[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Automatic activation of pre-existing RAID/LVM devices (was: RAID10 and RAID6)

On Sat, Jul 19, 2008 at 12:16:07PM +0200, Frans Pop wrote:
> On Saturday 19 July 2008, Jérémy Bobbio wrote:
> > > Regarding your patch: my last thoughts on it were that I'm still not
> > > sure whether it is a good idea to switch to automatic detection of
> > > RAID during init.d at this point, especially if we don't do the same
> > > for LVM at the same time.
> >
> > As far as I know, we do the same for LVM.  Actually, my patch was
> > mostly about making the initialization for MD a lot closer to what was
> > done for LVM!
> No, we do not currently activate pre-existing LVM volumes during partman 
> initialization. We only do that after "Configure LVM" is selected from 
> the main partman dialog.
> See do_initial_setup() in partman-lvm/choose-partitions/lvm/do_option.

*sigh* I got mislead by the "pvscan" and "vgscan" at the top of the
init.d/lvm script.

> The only thing we do during Guided partitioning is _detect_ if there are 
> any pre-existing LVM devices, but that is done without activating them.
> With your RAID patchset pre-existing RAID devices would show up 
> automatically, while LVM devices would not.
> My current feeling is that that bit should be dropped from your current 
> patchset and that we should consider that question separately for both 
> RAID and LVM afterwards, and we should at least consider crypto as well.

I have tested my patch as is in many different situations and I feel
pretty confident in its current form…

> I think I would be in favor of it (even without crypto), but it would 
> require quite a bit of testing effort.

If you are in favor of it, I think it would make more sense (at least to
me) to spend time testing and working on automatic activation of LVM
devices than to spend many hours debugging a new patch with different

Automatic activation of RAID and LVM devices really make sense as,
otherwise, existing devices just magically appears when one will select
the relevant "configure" step.  The user experience described in #391474
feels very wrong to me and that is what motivated the current patch.

Jérémy Bobbio                        .''`. 
lunar@debian.org                    : :Ⓐ  :  # apt-get install anarchism
                                    `. `'` 

Attachment: signature.asc
Description: Digital signature

Reply to: