn't this loaded into the kernel via userland tools like
device-mapper?  So while a kernel feature would be required for a new
type of kernel metadata, experimental uses of existing formats can
be done in userland.

--
 	      Stuart D. Gathman <stuart@bmsi.com>
     Business Management Systems Inc.  Phone: 703 591-0911 Fax: 703 591-6154
"Confutatis maledictis, flammis acribus addictis" - background song for
a Microsoft sponsored "Where do you want to go from here?" commercial.

Re: [linux-lvm] Snapshots

[Mike Snitzer]

On Thu, Feb 16 2012 at 11:46pm -0500,
Stuart D. Gathman <stuart@bmsi.com> wrote:

> Long ago, Nostradamus foresaw that on Feb 16, Mike Snitzer would write:
> 
> >>An elegant part of the LVM system is that the device mapper kernel support is
> >>very general, and new data structures can be experimented with entirely in user
> >>code - with a script language even.  Metadata for experimental structures does
> >>not have to stored with the main metadata.
> >
> >Please note that the dm-thinp code has metadata in the kernel (on-disk
> >format for btrees, etc) much like a filesystem would have.  So there is
> >both kernel and userspace (lvm2) metadata for dm-thinp.
> 
> Yes, but isn't this loaded into the kernel via userland tools like
> device-mapper?

The 'thin-pool' and 'thin' DM device tables are loaded from userspace
via DM interfaces.

The kernel metadata isn't loaded from userspace.  It is created and/or
changed by certain actions taken from userspace (via DM messages).

> So while a kernel feature would be required for a new
> type of kernel metadata, experimental uses of existing formats can
> be done in userland.

The kernel manages the kernel's metadata.  But the LVM metadata that
userspace uses to coordinate and manage the thin devices can be
changed independently.

The thin-provisioning-tools know the kernel's metadata format and can
check it (and in the future repair it).

Re: [linux-lvm] Snapshots

[Bryn M. Reeves]

On 02/16/2012 03:18 PM, Stuart D. Gathman wrote:
> Long ago, Nostradamus foresaw that on Feb 16, Mark Woodward would write:
> 
>> I guess I missed that. I need to check out the changes, looks like you have 
>> auto-grow with monitoring? I was looking more towards the "snapshots of 
>> snapshots" ability. Is that on the radar?
> 
> The ZumaStor 3rd party userland tools support that.

Has the zumastor stuff been updated lately? The last release I saw was
nearly four years ago and the LVM2 and device-mapper capabilities have
improved since.

Regards,
Bryn.

Re: [linux-lvm] Snapshots

[Mike Snitzer]

On Thu, Feb 16 2012 at  9:42am -0500,
Mark Woodward <markw@mohawksoft.com> wrote:

> On 02/16/2012 08:57 AM, Mike Snitzer wrote:
> >On Thu, Feb 16 2012 at  8:25am -0500,
> >Mark Woodward<markw@mohawksoft.com>  wrote:
> >
> >>I have been looking into LVM2 for a while now and while I think it
> >>is useful for a range of applications, its seems pretty limited in
> >>overall scope.
> >>
> >>Is there any active development happening? Most of the code seems
> >>like bug-fixes or minor tweaks. Is the snapshot system being
> >>improved? Specifically, snapshots of snapshots? thin provisioned
> >>snapshots or auto-expand (i.e. snapshots don't run out of space).
> >>If you look at technologies like ZFS snapshots are far better
> >>supported, but ZFS will probably never be real under Linux, and it
> >>really is far more than is really needed.
> >>
> >>So, I guess my questions are these: Is LVM in maintenance mode or is
> >>there active development? If it is being actively developed, is
> >>there a road map and is there a group or site specifically dedicated
> >>to the development?
> >Um, where are you even getting this idea that LVM2 is in maintenance
> >mode?  Or that snapshots haven't improved?
> >
> >Sorry to come off defensive but your entire post is founded on incorrect
> >understanding.
> 
> Understood, I apologize.
> >Anyway, if you look at the change history of the lvm2 repository (be it
> >cvs or git, cvs commits are mirrored to git) you'll see there have been
> >regular changes flowing in and most recently a very extensive evolution
> >of the code to add support for thin provisioning with highly efficient
> >snapshots (ala btrfs or ZFS).
> I guess I missed that. I need to check out the changes, looks like
> you have auto-grow with monitoring? I was looking more towards the
> "snapshots of snapshots" ability. Is that on the radar?

recursive snapshots of arbitrary depth (snapshot of snapshots ...) are
already supported with the new thinp snapshots.

