Re: [Aoetools-discuss] one array keeps disconnecting

[Scott R. <ska...@gm...>]

OK, with that in mind, we've decided to upgrade our controller system
(the one that mounts the four aoe arrays).  The kernel is
2.6.15-1-686-smp on Debian 3.1.  What I need to know is:

- What kernel should I upgrade to?
- What version of aoetools/vblade should I be using?
- What version of xfsprogs should I be using?

We're going to arrange for a maintenance window to bring everything
down, upgrade, run repairs on all of the arrays, and bring them back
online.

One final concern.  The arrays are all 5TB in size.  I understand that
xfs_repair had an issue with devices larger than 2TB.  Does the later
version correct this?

On 6/18/07, Ed L. Cashin <ec...@co...> wrote:
> On Fri, Jun 15, 2007 at 11:44:28PM -0700, Scott Raymond wrote:
> > Actually, I did paste the output of df from my ssh session in my
> > follow-up to Smart.
> >
> > I suppose I could have been more clear - I'm not a *nix n00b, just an
> > aoetools/vblade n00b.  :)
>
> Ah.  Sorry for the confusion.  More comments follow inline below.
>
> > We didn't see any unusual output in dmesg.
> > However, I did find something in /var/log/messages that might hold a
> > clue.  Sorry for the large paste:
> >
> ...
> > Jun  8 03:09:05 netstoragecontrol2 kernel:  [pg0+950020086/1070175232]
> > xfs_free_ag_extent+0xc4/0x604 [xfs]
> ...
> > Jun  8 03:09:05 netstoragecontrol2 kernel:  [sys_ftruncate64+220/250]
> > sys_ftruncate64+0xdc/0xfa
> > Jun  8 03:09:05 netstoragecontrol2 kernel:  [syscall_call+7/11]
> > syscall_call+0x7 /0xb Jun  8 03:09:05 netstoragecontrol2 kernel:
> > xfs_force_shutdown(etherd/e4.0,0x8) c alled from line 4125 of file
> > fs/xfs/xfs_bmap.c.  Return address = 0xf8d78068 Jun  8 03:22:17
> > netstoragecontrol2 kernel: xfs_force_shutdown(etherd/e4.0,0x1) c alled
> > from line 339 of file fs/xfs/xfs_rw.c.  Return address = 0xf8db4c71
> >
> > All I can tell from that is that xfs didn't like something on the
> > array and shut down the connection.  But I'll be damned if I know what
> > it was xfs didn't like.
>
> Again, there's no connection.  XFS shut down the filesystem.  Usually
> when that happens it is because there's a problem with the
> filesystem.  Some kernels have bugs in the xfs driver, and so a course
> of action might be to ...
>
>   * look in the changelogs for your kernel to see whether there is a
>     new version that has XFS bugfixes, upgrading if so,
>
>   * look for a new version of the xfstools, upgrading if a better one
>     is available,
>
>   * use xfs_check and/or xfs_repair to make sure there's nothing wrong
>     with the filesystem, or (even better) ...
>
>   * really make sure there's nothing wrong with the filesystem by
>     rebuilding it with "mkfs -t xfs ...".
>
> Good luck with your endeavors.  XFS can be a good filesystem but if
> you catch a bad release, it can result in some frustrating stumbling
> blocks like this.
>
> --
>   Ed L Cashin <ec...@co...>
>


-- 
Scott