[linux-lvm] Removing faulty disk from LVM2

Magnus Nilsson magnus at upcore.net
Fri Feb 4 14:34:52 UTC 2005


On Fri, 2005-02-04 at 14:06 +0000, Alasdair G Kergon wrote:
> On Fri, Feb 04, 2005 at 10:49:15AM +0100, Magnus Nilsson wrote:
> > Through pvmove'ing parts of the PV I've narrowed it down to PE #14858 of
> > 38154 4MB PE's. 
> 
> > Is there any way of just skipping that last PE and let me repair the
> > filesystem afterwards? Perhaps by vgcfgrestore'ing an edited backup,
> > excluding the broken PV?
> 
> 
> Firstly, pvmove --abort abandons a pvmove cleanly.
> 
> Two ideas:
>   Use dmsetup directly to hide the problematic PE.
> e.g. dmsetup table vg1-lvol1 to display the table,
>   Work out which sector(s) are the problem and replace them with
>   an 'error' line
> 
> 
>   Edit a (current) VG backup file to 'move' 1 extent elsewhere.
> e.g. 1 segment containing the problematic extent turns into 3 segments
> 
> Alasdair

Thanks for the tips.

I did try 'pvmove --abort' before mailing, but since the VG couldn't be
found the pvmove couldn't be aborted. I tried using -P with pvmove, but
the parameter wasn't recognised.

I managed to fix the problem myself though. I thought that maybe the
machine would behave differently if the disk was on a different
controller, so I moved the faulty disk from the Promise TX2 controller
to the motherboard. The machine still complained about the disk, but at
least it didn't hard-lock. That allowed the final pvmove to go through,
so I could vgreduce the PV. An fsck.ext3 discovered a couple of FS
hickups on one of the 3 LV's, but nothing major.

I hope that I will never have to use your ideas, ever. :)

//Magnus




More information about the linux-lvm mailing list