Re: btrfs and deduplication
On Tue, Sep 25, 2018 at 9:56 AM, Markus Raps <m.raps@rapsplace.de> wrote:
> Hi there,
>
> currently iam trying to get deduplication working in debian/btrfs
>
> #so i created a btrfs filesystem
>
> mkfs.btrfs /dev/vdb1
> mkdir /mnt/btrfs
> mount /dev/vdb1 /mnt/btrfs
>
> # create some random file
> dd if=/dev/urandom of=/mnt/btrfs/img bs=1M count=1024
> for i in {1..30}; do cp /mnt/btrfs/img /mnt/btrfs/img$i; done
>
> # dedup this stuff
> jdupes -S -B -r /mnt/btrfs/
> Examining 31 files, 1 dirs (in 1 specified)
> Deduplication done (30 files processed)
>
> raps-debian btrfs # btrfs fi df /mnt/btrfs/
> Data, single: total=31.50GiB, used=30.96GiB
> System, DUP: total=32.00MiB, used=16.00KiB
> Metadata, DUP: total=222.94MiB, used=33.69MiB
> GlobalReserve, single: total=33.56MiB, used=0.00B
>
> raps-debian btrfs # btrfs fi usage /mnt/btrfs/
> Overall:
> Device size: 32.00GiB
> Device allocated: 32.00GiB
> Device unallocated: 1.04MiB
> Device missing: 0.00B
> Used: 31.03GiB
> Free (estimated): 549.41MiB (min: 549.41MiB)
> Data ratio: 1.00
> Metadata ratio: 2.00
> Global reserve: 33.56MiB (used: 0.00B)
>
> Data,single: Size:31.50GiB, Used:30.96GiB
> /dev/vdb1 31.50GiB
>
> Metadata,DUP: Size:222.94MiB, Used:33.69MiB
> /dev/vdb1 445.88MiB
>
> System,DUP: Size:32.00MiB, Used:16.00KiB
> /dev/vdb1 64.00MiB
>
> Unallocated:
> /dev/vdb1 1.04MiB
> raps-debian btrfs #
>
>
>
> hm ... hasn't worked.
> wrong tool ? did i missed something
> or have i completely misunderstood deduplication?
I can not answer this, but as a long time user of btrfs who has never
tried deduplication I got curious, so I tried to recreate your result
on my debian testing workstation, and for me it worked. Here is the
log of exactly how I created the filesystem, and the output from btrfs
fi usage before and after running jdupes. As you can see from the
usage after deduplication, both the allocated and used space has
decreased to what would be expected.
Perhaps your kernel is too old? It is possible that you are being hit
by the issue described here:
https://btrfs.wiki.kernel.org/index.php/Changelog#By_feature
"The range for out-of-band deduplication implemented by the
EXTENT_SAME ioctl will split the range into 16MiB chunks. Up to now
this was the overall limit and effectively only the first 16MiB was
deduplicated."
I can't say I understand much about it, but it was fixed in kernel
4.18 which happens to be what I use.
# btrfs --version
btrfs-progs v4.17
# uname -a
Linux spacelab 4.18.0-1-amd64 #1 SMP Debian 4.18.6-1 (2018-09-06)
x86_64 GNU/Linux
# lvcreate spacelab -n dedup -L 100G
# mkfs.btrfs /dev/spacelab/dedup
# mount /dev/spacelab/dedup /mnt/
# dd if=/dev/urandom of=/mnt/img bs=1M count=1024
# for i in {1..30}; do cp /mnt/img /mnt/img$i;done
# btrfs fi usage /mnt
Overall:
Device size: 100.00GiB
Device allocated: 33.02GiB
Device unallocated: 66.98GiB
Device missing: 0.00B
Used: 31.04GiB
Free (estimated): 67.98GiB (min: 67.98GiB)
Data ratio: 1.00
Metadata ratio: 1.00
Global reserve: 33.23MiB (used: 0.00B)
Data,single: Size:32.01GiB, Used:31.01GiB
/dev/mapper/spacelab-dedup 32.01GiB
Metadata,single: Size:1.01GiB, Used:33.36MiB
/dev/mapper/spacelab-dedup 1.01GiB
System,single: Size:4.00MiB, Used:16.00KiB
/dev/mapper/spacelab-dedup 4.00MiB
Unallocated:
/dev/mapper/spacelab-dedup 66.98GiB
# jdupes -S -B -r /mnt
Scanning: 31 files, 1 items (in 1 specified)
Deduplication done (30 files processed)
# btrfs fi usage /mnt
Overall:
Device size: 100.00GiB
Device allocated: 2.02GiB
Device unallocated: 97.98GiB
Device missing: 0.00B
Used: 1.00GiB
Free (estimated): 97.99GiB (min: 97.99GiB)
Data ratio: 1.00
Metadata ratio: 1.00
Global reserve: 16.00MiB (used: 0.00B)
Data,single: Size:1.01GiB, Used:1.00GiB
/dev/mapper/spacelab-dedup 1.01GiB
Metadata,single: Size:1.01GiB, Used:1.38MiB
/dev/mapper/spacelab-dedup 1.01GiB
System,single: Size:4.00MiB, Used:16.00KiB
/dev/mapper/spacelab-dedup 4.00MiB
Unallocated:
/dev/mapper/spacelab-dedup 97.98GiB
Reply to: