[OmniOS-discuss] Clues for tracking down a drastic ZFS fs space difference?

[Chris Siebenmann]

> > From: Chris Siebenmann
> > [...] ZFS level; as you can see, it's added 22 GB of ZFS usage in
> > less than a month while losing 5 GB at the user level.
>
> There was a thread late last year on the developer list that described
> similar symptoms:
>
> http://lists.open-zfs.org/pipermail/developer/2014-October/000887.html
>
> Have you by any chance done zfs recv into this filesystem, and then
> further have the async_destroy feature enabled? If so, you could
> possibly be hitting this issue. zdb -bb on the pool would show an
> increasing amount of deferred frees over time.

 We have done 'zfs recv' into this filesystem (and all of the other
ones; they were all migrated from our old Solaris fileservers to our
new OmniOS ones).

 However I just ran 'zdb -bb' and it kind of suggests we aren't seeing
this:

Blocks  LSIZE   PSIZE   ASIZE     avg    comp   %Total  Type
[...]
    30   346K   62.5K    372K   12.4K    5.54     0.00  deferred free

In general, nothing in 'zdb -bb's list of types is over a GB in LSIZE
or PSIZE except 'ZFS plain file'.

(On the other hand, maybe 'deferred free' is not blocks that are going
to be freed but instead blocks serving as indexes to etc, in which case
this might be a significant number. I don't have data from anything
else to cross-check against.)

	- cks