Re: easily reproducible filesystem crash on rebuilding array

From: Dave Chinner <david@fromorbit.com>
To: Emmanuel Florac <eflorac@intellique.com>
Cc: xfs@oss.sgi.com
Subject: Re: easily reproducible filesystem crash on rebuilding array
Date: Sat, 20 Dec 2014 10:06:05 +1100	[thread overview]
Message-ID: <20141219230605.GF24183@dastard> (raw)
In-Reply-To: <20141219125720.56c5f74d@galadriel.home>

On Fri, Dec 19, 2014 at 12:57:20PM +0100, Emmanuel Florac wrote:
> Le Fri, 19 Dec 2014 09:58:43 +1100 vous écriviez:
> 
> > 
> > Where's the XFS error output? This is just the output from the
> > dump_stack() call in the xfs error message code...
> 
> Where is it supposed to display its errors? I thought it to be
> in /var/log/messages....

Depends on how you system is configured. have you turned down the
dmesg error level?

> > Still, that's implying a write IO error being reporte din IO
> > completion, not a read error, and that's different to the previous
> > issue you've reported. It's also indicative of an error coming from
> > the storage, not XFS...
> >
> > Do these problems *only* happen during or after a RAID rebuild?
> 
> Only while the rebuild process is running. All works fine afterwards.

Which pretty much points to a RAID controller rebuild bug.

> > > Phase 7 - verify and correct link counts...
> > > resetting inode 4294866029 nlinks from 2 to 5
> > > resetting inode 150323855504 nlinks from 13 to 12
> > > Metadata corruption detected at block 0x10809dc640/0x1000
> > > libxfs_writebufr: write verifer failed on bno 0x10809dc640/0x1000
> > > Metadata corruption detected at block 0x10809dc640/0x1000
> > > libxfs_writebufr: write verifer failed on bno 0x10809dc640/0x1000
> > > done
> > 
> > I'd suggest you should be upgrading xfsprogs, because that's an
> > error that shouldn't happen at the end of a repair. If the latest
> > version (3.2.2) doesn't fix this problem, then please send me a
> > compressed metadump so I can work out what corruption xfs_repair
> > isn't fixing properly.
> 
> It was 3.2.2 this time. However it seems to have fixe it anyway;
> running it a second time displays nothing special.
> Is a metadump of a clean filesystem of any use?

Not really.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs