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

Ted Cabeen ted.cabeen at lscg.ucsb.edu
Sat Nov 3 23:02:42 EDT 2018


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.

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.

If you need synchronous writes, use Optane for SLOG.

CentOS is fine, and pulls in the KMOD packages from the ZoL team easily.

--Ted

On 11/3/2018 5:32 PM, Ulli Horlacher via zfs-discuss wrote:
> On Sun 2018-11-04 (00:13), Gordan Bobic via zfs-discuss wrote:
>> IMO, disks much over 4TB are bad news. I suggest you stick with 4TB disks.
> 
> The server host has not so many disk bays.
> Why are disks bigger than 4 TB bad?
> Because of seek times?
> 
> 
>> Avoid shingled or helium filled disks, or those under 7200rpm.
> 
> Avoiding SMR and <7200rpm is evident, but why are helium filled disks a
> problem?
> 
> 
>> Consult last few years worth of Backblaze statistics when choosing the
>> disk brand and model to get.
> 
> https://www.backblaze.com/blog/hard-drive-stats-for-q2-2018/
> 
> HGST looks like a good choice.
> 
> 
>> I concur with you that RAIDZ2 is a reasonable choice for reliability, but
>> bear in mind that while this will be a reasonable choice for sequential
>> workloads, performance will be very poor for random read workloads (with
>> 7200 rpm disks, you will get 120 IOPS per vdev).
> 
> A F*EX server does mostly sequential reads and writes, there are no
> databases or something like this.
> 
> 
> 


More information about the zfs-discuss mailing list