Monitors drive I/O and forces HDD spindown after a given idle period. Resistant to S.M.A.R.T. reads.
Disk spindown has always been an issue for various TrueNAS / FreeNAS users. This
script utilizes iostat to detect I/O operations (reads, writes) on each disk.
If a disk was neither read nor written for a given period of time, it is
considered idle and is spun down.
Periodic reads of S.M.A.R.T. data performed by the smartctl service are excluded. This allows users to have S.M.A.R.T. reporting enabled while being able to automatically spin down disks. The script moreover is immune to the periodic disk temperature reads in newer versions of TrueNAS.
- Periodic S.M.A.R.T. reads do not interfere with spindown
- Support for ATA and SCSI devices
- Works with both TrueNAS Core and TrueNAS SCALE
- Can operate on single disks or whole ZFS pools
- Per-disk idle timer / Independent spindown
- Configurable idle timeout and poll interval
- Different idle timeouts for different disks / ZFS pools
- Automatic detection or explicit listing of drives / ZFS pools to monitor
- Ignoring of specific drives / ZFS pools (e.g. SSD with system dataset)
- Executable via
TasksasPost-Init Script, configurable via TrueNAS GUI - Allows script placement on encrypted pool
- Optional automatic shutdown after configurable idle time
Usage:
spindown_timer.sh [-h] [-q] [-v] [-l] [-d] [-o] [-c] [-m] [-u <MODE>] [-t <TIMEOUT>] [-p <POLL_TIME>] [-i <DRIVE>] [-s <TIMEOUT>]
Monitors drive I/O and forces HDD spindown after a given idle period.
Resistant to S.M.A.R.T. reads.
Operation is supported on either drive level (MODE = disk) with plain device
identifiers or zpool level (MODE = zpool) with zfs pool names. See -u for more
information. A drive is considered idle and gets spun down if there has been no
I/O operations on it for at least TIMEOUT seconds. I/O requests are detected
within multiple intervals with a length of POLL_TIME seconds. Detected reads or
writes reset the drives timer back to TIMEOUT.
Options:
-t TIMEOUT : Total spindown delay. Number of seconds a drive has to
experience no I/O activity before it is spun down (default: 3600).
-p POLL_TIME : I/O poll interval. Number of seconds to wait for I/O during a
single monitoring period (default: 600).
-s TIMEOUT : Shutdown timeout. If given and no drive is active for TIMEOUT
seconds, the system will be shut down.
-u MODE : Operation mode (default: disk).
If set to 'disk', the script operates with disk identifiers
(e.g. ada0) for all CLI arguments and monitors I/O using
iostat directly.
If set to 'zpool' the script operates with ZFS pool names
(e.g. zfsdata) for all CLI arguments and monitors I/O using
the iostat of zpool.
-i DRIVE : In automatic drive detection mode (default):
Ignores the given drive or zfs pool.
In manual mode [-m]:
Only monitor the specified drives or zfs pools. Multiple
drives or zfs pools can be given by repeating the -i option.
-m : Manual drive detection mode. If set, automatic drive detection
is disabled.
CAUTION: This inverts the -i option, which can then be used to
manually supply drives or zfs pools to monitor. All other drives
or zfs pools will be ignored.
-o : One shot mode. If set, the script performs exactly one I/O poll
interval, then immediately spins down drives that were idle for
the last <POLL_TIME> seconds, and exits. This option ignores
<TIMEOUT>. It can be useful, if you want to invoke to script
via cron.
-c : Check mode. Outputs drive power state after each POLL_TIME
seconds.
-q : Quiet mode. Outputs are suppressed set.
-v : Verbose mode. Prints additional information during execution.
-l : Syslog logging. If set, all output is logged to syslog instead
of stdout/stderr.
-d : Dry run. No actual spindown is performed.
-h : Print this help message.
Example usage:
spindown_timer.sh
spindown_timer.sh -q -t 3600 -p 600 -i ada0 -i ada1
spindown_timer.sh -q -m -i ada6 -i ada7 -i da0
spindown_timer.sh -u zpool -i freenas-boot
The following steps describe how to configure TrueNAS and deploy the script.
To prevent the smartctl daemon or TrueNAS from interfering with spun down disks,
open the TrueNAS GUI and navigate to Storage > Disks.
For every disk you want to spin down, click the Edit button. Set the HDD Standby option to Always On and Advanced Power Management to level 128 or
above.
Note: In older versions of FreeNAS it was required to set the S.M.A.R.T Power Mode to Standby. This setting was configured globally and was located under
Services > S.M.A.R.T. > Configure.
