On Thu, 24 Feb 2005 22:28:28 +0000
Dieter <netbsd@sopwith.solgatos.com> wrote:

> Power can and will fail.  Even with a UPS.  :-(
=3D> NVRAM cache.

On the other side: If power fails or the hardware breaks you are
fsck(8)ed anyway. In this situation all you can hope is that only a smal
damage has happend. Some FS tend to be more robust in such situations,
some less. For me it would be OK in this case when the FS is in a state
a few seconds before the crash. The last data written is lost, but the
FS can recover fast. (Yes, I read how WAFL works.)

> In reading this thread, it sounds like FFS is the preferred FS for
> avoiding data loss on *BSD.=20
Well. In practice there are only two local FS for (Net)BSD: FFS(1/2) and
LFS. LFS ist not really usable (at the moment), so...

> If one is forced to run Linux :-( which FS is the least likely
> to lose data, and how much worse is it than NetBSD's FFS?
At work, where I have to take care of a herd of Penguins, I use XFS. I
once tried JFS and it was way to buggy. (Lost data, crashes.) ext2: Wane
fsck(8) a 250 GB FS with lots of small files? No thanks. ext3: Never
tried because I have seen fsck(8) destroying an ext(2), thank you.=20
reiserfs: Still incomplete last time I checked (no quotas, trouble with
2 GB file size, ... An unexcusable design flaw for an FS designed in the
late 90's.) XFS works, speed is OK, there are repair and dump / restore
tools, snapshots, online resizeable, ... CXFS, ...
--=20


tsch=FC=DF,
       Jochen

Homepage: http://www.unixag-kl.fh-kl.de/~jkunz/