Btrfs devel - Jan 2011 - Defragmentation Not Working with Directory as Argument

Defragmentation is not working when a directory is supplied as the
argument.  When a file name is supplied as the argument,
defragmentation works well.

This can be demonstrated as follows:

# mount /dev/sdb2 /mnt/btrfs
# cd /mnt/btrfs/
# cp /usr/src/linux/fs/btrfs/zlib.c  .    (any suitable file will do,
small text files seemed appropriate)
# cp zlib.c fragmented-zlib.c
# btrfs fi sync /mnt/btrfs
# cat zlib.c >> fragmented-zlib.c
# btrfs fi sync /mnt/btrfs
# filefrag zlib.c fragmented-zlib.c
zlib.c: 1 extent found
fragmented-zlib.c: 2 extents found
# btrfs filesystem defragment /mnt/btrfs/
# btrfs fi sync /mnt/btrfs
# filefrag zlib.c fragmented-zlib.c
zlib.c: 1 extent found
fragmented-zlib.c: 2 extents found

Now, supply a filename as the argument

# btrfs filesystem defragment /mnt/btrfs/fragmented-zlib.c
# btrfs fi sync /mnt/btrfs
# filefrag zlib.c fragmented-zlib.c
zlib.c: 1 extent found
fragmented-zlib.c: 1 extent found

I''ve also tested defragmentation on my normal btrfs filesystems, and
have seen the same results.  I''ve never seen a file defragmented when
using a directory as the argument.

Also, I was wondering if one of the Btrfs developers could clarify the
intended scope of defragmentation when a directory is supplied as the
argument.

I have seen comments on the mailing list indicating that only the
directory supplied is intended to be defragmented, and defragmentation
will not recurse to subdirectories below.

However, if I look at the code, and read the documentation for the
older btrfsctl command (which seems to be the basis for the current
btrfs command), it seems like the intention is to recursively
defragment the entire tree below the directory supplied in the
argument.
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Monday, January 24, 2011 19:11:07 Mitch Harder wrote:
> Defragmentation is not working when a directory is supplied as the
> argument.  When a file name is supplied as the argument,
> defragmentation works well.
> 
> This can be demonstrated as follows:
> 
> # mount /dev/sdb2 /mnt/btrfs
> # cd /mnt/btrfs/
> # cp /usr/src/linux/fs/btrfs/zlib.c  .    (any suitable file will do,
> small text files seemed appropriate)
> # cp zlib.c fragmented-zlib.c
> # btrfs fi sync /mnt/btrfs
> # cat zlib.c >> fragmented-zlib.c
> # btrfs fi sync /mnt/btrfs
> # filefrag zlib.c fragmented-zlib.c
> zlib.c: 1 extent found
> fragmented-zlib.c: 2 extents found
> # btrfs filesystem defragment /mnt/btrfs/
> # btrfs fi sync /mnt/btrfs
> # filefrag zlib.c fragmented-zlib.c
> zlib.c: 1 extent found
> fragmented-zlib.c: 2 extents found
> 
> Now, supply a filename as the argument
> 
> # btrfs filesystem defragment /mnt/btrfs/fragmented-zlib.c
> # btrfs fi sync /mnt/btrfs
> # filefrag zlib.c fragmented-zlib.c
> zlib.c: 1 extent found
> fragmented-zlib.c: 1 extent found
> 
> I''ve also tested defragmentation on my normal btrfs filesystems,
and
> have seen the same results.  I''ve never seen a file defragmented
when
> using a directory as the argument.
That''s because this defragments only directory metadata.
> Also, I was wondering if one of the Btrfs developers could clarify the
> intended scope of defragmentation when a directory is supplied as the
> argument.
I submited apropriate patches few days back to update man and help messages
> I have seen comments on the mailing list indicating that only the
> directory supplied is intended to be defragmented, and defragmentation
> will not recurse to subdirectories below.
that''s true
> However, if I look at the code, and read the documentation for the
> older btrfsctl command (which seems to be the basis for the current
> btrfs command), it seems like the intention is to recursively
> defragment the entire tree below the directory supplied in the
> argument.
Then it looks like the userspace part of implementation changed, don''t
know
why though.

Regards,
-- 
Hubert Kario
QBS - Quality Business Software
02-656 Warszawa, ul. Ksawerów 30/85
tel. +48 (22) 646-61-51, 646-74-24
www.qbs.com.pl
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Mon, Jan 24, 2011 at 12:21 PM, Hubert Kario <hka@qbs.com.pl>
wrote:
> On Monday, January 24, 2011 19:11:07 Mitch Harder wrote:
>> Defragmentation is not working when a directory is supplied as the
>> argument.  When a file name is supplied as the argument,
>> defragmentation works well.
>>
>> This can be demonstrated as follows:
>>
>> # mount /dev/sdb2 /mnt/btrfs
>> # cd /mnt/btrfs/
>> # cp /usr/src/linux/fs/btrfs/zlib.c  .    (any suitable file will do,
>> small text files seemed appropriate)
>> # cp zlib.c fragmented-zlib.c
>> # btrfs fi sync /mnt/btrfs
>> # cat zlib.c >> fragmented-zlib.c
>> # btrfs fi sync /mnt/btrfs
>> # filefrag zlib.c fragmented-zlib.c
>> zlib.c: 1 extent found
>> fragmented-zlib.c: 2 extents found
>> # btrfs filesystem defragment /mnt/btrfs/
>> # btrfs fi sync /mnt/btrfs
>> # filefrag zlib.c fragmented-zlib.c
>> zlib.c: 1 extent found
>> fragmented-zlib.c: 2 extents found
>>
>> Now, supply a filename as the argument
>>
>> # btrfs filesystem defragment /mnt/btrfs/fragmented-zlib.c
>> # btrfs fi sync /mnt/btrfs
>> # filefrag zlib.c fragmented-zlib.c
>> zlib.c: 1 extent found
>> fragmented-zlib.c: 1 extent found
>>
>> I''ve also tested defragmentation on my normal btrfs
filesystems, and
>> have seen the same results.  I''ve never seen a file
defragmented when
>> using a directory as the argument.
>
> That''s because this defragments only directory metadata.
>
>> Also, I was wondering if one of the Btrfs developers could clarify the
>> intended scope of defragmentation when a directory is supplied as the
>> argument.
>
> I submited apropriate patches few days back to update man and help messages
>
>> I have seen comments on the mailing list indicating that only the
>> directory supplied is intended to be defragmented, and defragmentation
>> will not recurse to subdirectories below.
>
> that''s true
>
>> However, if I look at the code, and read the documentation for the
>> older btrfsctl command (which seems to be the basis for the current
>> btrfs command), it seems like the intention is to recursively
>> defragment the entire tree below the directory supplied in the
>> argument.
>
> Then it looks like the userspace part of implementation changed,
don''t know
> why though.
>
Thanks!  I see where they''ve added some clarification on this issue to
the wiki in the Problem FAQ.

https://btrfs.wiki.kernel.org/index.php/Problem_FAQ#Defragmenting_a_directory_doesn.27t_work

I understand now that the function performed when a directory is
supplied as the argument is different than the function performed when
a file or file wildcard is supplied as the argument.

I''m concerned that there will be a vast number of new btrfs users who
have spent their computer life immersed in the command-line syntax
convention where supplying a directory as an argument implies the
files located in that directory.

It might make sense to break out the defragmentation of leaf metadata
into a separate command so that it''s clear to users what operation is
being performed.

It will save a lot of frustration from both the new users who are
learning the new Btrfs concepts and the advanced users who have to
field the questions.
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html