Having the TrueNAS system dataset placed on a drive / ZFS pool prevents spindown. The system dataset should therefore be located on a disk that will not be spun down (e.g. the operating system SSD).
The location of the system dataset can be configured under System > System Dataset.
Copy the script to your NAS box and set the execution permission through chmod +x spindown_timer.sh.
That's it! The script can now be run, e.g., in a tmux session. However, an
automatic start during boot is highly recommended (see next section).
The spindown timer support two operation modes, as selected by the -u <MODE>
CLI argument (default: disk):
disk: Operates with plain disk identifiers (e.g.ada0,sda, ...). Each disk is monitored and spun down independently. Disks can even be spun down if they are not part of any ZFS pool (e.g. for hot standby). The CLI option-iwill expect disk names as found in/dev.zpool: Operates on a ZFS pool level. Disks are grouped by their associated ZFS pool. Disks are only spun down if all disks inside a ZFS pool (i.e. the ZFS pool as a whole) received no I/O for a given amount of time. The CLI option-iwill expect pool names (e.g.zfsdata,tank, ...) as output byzpool list.
You can also use a combination of both modes when running multiple instances of the script. See Using separate timeouts for different drives for more details.
There are multiple ways to start the spindown timer after system boot. The
easiest one is to register it as an Init Script via the TrueNAS GUI. This can
be done by opening the GUI and navigating to Tasks > Init/Shutdown Scripts.
Here, create a new Post Init task that executes spindown_timer.sh after
boot.
Note: Be sure to select Command as Type
Note: With FreeNAS-11.3 a Timeout was introduced. However, the spindown
script is never terminated by FreeNAS, regardless of the configured value.
Therefore, keep Timeout at the default value of 10 seconds for now.
If you placed the script at a location that is not available right after boot, a delayed start of the spindown timer is required. This, for example, applies when the script is located inside an encrypted pool, which needs to be unlocked prior to execution.
To automatically delay the start until the script file becomes available, the
helper script delayed_start.sh is provided. It takes the full path to the
spindown timer script as the first argument. All additional arguments are passed
to the called script once available. Example usage: ./delayed_start.sh /mnt/pool/spindown_timer.sh -t 3600 -p 600
The delayed_start.sh script, however, must again be placed in a location that
is available right after boot. To circumvent this problem, you can also use the
following one-liner directly within an Init/Shutdown Script, as shown in the
screenshot below. Set SCRIPT to the path where the spindown_timer.sh file is
stored and configure all desired call arguments by setting them via the ARGS
variable. The CHECK variable determines the delay between execution attempts
in seconds.
/bin/bash -c 'SCRIPT="/mnt/pool/spindown_timer.sh"; ARGS="-t 3600 -p 600"; CHECK=60; while true; do if [ -f "${SCRIPT}" ]; then ${SCRIPT} ${ARGS}; break; else sleep ${CHECK}; fi; done'
Note: Be sure to select Command as Type
Note: With FreeNAS-11.3 a Timeout was introduced. However, the spindown
script is never terminated by FreeNAS, regardless of the configured value.
Therefore, keep Timeout at the default value of 10 seconds for now.
You can verify execution of the script either via a process manager like htop
or simply by using the following command: ps -aux | grep "spindown_timer.sh"
When using a delayed start, keep in mind that it can take up to $CHECK seconds
before the script availability is updated and the spindown timer is finally
executed.
It can be useful to check the current power state of a drive. This can be done by using one of the following commands, depending on your device type.
The script features a check mode. If the CLI flag -c is supplied, the power
state of all monitored drives is output every POLL_TIME seconds, as set via
the -p option (default: 600 seconds). To monitor drive power states without
performing actual spindowns, the dry run flag -d can be set.
The following example checks the power state of all drives every 60 seconds and does perform no spindowns:
./spindown_timer.sh -d -c -p 60The current power mode of an ATA drive can be checked using the command
camcontrol epc $drive -c status -P, where $drive is the drive to check
(e.g., ada0).
It should return Current power state: Standby_z(0x00) for a spun down drive.
The current power mode of a SCSI drive can be checked through reading the
modepage 0x1a using the command camcontrol modepage $drive -m 0x1a, where
$drive is the drive to check (e.g., da0).
A spun down drive should be in one of the standby states Standby_y or
Standby_z.
A detailed description of the available SCSI modes can be found in
/usr/share/misc/scsi_modes.
In the following section, advanced usage scenarios are described.
In automatic mode (default) all drives of the system, excluding the ones
specified using the -i switch, are monitored and spun down if idle.
