[mythtv-users] IOBOUND errors after kernel upgrade from 2.6.31 to 2.6.32

[Brian Wood]

On Friday 07 May 2010 11:18:38 am Tom Dexter wrote:
> On Fri, May 7, 2010 at 12:48 PM, Brian Wood <beww at beww.org> wrote:
> > On Friday 07 May 2010 10:36:24 am John Drescher wrote:
> >> > From what I've read, the kernel folks seem determined to push ext4 out
> >> > the door, ready or not, I'm not sure why.
> >>
> >> It's been working well for me for over 2 years. That is since kernel
> >> 2.6.26 and ext4-dev. And I not talking about just Mythtv usage. I use
> >> it at work with at least 10TB of medical image storage. I am not using
> >> barriers and I also am not updating kernels frequently. So possibly
> >> that is why it has gone well for me.
> >
> > I just don't like to see ext4 used as a test bed.
> >
> > I miss the old "odd/even" model for kernel development, you could use an
> > even kernel if you wanted stability, and an odd one if you wanted to go
> > "bleeding edge" and accept the risks. Things could be tested on the
> > development tree before moving them to the stable (even) tree once proven
> > reliable.
> >
> > Linus eliminated the odd/even system, saying that by doing do he was
> > "making kernel development smoother".
> >
> > "Smoother" perhaps, for the developers, not necessarily for the users,
> > who are now treated as guinea pigs.
> 
> I'm a bit put off by the radical changes that appear to have gone into
> ext4 between those two versions of 2.6.31...the changes that appear to
> causing my performance issues.  My understanding is most of these
> recent changes were about correcting potential issues with data
> integrity, and they tended to be at the cost of performance.  It seems
> odd to push out such a radical change in the middle of a kernel
> version like that.  I guess they were concerned enough about the
> integrity issues to do that.

"Hiding" major changes by making only a minor version number change is typical 
of Microsoft, not Linux:

http://tech.slashdot.org/story/10/05/06/1734250/Security-Firm-Reveals-
Microsofts-Silent-Patches

> 
> I'm just sticking with the working kernel for the foreseeable future.
> Another option would be to either try going back to ext3 or perhaps
> trying XFS, but either one means temporarily copying a TB of
> recordings off to some drive I currently don't even have...not sure I
> want to deal with that any time soon.

1 TB drives are getting cheaper, assuming you can get one without firmware 
issues. I'm staying away from Seagate or WD now, but I may run out of 
manufacturers if the current trend keeps up.