Hi - looking for some input on potential approaches to replacing cache drives.

Use case for my Unraid system is primarily a Plex server with supporting apps, as well as Nextcloud, Immich, and a few misc docker containers. Array is 76TB XFS consisting of 5 drives and one parity. I have have not done any work with ZFS to this point.

I currently have 2x 1TB NVME's configured as a single BTRFS pool. Over the last few weeks, I've been getting persistent BTRFS errors. These NVME's are probably 6 years old at this point and I am looking to replace them (also 2x 1TB drives). While I'm at it I thought I'd get some opinions on perhaps changing things up and improving my cache approach. Specifically:

• for my use case would there be any benefit to formatting the new cache drives as ZFS?

• is there any benefit of not pooling the two NVME drives (I've seen posts where dedicating one cache drive to downloads improves performance)

• anything else I should consider.

Thanks all!

43 minutes ago, joshallen said:

for my use case would there be any benefit to formatting the new cache drives as ZFS?

Not really. btrfs has a lot of similar features to zfs like snapshots and compression. I will say there are more GUI tools available to manage zfs snapshots and datasets. These tools are however 3rd party and could be deprecated at any time if the community devices to stop keeping them up to date. (zfs master is kinda dying off right now anyways)

43 minutes ago, joshallen said:

is there any benefit of not pooling the two NVME drives (I've seen posts where dedicating one cache drive to downloads improves performance)

I would argue that you 100% should mirror them unless you keep regular full backups of the entire pool contents. I prefer to minimize my downtime by doing a mirror but that's a personal preference.

44 minutes ago, joshallen said:

anything else I should consider.

Not really. There's not really a right or wrong answer.

1 hour ago, joshallen said:

I've been getting persistent BTRFS errors. These NVME's are probably 6 years old at this point and I am looking to replace them (also 2x 1TB drives)

I would recommend posting the diagnostics; errors may not be device related.

Diagnostics attached. This is what I am seeing in the logs:

Oct 19 05:51:36 ATLANTIS kernel: BTRFS error (device nvme0n1p1): bdev /dev/nvme1n1p1 errs: wr 0, rd 0, flush 0, corrupt 453, gen 0

atlantis-diagnostics-20251028-1112.zip

Have you ran a btrfs scrub? there's FS corruption detected.

Edit: You're also having read errors. Might be more severe than I thought.

Edited October 28, 2025Oct 28 by MowMdown

Yes, I've scrubbed since I started noticing the errors. I've had a few crashes over the past few months too, while the system has been rock solid for years with regular upgrades as they came available.

UUID:             d0205282-dbbd-45ca-99c4-63bcc61a504b
Scrub started:    Tue Oct 28 11:58:28 2025
Status:           finished
Duration:         0:04:05
Total to scrub:   683.73GiB
Rate:             2.79GiB/s
Error summary:    csum=2
  Corrected:      0
  Uncorrectable:  2
  Unverified:     0

Btrfs is detecting data corruption on both devices, this is typically bad RAM, sugest running memtest.

Thanks. I'll run memtest and report back

2 minutes ago, joshallen said:

Thanks. I'll run memtest and report back

You may want to run it for an extended period of time, it might take 10-15-20 passes to detect errors.