Hi all,

After reading and watching a lot of content on ZFS and the new stuff in 7.1, I'm still confused on what the best path forward is.

I currently have an Array with 2 parity drives that I'm in the process of growing (I'm merging an old NAS into my Unraid server). The array houses all my media (photos, music, videos), which will account for about 19 TB when all NAS migration has been done.

I then have 1 SATA SSD for Frigate Camera footage, 1 NVME for dockers and another NVME for VMs. Again, here I will get another 2 NVMEs coming from the old NAS.

From what I gathered, a typical Unraid array is still the way to go after 7.1, but I need to do something with my cache pool for sure. Can someone guide me on what would be the best way to structure my drives?

There's nothing wrong with using the unraid array and parity. It comes down to what you want it to do and how you want to interact with it.

Do you want to move to ZFS?
-May require some data wiping...

Some Best Practices for a ZFS only Setup:

View ZFS Forum Info 

ZFS Unraid Docs

with 5 as a pool device only no array, no unraid parity.
I would recommend a raidz2 with 1 vdev.
(can Lose 2 disk with no data loss)

this would make a 5x 20 TB disk volume

Here’s a quick, realistic estimate for a single RAIDZ2 vdev of 5 × 20 TB drives:

• Raw (sum of disks): ~100 TB (decimal) ≈ 90.95 TiB of Total disk space

• RAIDZ2 usable (N−2 disks): ~60 TB ≈ 54.57 TiB realistic space usable in zfs (striping, partiy, snapshots....)

• After ZFS overhead/slop (~3%): ~58.2 TB ≈ 52.9 TiB allocatable

• Practical “keep ≤80% full” target: ~48 TB ≈ 42–43 TiB

This would be your Deep storage so shares would be setup to move things like camera data off into the zfs pool.

Edited September 2, 2025Sep 2 by bmartino1
Data

1. Long-Term Storage (Media, ~19 TB)

• Best option: Create a ZFS RAIDZ2 vdev with your 5 × 20 TB HDDs.

  • Usable: ~52–54 TiB after parity + ZFS overhead.

  • Great balance of redundancy (tolerates 2 disk failures) and usable space.

  • Data integrity benefits from ZFS (checksums, self-healing).

• You’ll want to create ZFS datasets for each type of data (e.g., media/photos, media/music, media/videos) so you can apply compression, quotas, or snapshots independently.

• Configure your Unraid shares to point at the ZFS pool for long-term storage, while still leveraging the Unraid array for parity-protected “drop space” if needed during migration.

---

2. High-Speed / Active Workloads

Keep your SSD/NVMe devices split by function:

• Cache_docker (1 TB NVMe): Dedicated for docker appdata. Use ZFS with recordsize=16K or 32K, compression enabled.

• Cache_vm (1 TB NVMe): Dedicated for VMs. Use ZFS with recordsize=16K, sync=always (if you want VM crash consistency, pair with a UPS).

• Cameras (2 TB SATA SSD): Dedicated ZFS pool for Frigate/NVR footage. Video files are sequential writes, so set recordsize=1M. Consider disabling compression here.

When your 2 extra NVMEs arrive:

• Mirror them (RAID1) to get redundancy + speed, and either:

  • Expand Cache_vm to a mirrored ZFS pool, or

  • Use them as a separate high-performance pool (e.g., fast_storage) for scratch space, databases, or active projects.

---

3. Share Strategy

• Set up Unraid shares so that:

  • Active writes land on the SSD pools (docker/VMs/cameras).

  • Media shares bypass cache and go directly to the ZFS RAIDZ2 pool.

• Use Mover or scripts only if you want to stage large incoming transfers on SSDs before migrating to HDDs. Otherwise, point media shares directly at the RAIDZ2 pool.

