LXC agregate file size significantly differs from zfs list and du -sh report

[Sasha]

Hi, guys.
I'm really in stuck how is it possible. zfs list shows

NAME                           USED  AVAIL     REFER  MOUNTPOINT
...
rpool/data/subvol-101-disk-3   427G   273G      427G  /rpool/data/subvol-101-disk-3
...

and df -h inside LXC 101 shows
Filesystem Size Used Avail Use% Mounted on
rpool/data/subvol-101-disk-3 700G 429G 272G 62% /var/www/files.local (storage mount point)

But correct files volume itself over 700G

 

[Dunuin]

Try zfs list -o space for additional info. And keep in mind that ZFS also does block level compression by default.

 

[Sasha]

I guess You are right about compression! Just look at this:

root@hq:~# zfs get compression rpool/data/subvol-101-disk-3
NAME                          PROPERTY     VALUE           SOURCE
rpool/data/subvol-101-disk-3  compression  on              inherited from rpool

root@hq:~# zfs get compressratio rpool/data/subvol-101-disk-3
NAME                          PROPERTY       VALUE  SOURCE
rpool/data/subvol-101-disk-3  compressratio  1.69x  -

It's interesting to figure out a cost of that... I mean performance cost...

 

[Dunuin]

"Compresson=on" is using LZ4 as algorhithm. There usually is no noticeable performance penalty when using it with SATA/SAS disks. ZFS also doesn't compress everything. It tries to compress blocks of a file and if that file isn't well compressable, it will skip the compression for those blocks. And LZ4 doesn't got a huge compression efficieny but it is super fast and needs very few CPU ressources. So unless you run some super fast NVMe SSDs and see a bottleneck there, it is usually best to keep LZ4 compression enabled.