Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 30 May 2001 00:27:14 -0700
From:      dannyman <dannyman@toldme.com>
To:        Jonathan Chen <jonathan.chen@itouch.co.nz>
Cc:        Bill Moran <wmoran@iowna.com>, freebsd-questions@FreeBSD.ORG
Subject:   Re: Remount Filesystems
Message-ID:  <20010530002714.L14366@dell.dannyland.org>
In-Reply-To: <20010525103021.B40969@itouchnz.itouch>; from jonathan.chen@itouch.co.nz on Fri, May 25, 2001 at 10:30:21AM %2B1200
References:  <SAK.2001.05.24.raeessor@support10> <20010525094056.B37339@itouchnz.itouch> <3B0D8A80.596CC3B7@iowna.com> <20010525103021.B40969@itouchnz.itouch>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, May 25, 2001 at 10:30:21AM +1200, Jonathan Chen wrote:
> On Thu, May 24, 2001 at 06:26:09PM -0400, Bill Moran wrote:
> 
> [...]
> > I don't understand why you'd bother with softupdates on / anyway? Just
> > from my perspective, I try to keep the / fs as reliable as possible and,
> > although softupdates is very reliable, it does have a slightly higher
> > incidence of crash corruption than standard sync.
> 
> Really? My understanding of softupdates was that it keeps that metadata
> in a more stable state, and thus makes your filesystem *less* prone to
> fsck problems.

Yes, it keeps meta-data in a more stable state by cacheing it
intelligently in memory before writing it.  Thus, the filesystem is less
prone to fsck problems, but you do risk losing data if the metadata is
not written to disk before a crash.

> Otherwise, why would anyone want to enable softupdates on a production
> system?

You can re-mount a filesystem without fscking it.  You get better
performance and availability, but you risk losing some data in a crash.
But then, for important storage operations, like writing e-mail to disk,
your application will block on sync() returning so that it knows that
the data was successfully committed to disk.

But then your data is being trusted to algorithms that haven't been
around for debugging unanticipated edge cases as long as UFS has. :)

Don't take my word for it, though, I've just been lurking around long
enough to regurgitate these arguments that others have made.  For the
record, I only enable SoftUpdates when it seems that the application
could use enhanced disk performance.  For example, an e-mail gateway,
e-mail server, perhaps LDAP, MySQL, or a log server ...

-danny

-- 
http://dannyman.toldme.com/

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010530002714.L14366>