Mailinglist Archive: opensuse (818 mails)

< Previous Next >
Re: [opensuse] Reiserfs (was ext4 vs btrfs)

On Thu, 12 Jul 2012, Duaine Hechler wrote:
On 07/12/2012 06:43 PM, Cristian Rodríguez wrote:
El 12/07/12 18:29, Hans Witvliet escribió:
I was wondering when should one choose ext4 and when btrfs...

If in doubt ... [see below]

Ofcourse i've googling for it, but in some comparisons they only measure
throughput in general. I agree that _is_ important, but i presume other
factors also play a role...
When dealing with config files (in "/etc/.." or so) it is probably very
handy to be able to do a roll-back.

But when dealing with a large collection of ISO's ?

ext3. I currently have ~1586. Is that a "large collection"?

Last time I checked, XFS worked better for larger files, things might
have changed in the meanwhile though.

XFS is still a beast you have to understand. It does and has been
working for years, but you should know what you're doing, and _why_!
And you have to set options that fit your use case!

I use BTRFS daily and while it usually works, the error handling is
still sketchy to say the least and will BUG or panic in such cases.. but
to be fair, it has improved significantly in recent kernels.

btrfs is still a fledgling with little / flaky error handling /
recovery. Just read the btrfs ML(-Archive). I probably will use it
"soon", but not before a stable fsck for it is available. I'd give it
a couple of years.

Filesystem corruption is also a possibility, personally I got hit once
in the last months and took me a while to figure out I needed to zero
the filesystem log to keep going. fsck will not help you in that case.


Well, I know it's a little passe, but I've been using reiserfs for
about 4 years and love it. The recovery is great ! And I've never had
it totally go belly up on me !

reiserfs is a dead fish driftin' in the water, a fish that has begun
smelling bad years ago. I've almost always[1] called it "reißwolffs"
(german for: (paper-|file-) "shredderfs"). And for a reason. I _do_
know, that a lot of people have never had any problems with it, but
_how_ it behaved when problems _did_ occur simply and outright
disqualified it for me. Classic example: put an image of a reiserfs
on reiserfs and do a reiserfsck [--rebuildtree] (not sure about the
--rebuildtree, but AFAIR usually that was the only useful thing you
could do anyway). The result will be

I for one wait for ext4 to stabilize more, watch btrfs quite
interested, but until further notice, I'll stick with ext3. On 16[2]
filesystems on this box alone, with ~16TiB (partitioned and
formatted!); and another 11.7TiB or so on 13 partitions on the other
box, also all ext3.

I've been using ext2 "forever" (since 2.0.35 at least), sadly I don't
remember since when I use ext3, about 4-6 years? More? Anyway, I've
never had _any_ problems I could attribute to the FS[3] and use ext3
exclusively by now (root partitions I hope without ext2 incompatible
options, I do have "ext3,ext2" in the FS field in fstab for / (and
/boot)) ...

Anyway, to sum it up:

... if you have to ask what FS you should use, then use ext3.

If you have special needs, ask again, ext4, XFS, JFS, zfs, or btrfs
might be more suited to those needs. Read up on the differences.
Unless you know _why_ another FS may be more suited _for you_, don't
even ask, just use ext3. If you do, do ask again.

As long as I see a steady stream of "BUG" mails in the btrfs ML
concerning non-ephemereal situations and with no (stable) btrfsck

JFTR & my 2¢,
-dnh, non-random sig this time ;)

[0] could even have been a patched 2.2.14+, but I doubt that.

[1] I eagerly jumped for it in 2.4.0-test*[0] when it was brand
spanking new, fresh in the kernel, tested and used while it became
"stable", lost data, went back to ext2 and _very_ closely followed
narrations of others about it, that did or did not have problems
with it. Never ever considered going back. Oh, and I looked at
the codebase too ... If you still like reißwolffs, I urge you to
do the same. Have an read ... About reiserfs4?

[2] # df -t ext3 | wc -l

[3] oh, and I did have HW-caused[4] _bad_ crashes, crashes where not even
the reset-button worked! (IIRC at least once the power-button
didn't too, not sure though!). IIRC, I've never had any FS
corruption (that wasn't fixed by fsck/journal-replay on
reboot). The one or two times I had something in lost+found were
when the HDD corrupted the relevant sectors. Anyway, error-
recovery on ext2/3 is much better than on any other FS plus
there's tools for the worst case (debugfs, e2undel(?), ext3grep,

[4] some HW would only work with some specific IO-Ports. Else: Hard
crash. Took me a while.

The steady state of disks is full. -- Ken Thompson
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse+owner@xxxxxxxxxxxx

< Previous Next >