Please review the upstream Linux >= 3.2 kernel code/docs too, it has
more documentation of the underlying capabilities, e.g.:

Documentation/device-mapper/thin-provisioning.txt

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/device-mapper/thin-provisioning.txt

(the "Introduction" being the most approachable overview)

Re: [linux-lvm] Snapshots

[Mark Woodward]

On 02/16/2012 10:43 AM, Mike Snitzer wrote:
> On Thu, Feb 16 2012 at  9:42am -0500,
> Mark Woodward<markw@mohawksoft.com>  wrote:
>
>> On 02/16/2012 08:57 AM, Mike Snitzer wrote:
>>> On Thu, Feb 16 2012 at  8:25am -0500,
>>> Mark Woodward<markw@mohawksoft.com>   wrote:
>>>
>>>> I have been looking into LVM2 for a while now and while I think it
>>>> is useful for a range of applications, its seems pretty limited in
>>>> overall scope.
>>>>
>>>> Is there any active development happening? Most of the code seems
>>>> like bug-fixes or minor tweaks. Is the snapshot system being
>>>> improved? Specifically, snapshots of snapshots? thin provisioned
>>>> snapshots or auto-expand (i.e. snapshots don't run out of space).
>>>> If you look at technologies like ZFS snapshots are far better
>>>> supported, but ZFS will probably never be real under Linux, and it
>>>> really is far more than is really needed.
>>>>
>>>> So, I guess my questions are these: Is LVM in maintenance mode or is
>>>> there active development? If it is being actively developed, is
>>>> there a road map and is there a group or site specifically dedicated
>>>> to the development?
>>> Um, where are you even getting this idea that LVM2 is in maintenance
>>> mode?  Or that snapshots haven't improved?
>>>
>>> Sorry to come off defensive but your entire post is founded on incorrect
>>> understanding.
>> Understood, I apologize.
>>> Anyway, if you look at the change history of the lvm2 repository (be it
>>> cvs or git, cvs commits are mirrored to git) you'll see there have been
>>> regular changes flowing in and most recently a very extensive evolution
>>> of the code to add support for thin provisioning with highly efficient
>>> snapshots (ala btrfs or ZFS).
>> I guess I missed that. I need to check out the changes, looks like
>> you have auto-grow with monitoring? I was looking more towards the
>> "snapshots of snapshots" ability. Is that on the radar?
> recursive snapshots of arbitrary depth (snapshot of snapshots ...) are
> already supported with the new thinp snapshots.
>
> Please review the upstream Linux>= 3.2 kernel code/docs too, it has
> more documentation of the underlying capabilities, e.g.:
>
> Documentation/device-mapper/thin-provisioning.txt
>
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/device-mapper/thin-provisioning.txt
>
> (the "Introduction" being the most approachable overview)
OK, that's where I missed it. I looked at the snapshots description. It 
is now part of thin-provisioning. Got it. Will it be in the 2.6 kernel 
or will there need to be an upgrade to 3.x series?

Re: [linux-lvm] Snapshots

[Mike Snitzer]

On Thu, Feb 16 2012 at 11:04am -0500,
Mark Woodward <markw@mohawksoft.com> wrote:

> On 02/16/2012 10:43 AM, Mike Snitzer wrote:
> >Please review the upstream Linux>= 3.2 kernel code/docs too, it has
> >more documentation of the underlying capabilities, e.g.:
> >
> >Documentation/device-mapper/thin-provisioning.txt
> >
> >http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/device-mapper/thin-provisioning.txt
> >
> >(the "Introduction" being the most approachable overview)
> OK, that's where I missed it. I looked at the snapshots description.
> It is now part of thin-provisioning. Got it. Will it be in the 2.6
> kernel or will there need to be an upgrade to 3.x series?

It went upstream in Linux 3.2.  So if using upstream kernels you'll need
Linux >= 3.2. (you'll also need the latest upstream lvm2 code for the
ability to create thinp devices and snapshots of them with lvm2).

But thinp will also be provided in the next RHEL6 release (6.3).

Mike