recommendations on how to recover a corrupted, LVM-based hard drive?

[Chris Murphy]

On Feb 17, 2014, at 8:05 AM, Matthew Miller <mattdm at fedoraproject.org> wrote:

> On Sun, Feb 16, 2014 at 11:52:16PM -0700, Chris Murphy wrote:
>>> Hallway talk, I'm afraid. However, we talked about actually having a
>>> session (or even a mini-track, if there's a lot of interest) at the next
>>> Flock about the near- and medium-term future of filesystems in Fedora.
>> Is Fedora.next to be biased toward production or testing?
> 
> I think the Fedora Desktop/Server/Cloud products we are talking about are
> more biased towards production than Fedora has sometimes been. I'm not sure
> that answers your question, though.

The question isn't literal. The point being, such a decision doesn't depend merely on the technical merits of each fs. There's a certain subjectivity to it, and there are other questions that need answering first, that makes it easier to make a compatible fs decision.


>> Should the user make that choice, and if so what changes are needed in the
>> installer for them to effectively make that choice? Are the decision
>> makers capable of saying "no" and actually establishing a handful of
>> installed layouts?
> 
> Actually, I'm not sure this is a question. It sounds like a rhetorical
> argument which happens to be in the form of a quetsion. 

There are three interdependent, non-literal, non-rhetorical questions designed to reveal an ideological statement. It's not a right or wrong whether users get to choose a file system via the easy path. It's an ideological question if the easy path is also a recommendation, and if it is then there shouldn't be a file system option - merely whether the user wants encryption or /home saved.


>> Right now the installer's auto/guided/easy path permits about 80 testable
>> outcomes. And yet not one of them permits Fedora 20 to be installed along
>> side Fedora 19. And not one permits a prior linux OS to be replaced while
>> keeping /home.
> 
> I agree that the last one seems like an important case which the installer
> could cover better.

If only that is added, and nothing is removed, we now have ~160 testable outcomes to the installer in the auto/guided/easy path. And I think 80 outcomes is nutty.

So back to the ideological question. If the bias is production, then I'd argue the bar needs to be raised with a rule that says anything offered in the installer must work. And to do that, necessarily a huge percent of installer features would have to be hidden by default (maybe revealed with an advanced boot time switch) because presently there's no way to test all outcomes now.


> Separately, if we're really interested in allowing/encouraging parallel
> installs, we probably should figure out how to make OSTree work officially.
> https://lwn.net/Articles/581811/

It'd be a better way, but are enough mirrors really going to sign up for hosting full blown expanded OSTree trees, meaning tens of thousands of files, rather than RPMs?

In the short term, it's already possible to do parallel Fedora 19 and Fedora 20 along side each other via custom partitioning without OSTree. The question is whether it's sufficiently common and useful for it to be a "recommended" outcome that enables easy/guided/auto partitioning to contend with; or should the user be compelled to use custom partitioning for such a case.



Chris Murphy