Re: Moving partition to an overlapping position

[K.G.]

On Wed, 20 Jul 2005 11:22:21 +0200 (MEST)
Szakacsits Szabolcs <address@hidden> wrote:

> 
> On Wed, 20 Jul 2005, K.G. wrote:
> 
> > All Parted operations are theoretically power-loss proof, including resizing
> > supported FS. 
> 
> I'm interested how. I know it's possible but it's not always very trivial.
> 
> In the general case ensuring consistency is difficult. Disks and
> controllers can lie that data is on the disk platter but in real they 
> keep it in their cache which can get lost during power outage.
> 
> They also reorder data for performance, so it's possible that what you
> think is already written out (sync also doesn't give guarantee), in real
> isn't but things what you didn't sync latter on are already on the
> platter.

I've run a couple of tests, violently unplugging the computer while resizing 
a HFS, and at least it worked fine for me :)

Of course I can't warranty it'll be ok every time, but I think there's a good
probability of non destruction anyway.

In contrast implementing things like moving a partition in a generic way by
spliting it and moving each part would produce something that has a high
probability of destruction in case of interruption. It could be ok using
a kind of journal anyway.

> > I think introducing non power-loss proof operations could be
> > disturbing. It would be better to write another program to perform
> > such operations.
> 
> Power outage is not a common case and everybody should have always backups
> because fatal disk failures are much-much more common.

This is another issue. Anyway people using Parted to resize/move non
destructively tend to have no backup but of their most important files...

Guillaume