I agree, all the talk of certain filesystems not needing defragmentation (not
just XFS), is all just talk. In the real world I have seen heavy
defragmentation on all my machines, regardless of filesystem. All it takes is
a limited amount of freespace and/or a lot of file creation/deletion. Mail
servers as already mentioned are a good example of the heavy file
creation/deletion scenario, as are audio/video editing workstations. On a/v
editing workstations I have seen greater than 90% fragmentation in a fairly
short time using virtually any filesystem. The ability to easily defrag XFS is
pretty significant in my choice of filesystems.
-----Original Message-----
From: linux-xfs-bounce@xxxxxxxxxxx on behalf of Fong Vang
Sent: Fri 8/26/2005 10:48 AM
To: Jan-Frode Myklebust
Cc: linux-xfs@xxxxxxxxxxx
Subject: Re: file system defragmentation
On 8/26/05, Jan-Frode Myklebust <Jan-Frode.Myklebust@xxxxxxxxxxx> wrote:
> On Fri, Aug 26, 2005 at 12:16:16AM -0700, Fong Vang wrote:
> > I've read in many places that file systems on Linux do not suffer the
> > same fragmentation problems of Windows systems. No one has provided a
> > clear explanation as to why fragmentation is not an issue for file
> > systems such as XFS.
>
>
> I think it is a problem for XFS also.. at least for our slow growing
> mailserver with lots of mbox-files.
>
>
> > Also, is there a way to determine the percentage of fragmentation on
> > an XFS file system?
>
> # xfs_db -c frag -r /dev/sdc1
> actual 59707, ideal 46970, fragmentation factor 21.33%
>
> and individual file fragmentation can be checked with xfs_bmap:
>
> # xfs_bmap usersmbox
> usersmbox:
> 0: [0..744815]: 31464808..32209623
> 1: [744816..744831]: 31464792..31464807
> 2: [744832..744959]: 31464664..31464791
> 3: [744960..745063]: 31464048..31464151
> 4: [745064..745215]: 31463896..31464047
> 5: [745216..745335]: 31463768..31463887
> 6: [745336..745471]: 31463632..31463767
> 7: [745472..745583]: 31463504..31463615
>
> and you can defrag either individual files:
>
> # xfs_fsr usersmbox
> # xfs_bmap usersmbox
> usersmbox:
> 0: [0..747423]: 142372792..143120215
>
> or a whole live partiion:
>
> # xfs_fsr /dev/sdc1
>
> (btw: i think xfs_db -c frag will be a bit slow in reporting the updated
> fragmentation after a xfs_fsr, at least that's the impression I got
> the last time I ran it)
This is very useful. Thank you. How stable is xfs_fsr?
> -jf
>
|