[linux-lvm] pvmove abort

[linux-lvm] pvmove abort

[Sebastian Walter]

Dear list,

I have a problem recovering from an aborted pvmove process. I have a
volume group called storage consisting of 2 partitions (PV's) with 2 TB
each (let's call them A and B). I added a third partition (let's call it
C) using vgextend. Now I did a pvmove A to move A's content to C but
recognized that it would take a long time so I canceled the pvmove
pressing ctrl-c.

After that I saw that some data was already written on the formerly
empty PV C. I tried to pvmove this new pv out of the volume group, but
pvmove refuses to work with the following error:

Skipping locked LV storage
Skipping mirror LV pvmove0
No data to move for storage

There is an additional LV on the VG now called pvmove0. I dont
understand what pvmove does, but this LV seems to be a mirror. Does this
mean that everything is still on the old PV and I can delete the LV
pvmove0? The VG in total seems to be ok.

How can I unlock the volume and free the pv? Any help is greatly
appreciated.

Best regards,
Sebastian

Re: [linux-lvm] pvmove abort

[Alasdair G Kergon]

pvmove --abort is the correct way to stop a pvmove.
Then see how much got moved and decide if you want to pvmove it back.

The man page explains how it works, but the data moving happens in
the kernel - in the background if you like - and after the move begins,
the 'pvmove' process just monitors progress and tidies up when the
kernel has finished what it's been asked to do.

Alasdair
-- 
agk@redhat.com

Re: [linux-lvm] pvmove abort

[Sebastian Walter]

"pvmove --abort" solved the problem. It cleaned the mirrored LV and
everything looks sane now. Thank you very much!


Alasdair G Kergon wrote:
> pvmove --abort is the correct way to stop a pvmove.
> Then see how much got moved and decide if you want to pvmove it back.
>
> The man page explains how it works, but the data moving happens in
> the kernel - in the background if you like - and after the move begins,
> the 'pvmove' process just monitors progress and tidies up when the
> kernel has finished what it's been asked to do.
>
> Alasdair
>