Re: [zfs] recordsize: unexpected increase of disk usage when increasing it

On Tue, Jan 18, 2022 at 3:23 PM Florent Rivoire <florent@rivoire.fr> wrote:
> I didn't think at all about compression (which I know exists and
> usually has very positive impact) during my tests...
> and it's indeed disabled in my tests (I'm only seeing it now):

I did my tests again: exactly the same procedure as my initial email,
just adding `-O compression=lz4` in zpool create.
And now, the allocated size of disks (shown by zpool list) is the same
in both cases (128K & 1M recordsize): 224G.
So, good news :)

Note: we can still see a difference in the "logicalused": 226G vs 232G
(those are the allocated sizes that I had with no compression):

recordsize 128K:
-----------------------
# zfs get all bench |egrep "(used|referenced|written)"
bench  used                  224G                   -
bench  referenced            224G                   -
bench  usedbysnapshots       0B                     -
bench  usedbydataset         224G                   -
bench  usedbychildren        1.83M                  -
bench  usedbyrefreservation  0B                     -
bench  written               224G                   -
bench  logicalused           226G                   -
bench  logicalreferenced     226G                   -

recordsize 1M:
-----------------------
# zfs get all bench |egrep "(used|referenced|written)"
bench  used                  224G                   -
bench  referenced            224G                   -
bench  usedbysnapshots       0B                     -
bench  usedbydataset         224G                   -
bench  usedbychildren        1.88M                  -
bench  usedbyrefreservation  0B                     -
bench  written               224G                   -
bench  logicalused           232G                   -
bench  logicalreferenced     232G                   -

-- 
Florent