[zfs-discuss] zvol with dedupe space accounting

Mikko Tanner mikko.tanner at gmail.com
Thu Feb 25 04:37:50 EST 2016


Hi,

This is because ext3 on top of zvol has no knowledge of dedup, which is 
on a lower layer. ext3 "sees" 10GB of space, that's it.

Now if you were to make N number of such zvols, then you'd only reserve 
the space for unique blocks _from ZFS perspective_. Each of the zvols 
would still only be 10GB to ext3.

Hope this helps.

BR,
Mikko Tanner

On 25.2.2016 5:32, Max Levine wrote:
> Hi All,
>
> It looks like deduplication is not accounted for when I have a ext3 on 
> top of a deduplicated zvol. I can only write as much as the size of 
> the zvol. see example:
>
> $  zpool create  -O dedup=on  tank /dev/zram0
> $  zfs create  -V 10G -o volblocksize=4k tank/vol01
> $  mkfs.ext3 /dev/zvol/tank/vol01
> $  mount /dev/zvol/tank/vol01 /vol01
> $  dd if=/dev/zero of=/vol01/out bs=1G count=11
> dd: error writing /vol01/out: No space left on device
> 10+0 records in
> 9+0 records out
> 10400808960 bytes (10 GB) copied, 31.2259 s, 333 MB/s
> $  zpool list
> NAME   SIZE  ALLOC   FREE  EXPANDSZ   FRAG    CAP  DEDUP  HEALTH  ALTROOT
> tank  76.5G  58.9M  76.4G         -     0%     0%  1246.41x  ONLINE  -
> $  zfs list
> NAME         USED  AVAIL  REFER  MOUNTPOINT
> tank        10.3G  73.4G    96K  /tank
> tank/vol01  10.3G  73.6G  10.1G  -
> $  zfs get written tank/vol01
> NAME        PROPERTY  VALUE    SOURCE
> tank/vol01  written   10.1G    -
> $
>
>
> Is this expected? how to take advantage of deduplication on a zvol in 
> this case?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20160225/acaba64e/attachment.html>


More information about the zfs-discuss mailing list