Problems with kernel 3.6.x (vm ?) (was : Is kernel 3.6.1 or filestreams option toxic ?)

[Dave Chinner]

On Sat, Oct 27, 2012 at 12:05:34AM +0200, Yann Dupont wrote:
> Le 26/10/2012 12:03, Yann Dupont a écrit :
> >Le 25/10/2012 23:10, Dave Chinner a écrit :
> >
> >I'll try now to reproduce this kind of behaviour on a verry little
> >volume (10 GB for exemple) so I can confirm or inform the given
> >scenario .
> >
> 
> This is reproductible. Here is how to do it :
> 
> - Started a 3.6.2 kernel.
> 
> - I created a fresh lvm volume on localdisk of 20 GB.

Can you reproduce the problem without LVM?

> - mkfs.xfs on it, with default options
> - mounted with default options
> - launch something that hammers this volume. I launched compilebench
> 0.6  on it
> - wait some time to fill memory,buffers, and be sure your disks are
> really busy. I waited some minutes after the initial 30 kernel
> unpacking in compilebench
> - hard reset the server (I'm using the Idrac of the server to
> generate a power cycle)
> - After some try, I finally had the impossibility to mount the xfs
> volume, with the error reported in previous mails. So far this is
> normal .

So it doesn't happen every time, and it may be power cycle related.
What is your "local disk"?

> 
> xfs_logprint don't say much :
> 
> xfs_logprint:
>     data device: 0xfe02
>     log device: 0xfe02 daddr: 10485792 length: 20480
> 
> Header 0x7c wanted 0xfeedbabe
> **********************************************************************
> * ERROR: header cycle=124         block=5414 *
> **********************************************************************

You didn't look past the initial error, did you? The file is only
482280 lines long, and 482200 lines of that are decoded log data....
:)

> I tried xfs_logprint -c , it gaves a 22M file. You can grab it here :
> http://filex.univ-nantes.fr/get?k=QnBXivz2J3LmzJ18uBV

I really need the raw log data, not the parsed output. The logprint
command to do that is "-C <file>", not "-c".

> - Rebooted 3.4.15
> - xfs_logprint gives the exact same result that with 3.6.2 (diff
> tells no differences)

Given that it's generated by the logprint application, I'd expect it
to be identical.

> but on 3.4.15, I can mount the volume without problem, log is
> replayed.

> for information here is xfs_info of the volume :
> 
> here is xfs_info output
> 
> root at label5:/mnt/debug# xfs_info /mnt/tempo
> meta-data=/dev/mapper/LocalDisk-crashdisk isize=256    agcount=8,
> agsize=655360 blks

How did you get a default of 8 AGs? That seems wrong.  What version
of mkfs.xfs are you using?

Cheers,

Dave.
-- 
Dave Chinner
david at fromorbit.com