Help with 8 x 18 TB + 2 x 500 GB SSD ZFS Setup

Rudde

Member
Mar 10, 2011
49
0
6
Hi.

I purchased 8 x 18 TB Drives to fill this compact NAS chassis I got. It fits one M.2 drive witch I currently use for a OS. 8x 3,5" drives and 4x 2,5" drives.

I've been out of the ZFS loop for a while, I have some arrays running, but I'm not entirely up to date on best practices. So my initial idea was to run the SSD's as redundant Metadata Special Device, and the rest of the spinning drives in a z-raid1 or 2 pool, but now with dRAID is that better? I don't think it is really all that relevant upgrading before these drives are out of fashion, so I don't think RAID expansion is all that relevant here. Is there any benefit to dRAID if that is not my goal, and I don't have a hot-spare?

I'm going to be running Linux, probably debian or ubuntu, so ZFS-on-Linux
 

dswartz

Active Member
Jul 14, 2011
554
64
28
Generally not worth hot sparing, IMO. What is your use case? If random access perf is important, I'd go with 4x2 raid1. Otherwise raidz2 with 6+2 drives. Raidz1 not generally recommended nowadays... Can't comment on draid without more info.
 

Rudde

Member
Mar 10, 2011
49
0
6
The wast majority of the storage will be medium sized pre-compressed media files. But I do also have a few hundred thousands jpgs in various sizes. I also would like to set up owncloud or something to to my family members where they can access and backup their stuff from computers and phones.

I will also probably be running a few minecraft servers of it. And databases, but not for production, only developmental/experimental/self-hosted purposes.

I was thinking of running Block size at 1M
 

GourmetSaint

New Member
Dec 18, 2020
4
0
1
If space is not a real concern, for simplicity, I would use Raid10 equivalent (4 x striped mirrors) for the HDDs (ie slow pool) and a separate mirror for the SSDs (fast pool). Much faster resilvering when using mirrors and great increase in read and write performance with no parity calcs.

If you need performance (ie VMs) use the fast pool and vice versa. Use the standard block size at pool level and create filesets with varying block sizes depending on the data types for those filesets. eg for media files - use 1M, for databases - use recommended block size matching the database requirements.