From: Chris Cox on
On Fri, 2010-07-02 at 19:54 +0000, Doug Freyburger wrote:
> Aragorn wrote:
> >
> > That said, reiserfs wasn't all that bad, but if it did ever go wrong,
> > then it went wrong very badly, and it lacked a decent toolchain like
> > ext2/3/4 or XFS
> Lacking such tools puts ReiserFS out of the quesiton for me. EXT3 works
> and it works fine for any of my production database boxes.
> I will point out that a part of the core tool chain is a working dump
> and restore for the filesystem format. That is something lacking across
> Linux in general. I get that Linus doesn't like dump and prefers tar.
> To me that is a significant feature missing and a point against using
> Linux in production at all. But it's only one point in a long list.
> Still, it is as ridiculous that those tools are missing as it is that I
> am not on SourceForge contributing to them. Such is how Linux works
> after all. If I think it's broken I can fix it myself.

I'm going to argue that the lack of a SPECIFIC non-portable (at all)
format dump is NOT a good reason to NOT use a filesystem.

Simply put, people that use lack of fs dump as a reason, REALLY need to
look at the reasons behind their argument (hint, there isn't one).

> > I have no real experience with JFS, but I presume it
> > also has a similarly elaborate toolchain, since it is the default
> > filesystem in AIX.
> AIX JFS2 has a complete tool set. It works great. Extremely stable in
> the face of basically everything. I've had disks fail to spin up after
> a power cycle where I shook them, reinserted them, and started them up
> again. A bunch of resync work later hello big Oracle database on them.

Linux JFS is NOT AIX JFS1 and definitely NOT AIX JFS2. It's not, it's
not, it's not. It fragments like crazy and it is not well maintained
(which is why even SUSE is removing support for it, one of the few that
supported it).

With that said, I believe that even JFS2 has fragmentation problems (as
does ext2/3). One of the least fragmentable filesystems... the infamous

> > Personally, I have however never heard of reiserfs
> > making files disappear, but I did read somewhere that it's rather prone
> > to damaging its own internal trees in the event of an unclean shutdown.

I do agree that the complex on disk structure of reiserfs makes it more
prone to errors (if you can call it that) when you TRY to damage the
filesystem (mind you, the technique can cause other hardware damage as
well, and WILL damage ext2, ext3, etc... if you just practice long
enough). So... taking the gamblers approach that when you do something
evil to the filesystem, reiserfs tends to receive the evil more readily
might actually be a good thing (?). It means you have bigger problems.
And the day your ext3/4, etc tanks due to hard power out.... who will
you cry to then?

Just because you roll the dice and keep landing on 7 or 11 when using
ext3, 4 DOES NOT mean you should keep rolling the dice. Sure, maybe
reiserfs means you need to roll a 5 or something... and sure, that means
higher risk... but the problem is in rolling the dice, NOT the
filesystem. Eliminate the dice roll.

> Regular UNIX filesystems have not been that bad since 1982. On a
> VAX-11-750. That's a long time ago.

:-) Are you talking Ultrix-32? I think you mean something like 1984.