[linux-lvm] Dangerous Recipes

[linux-lvm] Dangerous Recipes

[Joseph Kezar]

I have a single shared SCSI JBOD.  And it makes up my VolumeGroupMailData.

It is mounted read/write by only 1 node at any given time.  This is strictly
enforced.

Although, at boot both nodes vgchange -a y the Volume Group.
So it is picked up in vgscan as an active VG on node1 and node2
I am aware if I need to make a change, like adding a Logical Volume,
Extending/Reducing a Logical Volume, or adding/subtracting a physical volume
that the volume must be deactivated on the secondary node(in my case node
2).

If everything goes well enforced, am I still in trouble or at risk of data
corruption?

Re: [linux-lvm] Dangerous Recipes

[Patrick Caulfield]

On Mon, Jan 27, 2003 at 03:42:13PM -0500, Joseph Kezar wrote:
> I have a single shared SCSI JBOD.  And it makes up my VolumeGroupMailData.
> 
> It is mounted read/write by only 1 node at any given time.  This is strictly
> enforced.
> 
> Although, at boot both nodes vgchange -a y the Volume Group.
> So it is picked up in vgscan as an active VG on node1 and node2
> I am aware if I need to make a change, like adding a Logical Volume,
> Extending/Reducing a Logical Volume, or adding/subtracting a physical volume
> that the volume must be deactivated on the secondary node(in my case node
> 2).
> 
> If everything goes well enforced, am I still in trouble or at risk of data
> corruption?

If you are careful it should be fine. 

What you must always remember to do is "vgchange -an" on the secondary node
before doing the metadata changes, then a "vgscan;vgchange -ay" on that node
after the metadata changes.

-- 

patrick