On Thu, Oct 27, 2016 at 10:15:27PM +0200, Yuri D'Elia wrote:
> It's hard to reproduce at will. The incomplete journal is consistently
> the INBOX folder of an outlook365 server. But this is because this is
> also the slowest folder to sync, and thus most likely to be interrupted.
>
you wouldn't ha
On Fri, Oct 28, 2016 at 10:04:12AM +0200, Oswald Buddenhagen wrote:
> > As for truncation, this might still happen if the file is not fsynced
> > explicitly at critical transaction points (including before fclose).
> >
> you're not getting truncation, but data corruption, as that's what
> appendin
On Fri, Oct 28, 2016 at 10:59:42AM -0400, Theodore Ts'o wrote:
> On Fri, Oct 28, 2016 at 10:04:12AM +0200, Oswald Buddenhagen wrote:
> > > As for truncation, this might still happen if the file is not fsynced
> > > explicitly at critical transaction points (including before fclose).
> > >
> > you'
On Fri, Oct 28, 2016 at 07:21:58PM +0300, Evgeniy Berdnikov wrote:
>
> Userspace process generally DO NOT read raw blocks from disk.
> It uses kernel's buffer space, so buffered data should be consistent
> regardless of disk contents.
>
> Oswald is right, process completion (by exit() or inte
On Fri, Oct 28, 2016 at 02:36:42PM -0400, Theodore Ts'o wrote:
> On Fri, Oct 28, 2016 at 07:21:58PM +0300, Evgeniy Berdnikov wrote:
> > Oswald is right, process completion (by exit() or interrupt by signal)
> > can only truncate data, but can not corrupt. Only system crash can
> > lead to data c
On Fri, Oct 28, 2016 at 02:36:42PM -0400, Theodore Ts'o wrote:
> On Fri, Oct 28, 2016 at 07:21:58PM +0300, Evgeniy Berdnikov wrote:
...
> > Oswald is right, process completion (by exit() or interrupt by signal)
> > can only truncate data, but can not corrupt. Only system crash can
> > lead to da
