Re: barrier and commit options?

[Nicolas KOWALSKI <nicolas kowalski gmail com>]

On Fri, Jan 30, 2009 at 10:22:46AM -0500, Eric Sandeen wrote:
> Christian Kujau wrote:
> > On Fri, 30 Jan 2009, Nicolas KOWALSKI wrote:
> >> I know I may loose the last 30 seconds of "work" (it's just a home
> >> server), but is the filesystem at risk (corruption, whatever, ...) with
> >> these mount options ?
> > 
> > No, why would it? If certain mount options would make a filesystem prone 
> > to corruption I'd consider this a bug.
> 
> Well, that's not exactly true.  Turning off barriers, depending on your
> storage, could lead to corruption in some cases.  Mounting with
> data=writeback can expose stale data, which could even be a security issue.
> 
> But as long as you make these decisions consciously, they may fit your
> needs.
> 
> >  So apart from losing a few more 
> > seconds of work in case of an error, the fs should be fine.
> 
> This part is correct, barriers on and longer commit time should not
> affect filesystem consistency / integrity.

Ok, I'm more relaxed about my data then. :)

Thanks for your reply,
-- 
Nicolas