Re: Converting ext2/ext3 fs to ReiserFS or XFS Journaling File Systems

From: Christopher Browne (cbbrowne_at_acm.org)
Date: 05/28/04 

Date: 28 May 2004 03:52:48 GMT

The world rejoiced as "Ruby Tuesdays" <NoSpamPlease_rubytuzdaiz@yahoo.com> wrote:
> Is it possible to convert ext2/ext3 filesystem of live system to one of the
> most flexible/fastest (ReiserFS? XFS?)Linux 2.6 Journaling Filesystem? Do I
> have compile the tools to do that?

Some benchmarking I did on database-related activity showed JFS to be
fastest, by a small margin, a little while back. But that may well
have been skewed by what hardware I was running. (Quad-Xeon, with
RAID array of 12 SCSI drives, complete with battery-backed cache on
the controller.)

The way to "convert" to another filesystem is to do the following:

  1. Unmount the ext? filesystem.
  2. Do an appropriate mkfs on the device for the desired new
      filesystem.
  3. Mount the new filesystem.
  4. Use your backup system to recover the files back onto the now
      empty filesystem.

If #4 is not an option, then you have vastly bigger problems than
which filesystem is fastest. You should solve that problem before
attempting such transitions.

> Perhaps if I can get some recommendation of which one to use? Pluses
> and minuses of using one over the other. How about Veritas File
> Systems?

The best you can get are anecdotal results of:

 a) Someone claiming that, for _their_ unique workload, one FS is
    preferable to the others;

 b) Someone claiming that they have either suffered considerable
    data losses due to FS corruption, or that they have NOT thus
    suffered, under one FS or another.

None of it appears particularly provable. 

-- 
output = ("cbbrowne" "@" "cbbrowne.com")
http://cbbrowne.com/info/fs.html
"... and  the REALLY GOOD  THING, is that  after you have gone  to the
trouble of compiling that once, you can run it MANY MANY times!!!"
-- Arthur Norman

Relevant Pages

  • Re: 2.6.33-rc1: kernel BUG at fs/ext4/inode.c:1063 (sparc)
    ... But I can't compile 2.6.33-rc2:) ... but i still can not reproduce your bug. ... Filesystem magic number: 0xEF53 ... Reserved blocks gid: 0 ...
    (Linux-Kernel)
  • Re: 2.6.33-rc1: kernel BUG at fs/ext4/inode.c:1063 (sparc)
    ... But I can't compile 2.6.33-rc2:) ... but i still can not reproduce your bug. ... Filesystem magic number: 0xEF53 ... Reserved blocks gid: 0 ...
    (Linux-Kernel)
  • Re: Converting ext2/ext3 fs to ReiserFS or XFS Journaling File Systems
    ... > have compile the tools to do that? ... The way to "convert" to another filesystem is to do the following: ... If #4 is not an option, then you have vastly bigger problems than ... Someone claiming that, for _their_ unique workload, one FS is ...
    (comp.os.linux.setup)
  • Re: [SLE] disadvantages of 32bit SuSE10.1
    ... any given file can be bigger than your whole filesystem. ... but i'm guessing that with a dual xeon you won't notice much ... Maybe if you compile a ... compiles about 50% faster than my single-core boxes (all approximately ...
    (SuSE)
  • Re: URGENT RAM disk problem
    ... > time to compile has almost increased by 2-4 times compared to ... > non-RAM disk build). ... the ramdisk block dev in Linux has some non-obvious properties. ... files from a filesystem on a ramdisk are often ...
    (comp.os.linux.development.system)