lv-shared

lv-shared

[Mikulas Patocka]

Hi

Why did you hide the "lv-shared" volume? With this hidden, the size and 
percentage can't be seen.

Also, your changes to 2.02.62 broke shared store resizing.
lvresize -L 132M vgs/lv-shared no longer works

Mikulas

lv-shared

[Mike Snitzer]

On Mon, Mar 15 2010 at  3:00pm -0400,
Mikulas Patocka <mpatocka@redhat.com> wrote:

> Hi
> 
> Why did you hide the "lv-shared" volume? With this hidden, the size and 
> percentage can't be seen.

You can still see it with: lvs -a

I'll have to revisit why I made it hidden (aside from the fact that the
associated DM device is also hidden).  There was some subtle reason that
had to do with the refactoring of associations when I eliminated the
distinct ->shared_snapshot pointer (otherwise the vg_validate() method
would fail).
 
> Also, your changes to 2.02.62 broke shared store resizing.
> lvresize -L 132M vgs/lv-shared no longer works

OK, I can have a look but some of my changes also made it so that
info-by-uuid based multisnapshot cleanup actually works (needed for lvm2
>= 2.02.62).

Mike

lv-shared

[Mikulas Patocka]

On Mon, 15 Mar 2010, Mike Snitzer wrote:

> On Mon, Mar 15 2010 at  3:00pm -0400,
> Mikulas Patocka <mpatocka@redhat.com> wrote:
> 
> > Hi
> > 
> > Why did you hide the "lv-shared" volume? With this hidden, the size and 
> > percentage can't be seen.
> 
> You can still see it with: lvs -a
> 
> I'll have to revisit why I made it hidden (aside from the fact that the
> associated DM device is also hidden).  There was some subtle reason that
> had to do with the refactoring of associations when I eliminated the
> distinct ->shared_snapshot pointer (otherwise the vg_validate() method
> would fail).

I know it can be seen with lvs -a, but I think it's inconvenient for the 
admin:
- the admin won't see that there is a shared store taking up space
- the admin won't see that the origin is a shared snapshot, he sees 'o', 
but nothing that refers to that 'o'
- the percentage usage of the shared snapshot is hidden

I think this information is important and should be displayed with plain 
"lvs" command without any flags. (for example, the percentage usage is 
critical, because it can overflow) --- so "lv-shared" shouldn't be hidden.

Mikulas

lv-shared

[Mike Snitzer]

On Mon, Mar 15 2010 at  4:06pm -0400,
Mikulas Patocka <mpatocka@redhat.com> wrote:

> On Mon, 15 Mar 2010, Mike Snitzer wrote:
> 
> > On Mon, Mar 15 2010 at  3:00pm -0400,
> > Mikulas Patocka <mpatocka@redhat.com> wrote:
> > 
> > > Hi
> > > 
> > > Why did you hide the "lv-shared" volume? With this hidden, the size and 
> > > percentage can't be seen.
> > 
> > You can still see it with: lvs -a
> > 
> > I'll have to revisit why I made it hidden (aside from the fact that the
> > associated DM device is also hidden).  There was some subtle reason that
> > had to do with the refactoring of associations when I eliminated the
> > distinct ->shared_snapshot pointer (otherwise the vg_validate() method
> > would fail).
> 
> I know it can be seen with lvs -a, but I think it's inconvenient for the 
> admin:
> - the admin won't see that there is a shared store taking up space
> - the admin won't see that the origin is a shared snapshot, he sees 'o', 
> but nothing that refers to that 'o'
> - the percentage usage of the shared snapshot is hidden
> 
> I think this information is important and should be displayed with plain 
> "lvs" command without any flags. (for example, the percentage usage is 
> critical, because it can overflow) --- so "lv-shared" shouldn't be hidden.

Sure, not a problem.

I've updated the relevant patch, see:
http://people.redhat.com/msnitzer/patches/multisnap/lvm2/LVM2-2.02.62/lvm-shared-eliminate-shared_snapshot-in-lv.patch

The lib/metadata/metadata.c:snapshot_count() changes in this updated
patch address the vg_validate() issue I was seeing.  The -shared cow was
considered to be a normal "snapshot"; as such the vg_validate() checks
failed.

Mike

lv-shared

[Mike Snitzer]

On Mon, Mar 15 2010 at  6:01pm -0400,
Mike Snitzer <snitzer@redhat.com> wrote:

> On Mon, Mar 15 2010 at  4:06pm -0400,
> Mikulas Patocka <mpatocka@redhat.com> wrote:
> 
> > On Mon, 15 Mar 2010, Mike Snitzer wrote:
> > 
> > > On Mon, Mar 15 2010 at  3:00pm -0400,
> > > Mikulas Patocka <mpatocka@redhat.com> wrote:
> > > 
> > > > Hi
> > > > 
> > > > Why did you hide the "lv-shared" volume? With this hidden, the size and 
> > > > percentage can't be seen.
> > > 
> > > You can still see it with: lvs -a
> > > 
> > > I'll have to revisit why I made it hidden (aside from the fact that the
> > > associated DM device is also hidden).  There was some subtle reason that
> > > had to do with the refactoring of associations when I eliminated the
> > > distinct ->shared_snapshot pointer (otherwise the vg_validate() method
> > > would fail).
> > 
> > I know it can be seen with lvs -a, but I think it's inconvenient for the 
> > admin:
> > - the admin won't see that there is a shared store taking up space
> > - the admin won't see that the origin is a shared snapshot, he sees 'o', 
> > but nothing that refers to that 'o'
> > - the percentage usage of the shared snapshot is hidden
> > 
> > I think this information is important and should be displayed with plain 
> > "lvs" command without any flags. (for example, the percentage usage is 
> > critical, because it can overflow) --- so "lv-shared" shouldn't be hidden.
> 
> Sure, not a problem.
> 
> I've updated the relevant patch, see:
> http://people.redhat.com/msnitzer/patches/multisnap/lvm2/LVM2-2.02.62/lvm-shared-eliminate-shared_snapshot-in-lv.patch
> 
> The lib/metadata/metadata.c:snapshot_count() changes in this updated
> patch address the vg_validate() issue I was seeing.  The -shared cow was
> considered to be a normal "snapshot"; as such the vg_validate() checks
> failed.

