[zfs-discuss] recommendations for a 50 TB zpool?

Ted Cabeen ted.cabeen at lscg.ucsb.edu
Mon Nov 5 12:54:37 EST 2018


On 11/4/18 1:46 AM, Ulli Horlacher via zfs-discuss wrote:
> On Sat 2018-11-03 (20:02), Ted Cabeen via zfs-discuss wrote:
>> I have a 700TB array built on 12-drive RAIDZ2 vdevs.  Works fine,
>> rebuilds are long, but not super long, and should get better now that
>> sequential resilver is almost released.
> 
> "sequential resilver" is a new ZFS feature?
> I have not used ZFS for the last 3 years, so I am a little bit out of
> business :-}

Yes, the current 0.7.X resilver code just follows the tree when 
re-silvering, even when that results in pulling data from all-over the 
drive.  The new code allows for re-ordering of the reads to improve 
efficiency.

>> Do you need synchronous writes?  If not, with sequential reads, you
>> don't need SLOG or L2ARC, just put in a decent amount of memory, and set
>> recordsize to 1MB.
> 
> "synchronous writes" means data will be written directly to disks?
> We have a good UPS, so synchronous writes are not that important.
> The server will have 32 GB of RAM, the application needs less than 1 GB.

A UPS is important, but that doesn't really affect synchronous writes. 
The question with sync writes is "In a crash or other system failure, 
can you tolerate losing up to 5 seconds of writes that were reported to 
the application as written successfully, but which were actually lost?" 
In some cases, that's not a problem.  In others, it is.  YMMV.

--Ted


More information about the zfs-discuss mailing list