---

4. Migration Plan

1. Finish merging old NAS → Unraid.

2. Create your ZFS RAIDZ2 pool for the 20 TB disks.

3. Move media shares from XFS array → ZFS pool.

4. Retire/repurpose the Unraid array if desired (once you trust the ZFS pool).

Great info! Thanks for this!

I have enough backups of everything, so wiping isn't a problem at all. What are the advantages of the ZFS pool instead of the traditional array?

The ZFS pool for cache makes total sense, but:

Do I add 1TB to each cache in mirror:

• 2 x 1TB ZFS in mirror for docker

• 2 x 1TB ZFS in mirror for VMs

Or do I merge everything in 1 big 4tb pool in let's say zraid1 or 2?

4 minutes ago, Jenfil said:

Great info! Thanks for this!

I have enough backups of everything, so wiping isn't a problem at all. What are the advantages of the ZFS pool instead of the traditional array?

The ZFS pool for cache makes total sense, but:

Do I add 1TB to each cache in mirror:

• 2 x 1TB ZFS in mirror for docker

• 2 x 1TB ZFS in mirror for VMs

Or do I merge everything in 1 big 4tb pool in let's say zraid1 or 2?

good questions.

you need a min of 3 disks for a raid z1 and 5 for a raid z2. this is due to how many disk can fail and still ahve the storage and capacity for data.

I'm going to use AI to help answer this with some summary.

Q3. So what’s the recommended structure?

• HDDs (20 TB × 5): ZFS RAIDZ2 → main media pool (~52–54 TiB).

• Cache – Docker: 2 × 1 TB NVMe mirror.

• Cache – VM: 2 × 1 TB NVMe mirror.

• Cache – Cameras: 2 TB SATA SSD (standalone).

This gives you ZFS integrity + performance while still isolating workloads.
Later, if you want more space/flexibility, you can migrate the two NVMe mirrors into a single striped mirror pool.

The typical unraid user will not generally benefit from moving away from the traditional unraid array to a zpool setup. The only downside to not doing a zpool is losing out on the data integrity and verification features. Media playback does not benefit from speed of zfs data striping. Media storage does not benefit from the compression on zfs disks. For anything that is wrote once and read many times will not benefit from being stored on a zpool, with the exception I listed previously.

As for your cache pools using mirrored pools is the smart move due to cache not having it's own parity system in place in the event of a failed disk.

I personally would keep the unraid array and just combining your two 1TB SSDs into a single zfs mirror otherwise you get zero benefits from having single devices zfs zpools as you currently have.

Im in the camp of "if you don't know why you need it, you probably don't need it."

Another downside of going zfs is unraid does not present any GUI accessible options for disk recovery for zfs zpools. Everything you will need to fix/repair degraded zpools is done through command line. Are you prepared to deal with this in the event of a disk failure?

The largest downsides are all your disks must work in tandem which means no individual spin down of disks and also if you lose more disks than parity (raidz level) you lose the ENTIRE zpool worth of data.

Edited September 2, 2025Sep 2 by MowMdown

Alright, looks like I know what to do now :) Thanks bmartino and AI

18 hours ago, MowMdown said:

Another downside of going zfs is unraid does not present any GUI accessible options for disk recovery for zfs zpools. Everything you will need to fix/repair degraded zpools is done through command line. Are you prepared to deal with this in the event of a disk failure?

The largest downsides are all your disks must work in tandem which means no individual spin down of disks and also if you lose more disks than parity (raidz level) you lose the ENTIRE zpool worth of data.

Oh wow, this is extremely valuable, thanks!!

18 hours ago, MowMdown said:

Another downside of going zfs is unraid does not present any GUI accessible options for disk recovery for zfs zpools. Everything you will need to fix/repair degraded zpools is done through command line. Are you prepared to deal with this in the event of a disk failure?

Just to be clear, you can use the GUI to replace/upgrade a zfs pool device; you may need to use the CLI if the pool is unmountable for some reason, since zfs doesn't have an fsck, there are some options that can be tried, but only via CLI.