In scenarios where only a small subset of all available drives should be spun
down, the manual mode can be used by setting the -m flag. This disables
automatic drive detection and inverts the -i switch. It can then be used to
specify all drives that should explicitly get monitored and spun down when idle.
An example in which only the drives ada3 and ada6 are monitored looks like
this:
./spindown_timer.sh -m -i ada3 -i ada6To verify the drive selection, a list of all drives that are being monitored by the running script instance is printed directly after starting the script (except in quiet mode [-q]).
It is possible to run multiple instances of the spindown timer script with
independent TIMEOUT values for different drives simultaneously. Just make sure
that the sets of monitored drives are distinct, i.e., each drive is managed by
only one instance of the spindown timer script.
In the following example, all drives except ada0 and ada1 are spun down
after being idle for 3600 seconds. The drives ada0 and ada1 are instead
already spun down after 600 seconds of being idle:
./spindown_timer.sh -t 3600 -i ada0 -i ada1 # Automatic drive detection
./spindown_timer.sh -m -t 600 -i ada0 -i ada1 # Manual modeAnother example is to operate on ZFS pool basis by default but spin down a set of additional disks that are not part of any ZFS pool yet (e.g., for hot standby):
./spindown_timer.sh -u zpool -i freenas-boot -i ssd # Spindown all ZFS pools except 'freenas-boot' and 'ssd'
./spindown_timer.sh -u disk -m -i ada8 -i ada9 # Additionally spin down disk drives 'ada8' and 'ada9' that are not part of any ZFS pool yetTo start all required spindown timer instances you can simply create multiple
Post Init Scripts, as described above in the Section Automatic start at
boot.
In one shot mode, the script performs exactly one single I/O poll interval, then
immediately spins down drives that were idle for the last POLL_TIME seconds,
and exits. No continuous monitoring is performed in this mode.
This option ignores any specified TIMEOUT value. It can be useful, if you want
to invoke the script via cron to spin down drives at only specific times of the day.
Notice: Make sure to have only one instance of the spindown script running
for one set of disks at the same time. Otherwise, the script instances could
interfere with each other and cause unexpected spin ups/downs. In other words:
Make sure that your cron job triggers not faster than your POLL_TIME value.
When a timeout is given via the -s argument, the system will be shut down by
the script if all monitored drives were idle for the specified number of
seconds. This feature can be used to automatically shut down a system that might
be woken via wake-on-LAN (WOL) later on.
Setting TIMEOUT to 0 results in no shutdown.
This script was successfully tested on the following OS versions:
TrueNAS-13.0-U5.1 (Core)TrueNAS-13.0-U4 (Core)TrueNAS-13.0-U3.1 (Core)TrueNAS-12.0-U8 (Core)TrueNAS-12.0-U7 (Core)TrueNAS-12.0-U6.1 (Core)TrueNAS-12.0-U6 (Core)TrueNAS-12.0-U5.1 (Core)TrueNAS-12.0-U5 (Core)TrueNAS-12.0-U3.1 (Core)TrueNAS-12.0-U1.1 (Core)TrueNAS-12.0-U1 (Core)TrueNAS-12.0 (Core)FreeNAS-11.3-U5FreeNAS-11.3-U4.1FreeNAS-11.3-U3.2FreeNAS-11.3-U3.1FreeNAS-11.3FreeNAS-11.2-U7FreeNAS-11.2-U4.1
TrueNAS SCALE 22.12.0TrueNAS SCALE 22.02.3
Intermediate OS versions not listed here have not been explicitly tested, but the script will most likely be compatible.
Heavily spinning disk drives up and down increases disk wear. Before deploying
this script, consider carefully which of your drives are frequently accessed and
should therefore not be aggressively spun down. A good rule of thumb is to keep
disk spin-ups below 5 per 24 hours. You can keep an eye on your drives
Load_Cycle_Count and Start_Stop_Count S.M.A.R.T values to monitor the number
of performed spin-ups.
Please do not spin down your drives in an enterprise environment. Only consider using this technique with small NAS setups for home use, which idle most time of the day, and select a timeout value appropriate to your usage behavior.
Another typical use case is spinning down drives that are only used once a day (e.g., for mirroring of files or backups).
Bug report and contributions are very welcome! Feel free to open a new issue or submit a merge request :)
The script is heavily inspired by: https://serverfault.com/a/969252
My work helped you in some way, or you just like it? Awesome!
If you want to support me, you can consider buying me a coffee/tea/mate. Thank You! <3