Mikulas,

I've cleaned up my patches that layer on your a bit, and updated to
2.02.63:
http://people.redhat.com/msnitzer/patches/multisnap/lvm2/LVM2-2.02.63/

This patch saw some changes becuase I moved some functions around:
http://people.redhat.com/msnitzer/patches/multisnap/lvm2/LVM2-2.02.63/lvm-shared-additional-wrappers.patch

One fix associated with the changes is detailed at the end of the
following patch's header:
http://people.redhat.com/msnitzer/patches/multisnap/lvm2/LVM2-2.02.63/lvm-shared-eliminate-shared_snapshot-in-lv.patch

And I implemented the ACCESS_HIDDEN_LV flag we talked about with this
patch:
http://people.redhat.com/msnitzer/patches/multisnap/lvm2/LVM2-2.02.63/lvm-shared-add-ACCESS_HIDDEN_LV-flag.patch

With the result being:
# lvs
  LV               VG   Attr   LSize Origin  Snap%  Move Log Copy%  Convert
  testlv1          test owi-a- 4.00g                                       
  [testlv1-shared] test swi--- 1.00g testlv1   0.01                        
  testlv1_snap     test swi-a- 4.00g testlv1                

Mike

lv-shared

[Mike Snitzer]

On Mon, Mar 15 2010 at  3:29pm -0400,
Mike Snitzer <snitzer@redhat.com> wrote:

> On Mon, Mar 15 2010 at  3:00pm -0400,
> Mikulas Patocka <mpatocka@redhat.com> wrote:
> 
> > Also, your changes to 2.02.62 broke shared store resizing.
> > lvresize -L 132M vgs/lv-shared no longer works
> 
> OK, I can have a look but some of my changes also made it so that
> info-by-uuid based multisnapshot cleanup actually works (needed for lvm2
> >= 2.02.62).

I cannot reproduce this:

# lvcreate -L 1G --sharedstore mikulas -s test/testlv1
  Logical volume "testlv1-shared" created

# lvs -a
  LV               VG   Attr   LSize Origin  Snap%  Move Log Copy%  Convert
  testlv1          test owi-a- 4.00g                                       
  [testlv1-shared] test swi--- 1.00g testlv1   0.00                        

# lvcreate -s -n testlv1_snap test/testlv1
  Logical volume "testlv1_snap" created

# lvresize -L 1.5G test/testlv1-shared
  Extending logical volume testlv1-shared to 1.50 GiB
  Logical volume testlv1-shared successfully resized

# lvs -a
  LV               VG   Attr   LSize Origin  Snap%  Move Log Copy%  Convert
  testlv1          test owi-a- 4.00g                                       
  [testlv1-shared] test swi--- 1.50g testlv1   0.00                        
  testlv1_snap     test swi-a- 4.00g testlv1                               

lv-shared

[Mikulas Patocka]

On Mon, 15 Mar 2010, Mike Snitzer wrote:

> On Mon, Mar 15 2010 at  3:29pm -0400,
> Mike Snitzer <snitzer@redhat.com> wrote:
> 
> > On Mon, Mar 15 2010 at  3:00pm -0400,
> > Mikulas Patocka <mpatocka@redhat.com> wrote:
> > 
> > > Also, your changes to 2.02.62 broke shared store resizing.
> > > lvresize -L 132M vgs/lv-shared no longer works
> > 
> > OK, I can have a look but some of my changes also made it so that
> > info-by-uuid based multisnapshot cleanup actually works (needed for lvm2
> > >= 2.02.62).
> 
> I cannot reproduce this:
> 
> # lvcreate -L 1G --sharedstore mikulas -s test/testlv1
>   Logical volume "testlv1-shared" created
> 
> # lvs -a
>   LV               VG   Attr   LSize Origin  Snap%  Move Log Copy%  Convert
>   testlv1          test owi-a- 4.00g                                       
>   [testlv1-shared] test swi--- 1.00g testlv1   0.00                        
> 
> # lvcreate -s -n testlv1_snap test/testlv1
>   Logical volume "testlv1_snap" created
> 
> # lvresize -L 1.5G test/testlv1-shared
>   Extending logical volume testlv1-shared to 1.50 GiB
>   Logical volume testlv1-shared successfully resized
> 
> # lvs -a
>   LV               VG   Attr   LSize Origin  Snap%  Move Log Copy%  Convert
>   testlv1          test owi-a- 4.00g                                       
>   [testlv1-shared] test swi--- 1.50g testlv1   0.00                        
>   testlv1_snap     test swi-a- 4.00g testlv1                               

Now I can't reproduce it too with the same code, so I don't know what 
happened :-(

Mikulas