On Fri, Jun 30, 2000 at 02:59:41PM -0400, jpb wrote:
>
> How has your experience been with Reiser? I've got a server I'm in the
> process of setting up and have been considering Reiserfs but am kind of
> leery of experimental filesystems.
>
If it's stability and reliability you're worried about (which are probably
the things one should be worried about, of course): rock solid.
I've been consistently amazed at how non-beta and polished this
"experimental" filesystem seems, and the thoroughness of the
utilities for it -- it already has a resizer and all that cool jazz,
which we didn't have for ext2 for years.
This is how it went for me:
1. Downloaded the patch (these Reiser guys are fanatical about this,
they seem to have patches out for kernel versions almost before the
kernels themselves hit the FTP servers).
2. Apply patch - no problems. Select the option in the kernel
configuration - no problem. Kernel compiles fine.
3. Compiled the utilities - a simple "make && make install" issue.
WHAT'S THIS?? Man pages, for an experimental filesystem with
experimental utilities? Complete man pages, that make sense? Whoa.
I don't think we're in Kansas anymore.
4. After booting the new kernel, I checked the man page for mkreiserfs,
and decided the default options would be fine. I ran mkreiserfs on a
Linux partition that I'd cleared out (Reiser doesn't have a partition
ID AFAIK, so one just runs the mkreiserfs utility on a partition one
doesn't mind losing -- no bothering with partition type)
5. Mounted it -- wow, that was simple. Cute advertisement for MP3.com
when a Reiser partition is mounted, by the way. First open-source
filesystem with corporate sponsorship...
6. Now, the fun. I made sure everyone was logged off of the machine,
and I casually flipped the power button off and back on. I watched the
ext2 partition take 5 minutes to fsck, and then... WOW. 3 seconds for
the Reiser of the same size.
Now... on to the performance issue. Two filesystems being otherwise
equal, journaling slows things down a lot, which is why ext3 is such
a dog. Reiser on the other hand is supposed to be miles ahead of
ext2 in terms of performance with the journaling off, and still maintain
a substantial lead with journaling on.
I haven't done any benchmarks, but I haven't noticed much difference.
I'm sure there's a performance increase, but it's not so huge as to
cause a person to exlaim, "Wow, this filesystem access is FAST!"
On a faster hard drive it might make more of a noticable difference,
but the hard drive in the server here is only pulling 10MB/sec.
Thoughts and caveats:
-ReiserFS does not (yet?) support the file attributes that
ext2 does... immutable, append-only, etc. This isn't a very
big deal because most people don't even know those file
attributes exist, and very few people use them, but if you rely
on any of those attributes then you'll have to keep an ext2
filesystem or two around.
-ReiserFS doesn't (yet?) support all of the mount options that
ext2 does, such as noatime and grpid, two of my favorites. So
again, it might not be right for replacign every single
filesystem. It is having its own mount options added to it,
such as an option to disable logging, and option to speed up
access at the cost of using extra disk space, and a (!!!) mount-time
filesystem resizer.
-Although AFAIK you can upgrade to new versions of the ReiserFS
kernel code and maintain compatibility with your filesystem,
old versions of the ReiserFS kernel code can't seem to handle
Resier filesystems created with newer versions... that is to say,
I first patched ReiserFS into the 2.4.0-test1 kernel. While trying
to troubleshoot an unrelated problem we were having, I temporarily
downgraded to 2.2.16, along with an older version of the ReiserFS
patch.... but it couldn't seem to mount the ReiserFS filesystems
I had created under 2.4.0-test1. Actually, I just realized I forgot
to recompile the "new" versions of the utilities that corresponded
with the 2.2.16 kernel, so that might have been the cause of the
problem.
-Although they're doing a great job right now of producing ReiserFS
patches for new kernels very quickly, one still does have to wait,
unless one wants to do some manual patching. I personally was ready
to try 2.4.0-test2 the day it came out, but I had to wait a bit on
the patch. I think Reiser is *SUPPOSED* to be merged into the kernel
before the 2.4 release... can anyone quote a definate yae or nae
on that?
In short, I'm very pleased by the filesystem and how well it works.
I have not had one problem, and I was amazed by the stability and
polish of it all. Definately give it a try on a filesystem or
two, to see what you think. You can also use it as an excuse to
try out a 2.4.0-test kernel if you haven't already... the lure of
the leading edge. I've been using 2.4 for a couple weeks now,
and haven't run into any show-stoppers or anything that required
a reboot.
Good luck!
--
Craig McPherson
Network Admin
Baptist Student Union
Fayetteville, Arkansas
_______________________________________________
Stormlinux-users-list mailing list
[EMAIL PROTECTED]
http://www.stormix.com/community/lists/listinfo/stormlinux-users-list
