Re: ZFS on root with dedupe

On Apr 22, 2023, at 1:52 PM, Andrea Venturoli <ml@netfence.it> wrote:

> On 4/22/23 17:46, Paul Mather wrote:
> 
>> Deduplication is a per-fileset option
> 
> Oh!
> I always thought it was a per-pool setting!!!
> Did it change, perhaps after the switch to OpenZFS?


I believe it has always been a per-fileset property.  I think it is seen as a "per-pool" setting because of the performance impact it can have on the entire pool by enabling it on a fileset.

(
> In fact the handbook says:
> 
>> To activate deduplication, set the dedup property on the target pool:
>> # zfs set dedup=on pool
> 
> Isn't this incoherent?
> I mean:
> zfs set ... *dataset*
> *zpool* set ... pool
> 
> Should it be corrected?


The handbook appears out of date in this case.  Even the example, as you point out, is confusing because it references a zfs(8) command, which acts upon filesets.

For 13-STABLE (and the most recent 13.2-RELEASE), the zfsconcepts(7) and zfsprops(7) manual pages both refer to deduplication being enabled for a "file system" or "dataset".  The zfsprops(7) manual page includes the "dedup" property in per-fileset (or dataset as the manual page calls it) properties: "Configures deduplication for a dataset."  The zfsconcepts(7) manual page states, "If a file system has the dedup property enabled, duplicate data blocks are removed synchronously."  I interpret "file system" to be the same as fileset/dataset.  These are fileset/dataset properties, not pool properties.

It's difficult to keep the FreeBSD Handbook accurate because it potentially covers many simultaneous supported versions of FreeBSD.  I always treat the manual pages for an installed FreeBSD version to be more accurate than the Handbook.

Cheers,

Paul.