Help designing new pool(s)

[espenfjo]

Hi

I am planning on upgrading an old Synology to a TrueNAS installation, and I will be using 5x6TB + 5x4TB HDDs as my primary storage.

The system will be mainly used for archive, as well as Plex streaming and adding of Plex content. I will run a couple of VMs and Dockers/apps as well, but these should all be disk-friendly and not use much IO.

The plan for both the 6TB and 4TB drives are obviously RAIDZ2, but I am not sure if I should split these two vdevs into two pools or one joint pool.
Im also unsure if I should tweak the record size either up or down from the default 128k (probably 1M for the VM and media datasets?), and possible use a raidz2 SSD for my metadata store.

According to Wendell`s magic "find command" my current file sizes are distributed as follows:

Code:

 
1k: 5427993
  2k: 5585454
  4k: 7946447
  8k: 6034212
 16k: 4947933
 32k: 3164960
 64k: 1471199
128k: 531740
256k: 194017
512k:  85942
  1M:  36179
  2M:  27097
  4M:  11684
  8M:   7177
 16M:  24798
 32M:   7952
 64M:  13971
128M:    337
256M:    334
512M:   1259
  1G:   1343
  2G:   1571
  4G:    459
  8G:    224

 

[NugentS]

Well 10*4 = 32TB
5*6 Z2 18, +5*4 Z2 12 = 30TB

For metadata are you talking a special vdev or L2ARC (metadata only)?

record size for the media data should be 1M. VM's I believe smaller

 

[espenfjo]

Would you recommend doing a 10 wide pool instead of two separate with the 4/6 disks?

Thinking a special vdev for the metadata, possibly partitioned with a partition for the small files (special_small_blocks). edit: A separate partition here is perhaps not needed. Metadata and small files will perhaps live on the same special vdev.

 

[Davvo]

Would you recommend doing a 10 wide pool instead of two separate with the 4/6 disks?

Imho no. You would have less redundancy for 2 TB.

 

[Etorix]

Would you recommend doing a 10 wide pool instead of two separate with the 4/6 disks?

It depends whether you plan to replace drives and/or add vdevs to increase capacity.
As it stands, two vdevs with drives of the same size is most appropriate.

Thinking a special vdev for the metadata, possibly partitioned with a partition for the small files (special_small_blocks). edit: A separate partition here is perhaps not needed. Metadata and small files will perhaps live on the same special vdev.

Metadata is part of the pool data: Loose the special vdev, loose the whole pool. A special vdev would have to be at least a three-way mirror to provide enough redundancy.
If the pool is mostly archive, i.e. mostly read and few writes, much of the same benefits can be achieved with a persistant metadata-only L2ARC. No redundancy needed here, just a single SSD, but there must be enough RAM to manage the L2ARC.
If you do buy three SSDs, use the other two for a dedicated pool for VMs and apps.