DragonFly users List (threaded) for 2005-03
Re: dragonfly pdf documentation
Adrian Nida wrote:
Joerg Sonnenberger wrote:
On Wed, Mar 30, 2005 at 11:57:21PM +1000, Steven Shaw wrote:
Reiserfs and ext3 don't need fsck either though do they? Perhaps you
were running early/unstable versions of either ext3 or reiserfs?
In theory yes. In practise ext3 can destroy directories under certain
circumstance, reiserfs its whole tree. It happened over the years
quite a few times to various people.
Heh, I really hate admitting this, but I lost an entire disk that was
running ext3. I'm sure it was something *I* did. However, when every
single directory and file ends up in /lost+found, it's hard to figure
out what cause it. Of course, praying to the backup gods only works if
the person responsible for backups has given them a sacrifice. Arrgh,
that was a bad day.
Not to worry. I once had a ReiserFS live for a whole month with
no problems. As it was on a rather massive UPS, it might have
gone even longer if I hadn't needed the drive for testing
OTOH, the half-dozen attempts *before* that one never lasted
through either a single hard-read error or a couple of power
outages. 'Good in concept' is ReiserFS, and 'maybe someday..',
but for *now* one best pray to the Gods of harware and power.
And ext2 and 3? Fewer worries about fragmentation
than MS NTFS, but not as robust if/as/when subjected
XFS and JFS, OTOH, were 'donated' by SGI and IBM. Both
have had scores of years, thousands of man-hours, and
*serious* money invested in development and test.
Both are rock solid, as is MKM's Berkeley FFS --> UFS.
Boot from XFS, use JFS for non-boot storage in the
places it serves better than XFS, and you have the
foundations for as durable a Linux as can be.
For sure it will guarantee a more affordable
'entry price', lots more service revenue, and
an easy upgrade to AIX-5L, which has to be a
large part of the reason IBM love's Linux.
For those of us with a more austere business
model, the *BSD's are a much lighter burden.