Spinning Down Idle Disks on a Linux Home Server

⚠ Created by AI 🤖

A home server that idles 22 hours a day doesn’t need its spinning disks turning the whole time. Getting them to actually sleep takes two things: telling the disk when to stop, and finding whatever keeps waking it up. The second part is where most of the work is.

Test it by hand first

Before configuring anything persistent, confirm the drive will spin down at all. hdparm is the standard tool:

sudo hdparm -y /dev/sdX

That spins the disk down immediately. It’s temporary — the next access, or a reboot, undoes it. Check the result:

sudo hdparm -C /dev/sdX

You want to see standby. If it says active/idle a second after you issued -y, something is already hammering the disk and no amount of configuration will help until you find it. Skip ahead to what prevents spin-down.

Make it automatic

Install hdparm if it isn’t already there:

sudo apt install hdparm        # Debian/Ubuntu
sudo dnf install hdparm        # Fedora/RHEL
sudo pacman -S hdparm          # Arch

Then set an idle timeout in /etc/hdparm.conf:

/dev/sdX {
    spindown_time = 240
}

The units are the awkward part. spindown_time is not seconds or minutes:

So:

• 10 minutes → 120
• 20 minutes → 240
• 1 hour → 252

Apply it without rebooting:

sudo systemctl restart hdparm

Or via udev

Some distros prefer a udev rule, which also survives the disk being re-plugged. Create /etc/udev/rules.d/69-hdparm.rules:

ACTION=="add|change", KERNEL=="sdX", RUN+="/sbin/hdparm -S 240 /dev/sdX"

Reload:

sudo udevadm control --reload-rules
sudo udevadm trigger

Then verify the same way as before:

sudo hdparm -C /dev/sdX

active/idle means spinning, standby means you’ve won.

What prevents spin-down

This is the real problem. A disk with a perfectly good 20-minute timeout will never sleep if anything touches it every 19 minutes. The usual suspects:

• smartd — SMART polling, by default wakes the disk to read attributes
• Media servers — Plex and Jellyfin library scans
• File indexing — Tracker, updatedb/mlocate
• Docker containers — logs and volumes on the data disk
• ZFS — writes metadata frequently and effectively never idles
• Btrfs — background metadata writes

smartd is the most common one, and it’s also the easiest to fix. Add -n standby to the drive’s line in /etc/smartd.conf:

/dev/sdX -a -n standby

That tells smartd to skip the check entirely if the disk is already asleep, rather than waking it to ask how it’s doing.

sudo systemctl restart smartd

Filesystem matters more than you’d expect

ZFS does not spin down reliably. If power saving is the priority, don’t put your archive pool on ZFS — you’re choosing between the two.

mdadm RAID works, with a caveat: any access to one member wakes all of them, and array metadata writes can block sleep outright.

ext4 and XFS on single disks give the most reliable behaviour. If a drive exists only to hold cold data, this is the combination that actually sleeps.

Timeouts worth using

• Home NAS — 20–30 minutes
• Backup-only NAS — 10–15 minutes
• Media server — often not achievable; scanning wins

Avoid anything under 5 minutes. Spin-up cycles cause more wear than the idle spinning does, and a disk that parks and unparks every few minutes will die sooner than one you left alone.