easily reproducible filesystem crash on rebuilding array

[Ferenc Kovacs]

2015. szept. 17. 8:17 ezt írta ("Emmanuel Florac" <eflorac at intellique.com>):
>
> Le Wed, 16 Sep 2015 15:50:13 +0200 vous écriviez:
>
> > Hi,
> >
> > have you found a resolution to your problem?
> > we are facing a similar issue(fs corruption every time when
> > rebuilding the array) with an Adaptec Series 8 Raid controller.
>
> The problem didn't occur anymore after I deactivated
> individual disks write caching in the "controller settings". This is
> only adjustable through the RAID BIOS though, not through arcconf
> apparently.
> This seems to apply to all of 5xx5, 6xx5, 7xx5 series. I never rebuilt
> an array with a 8xx5 yet (though I have some). I don't know why this is
> not the default settings (default keeps the disks write cache active
> instead). Adaptec support and engineering deny that the problem exists,
> though I've sent many logs and defined a test procedure that allow to
> reproduce the problem reliably enough.
> Thanks god for XFS. ext4 is completely unaware of the corruptions, but
> they're present nonetheless (md5summing files before, during and
> after rebuild)...
>

Thanks for the info!

> BTW I'm curious, where did you find this address? This isn't the one
> I'm using on the XFS mailing list...

I was using the gmane archives to read your thread and the there the domain
part is truncated for privacy(even from your signature) so I googled your
mail handle with your full name and this address was the first result.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://oss.sgi.com/pipermail/xfs/attachments/20150917/9670afe0/attachment.html>