On Thu, Feb 10, 2000 at 11:21:08PM -0500, Len Budney wrote:
> Sam <[EMAIL PROTECTED]> wrote:
> > On Thu, 10 Feb 2000, Len Budney wrote:
> >
> > > Which brings us back to your mistake; you make a claim about
> > > ``journalling filesystems'' which is true for some, and false for
> > > others...
> >
> > I did not say that all journaling file systems do not sync both the data
> > and the metadata, just that it's not a requirement for implementing a
> > journaled filesystem.
>
> Okay, conceded. We agree that ``reliable'' is not spelled ``journalling''.
> I mistook you to be implying that journalled filesystems generally are
> subject to email loss. I apologize for taking your statement farther
> than you meant it.
>
> What threw me for a loop was your apparent focus on journalling itself; it
> seemed like a red herring. Nobody made any silly claims about journalling
> on this list; indeed there was a recent discussion about the reliability
> impact of soft updates on FFS--giving us a good example of both a reliable
> journalled filesystem, and an unreliable journalled filesystem.
Thanks. Yes. Let's make the clarification. FFS with soft updates is NOT a
journaling file system.
It will however be significally faster than any journaling filesystem(*) but
it still needs to get fsyncs AFTER e.g. a return from rename() to be sure
that it effectively has happened on disk.
It DOES NOT MEAN that rename() is no longer an atomic operation in respect
to other applications, but if you get a crash between the rename() and the
fsync() you cannot be sure that the rename has been done when you come up
again.
The impact of qmail is however dim to me.
Please enlighten upon the problems that can arise.
Do we have to add extra fsyncs here?
/magnus
(*) Kirk McKusick is presenting these tests at Usenix this summer, he said
to me yesterday.
--
http://x42.com/
