Importance of LVM (was Re: Partitioning criteria revision proposal)

[Adam Williamson]

On Thu, 2012-10-25 at 19:23 -0600, Chris Murphy wrote:
> On Oct 25, 2012, at 4:32 PM, Kamil Paral <kparal at redhat.com> wrote:
> > The default was flipped to raw partitions in early builds of F18
> Anaconda, and that was unfortunate, because it lacked a proper
> discussion and announcement (it was quite a surprise for QA). It is
> still (barely) time to flip the default back, to what it always was.
> But there's no way enough time to _start_ the discussion now. It would
> consume weeks and by that time Beta should be out.
> 
> I brought this up just over two months ago on both the anaconda and
> test lists and there was not all that much discussion then. So I don't
> see why it's such a big deal now.
> 
> https://www.redhat.com/archives/anaconda-devel-list/2012-August/msg00116.html
> 
> From an autopartition point of view, putting in code that's just going
> to be removed again come btrfs time doesn't make sense me. The
> autopartitioning for btrfs obviates lvm and md raid (except where md
> is needed to support a prior full disk IMSM RAID setup).
> 
> I'm not convinced it makes sense to wedge LVM for autopartitioning
> when it's not needed in the next release. One release without it is
> not such a big deal, really.

Nothing gets 'wedged in' anywhere, there is no code to 'put in' (nor
will any of the code that exists get 'removed' even when we default to
btrfs, I don't think).

I already posted the patch: it's two lines. All the code for LVM
autopart already exists and is the same code that has existed for years.
The patch simply changes the way the autopart code is called from
'please don't use LVM' to 'please use LVM'. It is two lines.
http://fpaste.org/w1vE/

> I like LVM. But I don't care about LVM as default for autopart one way
> or another. We're just post beta freeze, and this is coming up for
> serious conversation now? I think it needs to be let go. I think Jesse
> Keating's reply is a sufficiently good and timely explanation for
> having set expectations well prior to now.

The tricky thing is that the argument kind of cuts both ways: the thing
that's the big change here was changing from LVM-by-default to
raw-ext4-by-default, and that should have been clearly publicised and
discussed. The fact that there are people just now finding out that it
happened and being unhappy about it rather indicates that the planned
change _wasn't_ properly communicated.

I don't really see any great options here, unfortunately.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net