Subject: Re: what's it take to get a journaling filesystem?
To: None <tech-kern@netbsd.org>
From: Matthew Mondor <mm_lists@pulsar-zone.net>
List: tech-kern
Date: 12/13/2005 05:44:20
On Mon, 12 Dec 2005 22:39:18 -0500
Sean Davis <dive-nb@endersgame.net> wrote:
> Is anyone working on such a thing? FFS is great, and suits my purposes
> perfectly, except when I have to recover from a crash. When I started using
> NetBSD, I had perhaps 4GB of disk space to deal with, so it was a non-issue
> then, but now, it's getting quite annoying.
This should require confirmation of a DragonFlyBSD developer, but I
believe that Matt Dillon currently works on FS-independent journalling
support, although it apparently is incomplete and unstable at current
time. I'm not sure about the approach it uses nor about it's actual
development status, I only read a few status report emails from Matt
about it. It might be worth investigating. If it uses something like a
directory on the FS for log files, there surely also are changes to fsck
involved... If I understand, one of the goals of this logging also is
to support live mirrors.
There was once some academic project where students would have worked on
adding journalling to FFS on one of the BSDs (I don't remember if it's
FreeBSD or NetBSD, but I contacted the project manager a year ago or so
and never got a reply, I assume that there was a change of plans).
LFS indeed has nice points. I used it a few years back and also was
impressed by its performance. However, after a week or so of uptime the
box would crash hard (and I couldn't get a core dump to be produced
about it at the time). I reverted to FFS only but would like to give
LFS a try again soon, considering that I recently was told that it's now
more stable than it used to be.
Apparently that LFS would have slower read performance than FFS after
being used for some time, though, from what I read some time back. Real
world performance tests and benchmarks definitely would be useful...
Another aspect I would like to have more information about pertains to
potential use of LFS on flash devices. If I understand a bit how LFS
works, wouldn't it potentially allow more wear distribution than FFS or
FAT? Knowing that LFS requires a fair partition size to be able to work
properly, though (I remember testing with various small filesystems and
it was useless for floppy-sized ones, obviously :) But the size of
flash storage devices is increasing and their price lowering nowadays.
Matt
--
Note: Please only reply on the list, other mail is blocked by default.
Private messages from your address can be allowed by first asking.