Letting disks sleep

What keeps hard disks awake on an OMV box with this plugin, and how to find the culprit. Marked [measured] where seen on the reference TS-253D (OMV 8, ZFS mirror of two SATA disks, 2026-10-17/18); the rest is general Linux / OMV behaviour that has not been measured here.

How a disk falls asleep

OMV sets the spindown time (Storage › Disks › edit: Spindown time, Advanced Power Management) and applies it at boot with the smartctl-hdparm service (smartctl --set=standby,N; OMV 8 does not use hdparm). The drive then sleeps after that long without any command. Not only reads and writes count: any S.M.A.R.T. query to an awake drive, even smartctl -n standby, restarts its standby timer [measured]. So a disk only sleeps if nothing touches it for the whole spindown time.

What the plugin does

qnapd never wakes a sleeping disk and, since 8.0.1, does not keep an idle one awake [measured]:

• temperatures and health with smartctl -n standby (skips sleeping disks), and only for disks that did I/O since the last check, or every 30 minutes;
• activity from /sys/block/*/stat, pool health from zpool status / /proc/mdstat (kernel state, no disk access);
• never the drivetemp hwmon driver.

With qnapd running, the reference disks reach standby 12–14 minutes after their last wake with a 10-minute spindown [measured] (drives add a little to the nominal time).

Keeps disks awake: change or remove

Cause	Why	What to do
OMV S.M.A.R.T. Power mode: Never	smartd checks every disk on each check interval (OMV default 30 min), waking sleeping ones [measured]	Storage › S.M.A.R.T. › Settings: Power mode = Standby (skips sleeping disks)
drivetemp kernel module	every read of its hwmon temperature sends a command to the drive; whatever reads it (OMV’s Sensors widget, sensors, collectd’s sensors plugin, Netdata, fan scripts) keeps the disk awake or wakes it	Don’t load it: delete its line from /etc/modules-load.d/*.conf. The plugin already shows disk temperatures without it
Other temperature or fan tools (tempmon scripts, hddtemp, hddfancontrol, other fan plugins)	usually smartctl without -n standby, which wakes the disk [measured: a tempmon setup did]	Remove them; qnapd does fan control from disk temperatures itself
An older openmediavault-qnap (< 8.0.1)	queried idle disks every minute [measured]	Update
Scheduled S.M.A.R.T. self-tests, scrubs, snapshots, rsync / backup jobs	wake the disks by design	Schedule them together, at a time the disks may wake

Often keeps disks awake: check if you use it

Not measured on the reference box; common causes on any Linux NAS:

• Apps writing to the data pool: Docker containers (databases, logs, downloaders), media servers scanning libraries (Plex, Jellyfin), Syncthing, Nextcloud. Put their config and data on an SSD (the boot disk or an NVMe pool) and keep only bulk media on the HDDs.
• ZFS atime=on: reading a file writes its access time. Use zfs set atime=off <pool> (or relatime=on).
• Clients browsing shares: macOS Spotlight and Finder, Windows thumbnailing, DLNA. Spotlight can be turned off per SMB share (spotlight = no, the OMV default).
• Swap or logs on a data disk. OMV keeps both on the system disk by default.

Finding what wakes a disk

1. Note the disk’s I/O counters, wait, compare. Reading them does not touch the disk:

   awk '{print $1 + $5}' /sys/block/sda/stat
   

   A rising number means something read or wrote. Flat but still awake means a command such as smartctl (step 3).

2. File access by process: fatrace -f W (Debian package fatrace) or iotop -oa.
3. Commands: look for timers and services that run smartctl (grep -rl smartctl /etc/cron* /etc/systemd), and check the S.M.A.R.T. power mode above.
4. Check the state once, at the end: smartctl -n standby -i /dev/sda exits with status 2 when the disk is asleep. Every check on an awake disk restarts its timer, so polling it keeps it awake.

OMV’s Diagnostics › Performance statistics › Disk I/O graphs show gaps or flat lines while disks sleep. They read kernel counters and do not wake disks.