OpenZFS on OS X

Hello,

It seems that the power saving states on Apple Silicon are more aggressive than Intel macs, putting my HDDs to sleep, which ZFS does not appreciate.
I'm running a OWC Thunderbay 8 thunderbolt 3 drive bay (8 drives). HDDs are WD HC580 24TB running latest firmware.

I've tried:
1) setting "never" to "put drives to sleep when possible" in the settings gui.
2) sudo pmset -a disksleep 0
3) caffeinate -ims as a LaunchDaemon

In all of the above Apple Silicon will spin the drives down after a couple minutes.
Where as Intel mac will keep the drives running no problem.

The only thing I've had luck doing is doing "smartctl -a /dev/diskXX > /dev/null" every 30 secs or so to each drive, but this seems fragile.

Curious if others have encountered this problem and/or have any suggested workarounds, etc.

Cheers,

Calico

can't say i've noticed this, but what are the symptoms you're seeing? just having to wait for them to spin up again, or does ZFS not appreciating it mean something more serious?

Interesting that you're not seeing the same issues.

Basically the harddrives will spin down. Once that happens, I typically get a kernel panic. Although I'm also getting kernel panics at other times as well.
Perhaps it's more about the interactions of the Apple Silicon mac and the OWC Thunderbay 8 running ZFS, etc. (Again Intel macs don't seem to have any problems, either with macos openzfs or running ubuntu and zfs).

I'm curious about your Apple Silicon setup.

Do your HDDs spin down at all or does ZFS just gracefully handle it when they do?
How are you connecting HDDs to your Apple Silicon system?

i have a Mac Studio (M4 Max) with 128 GB of RAM connected to three of the same enclosures as you (OWC Thunderbay 8). i'm running macOS Sequoia 15.5, OpenZFS 2.2.3 (OpenZFSonOsX-2.2.3-Sequoia-15-arm64.pkg). looks like there's a 2.3.0 release (that was never announced on here?) from three weeks ago that i haven't installed. might be helpful to get your various software versions.

i don't really know if my hard drives spin down at all. there's certainly plenty of time when at least one pool is getting no reads or writes, but i do have "Put hard disks to sleep when possible" off in the energy settings.

On versions previous to 2.3.0 I never actually noticed disks spinning down except when the system is put to sleep or a while after the pool is exported (matching the system settings for drive sleep). Since upgrading to 2.3.0 by backup pool now spins down around 10 minutes after a backup completes, which I'm generally happy for it to do (the disks being used are more archival/backup optimised anyway).

Can't comment on any specific difference between Intel and ARM specifically, but I do think something changed in 2.3.0 to allow drives to actually sleep during normal use, where before they simply wouldn't even when a pool was completely idle for some time.