xfs
[Top] [All Lists]

Bug#713994: XFS repair

To: Wes Deviers <yhetti@xxxxxxxxx>, 713994@xxxxxxxxxxxxxxx
Subject: Bug#713994: XFS repair
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Tue, 25 Jun 2013 08:47:53 +1000
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <CAMjbrKp164a2CU2SPn94NGP+urwKUCtGz2vKyt6snSWaYdHebA@xxxxxxxxxxxxxx>
References: <CAMjbrKp164a2CU2SPn94NGP+urwKUCtGz2vKyt6snSWaYdHebA@xxxxxxxxxxxxxx>
Reply-to: Dave Chinner <david@xxxxxxxxxxxxx>, 713994@xxxxxxxxxxxxxxx
Resent-cc: XFS Development Team <xfs@xxxxxxxxxxx>
Resent-date: Mon, 24 Jun 2013 22:51:13 +0000
Resent-from: Dave Chinner <david@xxxxxxxxxxxxx>
Resent-message-id: <handler.713994.B713994.137211409414191@xxxxxxxxxxxxxxx>
Resent-sender: Debian BTS <debbugs@xxxxxxxxxxxxxxxxxxxx>
Resent-to: debian-bugs-dist@xxxxxxxxxxxxxxxx
User-agent: Mutt/1.5.21 (2010-09-15)
On Mon, Jun 24, 2013 at 12:43:55PM -0400, Wes Deviers wrote:
> Update:
> 
> xfs_repair has fixed the problem:
> 
> root@ashdbtrk-s008-02:/home/wdeviers# xfs_check /dev/vg0/mysql
> agf_freeblks 327676, counted 655100 in ag 4
> agf_longest 327676, counted 655100 in ag 4
> block 12/1 expected type unknown got btbno
> block 12/4 expected type unknown got free2
> ...
> block 12/45268 expected type unknown got free2
> block 12/45269 expected type unknown got free2
> block 12/45270 expected type unknown got free2
> block 12/45271 expected type unknown got free2
> block 12/45272 expected type unknown got free2
> block 12/45273 expected type unknown got free2
> block 12/45274 expected type unknown got free2
> block 12/45275 expected type unknown got free2
> 
> root@ashdbtrk-s008-02:/home/wdeviers# xfs_repair /dev/vg0/mysql
> Phase 1 - find and verify superblock...
> Phase 2 - using internal log
>         - zero log...
>         - scan filesystem freespace and inode maps...
> primary/secondary superblock 4 conflict - AG superblock geometry info
> conflicts with filesystem geometry
> bad sequence # 12 for agf 4
> bad length 327680 for agf 4, should be 655104
> bad sequence # 12 for agi 4
> bad length # 327680 for agi 4, should be 655104
> reset bad sb for ag 4
> reset bad agf for ag 4
> reset bad agi for ag 4
> sb_fdblocks 28836636, counted 28509212

The AG headers for AG #4 have been overwritten with header from a
different AG# (12 instead of 4) that has a different length (327680
instead of 655104).  The sequence number mismatch is why the runtime
errors are being thrown

However, the superblock for AG 4 has the wrong geometry and that's
indicative of it belonging to some other filesystem. i.e. a
misdirected write. I'd be taking a careful look at your storage,
because misdirected writes are evil and *extremely hard* to diagnose
without specialised equipment...

Anyway, you're running an old, old kernel (2.6.32-5-amd64) and
xfsprogs, so unless you can reproduce this reliably on a couple of
different types of hardware with a more recent kernel and xfsprogs,
we aren't really going to be able to do anything about your problem.

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>