Re: md raid6 deadlock on write

[NeilBrown <neilb@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 1215 bytes --]

On Fri, 6 Jul 2012 14:55:24 +0100 Jose Manuel dos Santos Calhariz
<jose.calhariz@netvisao.pt> wrote:

> On Wed, Jul 04, 2012 at 11:24:11AM +0100, Jose Manuel dos Santos Calhariz wrote:
> > On Wed, Jul 04, 2012 at 11:38:11AM +1000, NeilBrown wrote:
> > > On Mon, 2 Jul 2012 23:15:08 +0100 Jose Manuel dos Santos Calhariz
> > > <jose.calhariz@netvisao.pt> wrote:
> > > 
> > > > 
> > > > We have a group of servers with a LVM over a RAID6 of 16 drives.
> > > > During normal work loads, sometimes, the md raid enter on deadlock for
> > > > writes and only a power off/power on allows to recover the machine.
> > > 
> > > This might be fixed by the following commit which was recently included in
> > > 3.5-rc. If could test with that I'd appreciate it.
> > 
> > We will do it, at first opportunity.
> 
> We have two machines that are running fio for 24 hours without
> problems.  So the bug seams to be fixed, thank you.

Thanks for testing and reported.


> 
> Any possibility of the fix being ported to kernel 3.2?

It seems that I didn't tag that patch for -stable so it won't automatically
get included.  I'll send it the old way - maybe it'll get into 3.2.23.

Thanks,
NeilBrown


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]