Bug#348449: kernel-image-2.6.8-2-k7: Oops during lv moving on resyncing raid5 md pv
On Tue, Jan 17, 2006 at 02:31:47AM +0100, Jeroen van Wolffelaar wrote:
> I did a lvremove of a snapshot of a logical volume that partly was on
> md9. This lvremove command hangs, and I got the below entry in my
> syslog. Resync is continuing but very slowly, while the lvremove is
> still also active, and seems uninterrupteable. A parallel lvdisplay
> hangs before displaying anything.
In the end, system didn't recover after 6 ours, and power cycle was
required.
Also pvmove's crashed at some point, similarly requiring a power cycle. On
reboot, lvm didn't even start anymore, and the kernel oopsed just while
trying to start LVM. Something very fishy, I don't have the last oops
because /var/log is on LVM.
With a grml 0.5 boot cd (kernel 2.6.13.4) though, everything worked and
could be recovered, pvmove resumed, etc. So I assume this bug has meanwhile
been fixed in later kernel versions. There was some minor data loss though,
which in itself would warrant a point-release-upload if the bug were to be
found. But eh, I don't think that's a very achieveable thing to do, unless
someone happens to have a good idea on what the kernel bug might be. In any
case, *I* am not capable of finding it. So if you don't think it's doable
to find & fix this bug, please close it, I'm pretty sure etch will be safe
for this.
--Jeroen
--
Jeroen van Wolffelaar
jeroen@wolffelaar.nl
http://jeroen.A-Eskwadraat.nl
Reply to: