hd-idle

Reimplementation of Christian Mueller's hd-idle with some extra features.

hd-idle is a utility program for spinning-down external disks after a period of idle time. Since most external IDE disk enclosures don't support setting the IDE idle timer, a program like hd-idle is required to spin down idle disks automatically.

Important note: hd-idle is not compatible with the usage of disk monitoring tools like smartmontools.

Index

• Extra features
  • Support ATA commands
  • Monitor the skew between monitoring cycles
  • Resolve symlinks in runtime
  • Log disk spin up
  • Use disk partitions or device mapper to calculate activity
• Install
  • Precompiled binaries
  • Build from source
• Run hd-idle
• Configuration
• Understand the logs
  • Standard log
  • Log file
• Warning on spinning down disks
• Troubleshot
  • Disks won't spin down
  • LUKS support
  • SCSI response not ok

Extra features

List of extra features compared to the original hd-idle:

Support ATA commands

The implementation of hd-idle written by Christian Mueller relies on the SCSI api to work. When listing the drives by id, disks starting with usb stop using the original implementation, but disk starting with ata do not.

$ ls /dev/disk/by-id/

ata-WDC_WD40EZRX-
ata-WDC_WD50EZRX-
usb-WD_My_Book_1140_

hdparm on the other hand always stops the drives without any problems. It uses ATA api calls to send disks to standby. hd-idle comes with ATA commands support to replicate hdparm's api calls.

Monitor the skew between monitoring cycles

Identify if the sleep took longer than expected and reset the spun down flag if it waited too long for the main loop sleep. This should capture suspend events as well as excessive machine load.

Resolve symlinks in runtime

hd-idle can resolve disk symlinks also in runtime. Disks added after application's start won't be hidden.

Log disk spin up

Show in standard output when disks spin up.

Use disk partitions or device mapper to calculate activity

The disk activity is calculated by watching read/write changes on partition or device mapper level instead of disk level. This is required for kernels newer than 5.4 LTS, because disk monitoring tools change read/write values on disk level, although there's no real activity on the disk itself. When using LUKS, activity will happen on the device mapper device mapped to the corresponding disk.

Install

There are various ways of installing hd-idle:

Precompiled binaries

Precompiled binaries for released versions are available in the releases section.

Build from source

To build hd-idle from the source code yourself you need to have a working Go environment with version 1.16 or greater installed.

Open a terminal and execute these commands:

git clone https://github.com/adelolmo/hd-idle
cd hd-idle
make

On Debian you can also build the package yourself using dpkg-buildpackage:

git clone https://github.com/adelolmo/hd-idle.git
cd hd-idle
dpkg-buildpackage -a armhf -us -uc -b

In the example above, the package is built for armhf, but you can build it also for the platforms i386, amd64, and arm64 by substituting the parameter -a.

Then install the package:

# dpkg -i ../hd-idle*.deb

Run hd-idle

In order to run hd-idle, type:

$ hd-idle

This will start hd-idle with the default options, causing all SCSI (read: USB, Firewire, SCSI, ...) hard disks to spin down after 10 minutes of inactivity.

If the Debian package was installed, after editing /etc/default/hd-idle and enabling it (START_HD_IDLE=true), run hd-idle with:

# systemctl start hd-idle

To enable hd-idle on reboot:

# systemctl enable hd-idle    

Please note that hd-idle uses /proc/diskstats to read disk statistics. If this file is not present, hd-idle won't work.

In case of problems, use the debug option -d to get further information.

Configuration

Command line options:

• -a name
  Set device name of disks for subsequent idle-time parameters -i. This parameter is optional in the sense that there's a default entry for all disks which are not named otherwise by using this parameter. This can also be a symlink (e.g. /dev/disk/by-uuid/...)

• -i idle_time
  Idle time in seconds for the currently named disk(s) (-a name) or for all disks. Setting this value to 0 will never spin down the disk(s).

• -c command_type
  Api call to stop the device. Possible values are scsi (default value) and ata.

• -p power_condition
  Power condition to send with the issued SCSI START STOP UNIT command. Possible values are 0-15 (inclusive). The default value of 0 works fine for disks accessible via the SCSI layer (USB, IEEE1394, ...), but it will NOT work as intended with real SCSI / SAS disks. A stopped SAS disk will not start up automatically on access, but requires a startup command for reactivation. Useful values for SAS disks are 2 for idle and 3 for standby.

• -s symlink_policy
  Set the policy to resolve symlinks for devices. If set to 0, symlinks are resolved only on start. If set to 1, symlinks are also resolved on runtime until success. By default symlinks are only resolved on start. If the symlink doesn't resolve to a device, the default configuration will be applied.

• -l logfile
  Name of logfile (written only after a disk has spun up or down). Please note that this option might cause the disk which holds the logfile to spin up just because another disk had some activity. On single-disk systems, this option should not cause any additional spinups. On systems with more than one disk, the disk where the log is written will be spun up. On raspberry based systems the log should be written to the SD card.

Miscellaneous options:

• -t disk
  Spin-down the specified disk immediately and exit.

• -d
  Debug mode. It will print debugging info to stdout/stderr (/var/log/syslog if started with systemctl)

• -h
  Print usage information.

Regarding the parameter -a:

The parameter -a can be used to set a filter on the disk's device name (omit /dev/) for subsequent idle-time settings.

1. A -i option before the first -a option will set the default idle time.

2. In order to disable spin-down of disks per default, and then re-enable spin-down on selected disks, set the default idle time to 0.

   Example:

   hd-idle -i 0 -a sda -i 300 -a sdb -i 1200
   

   This example sets the default idle time to 0 (meaning hd-idle will never try to spin down a disk) and the default api command to scsi, then sets explicit idle times for disks which have the string sda or sdb in their device name.

3. The option -c allows to set the api call that sends the spindown command. Possible values are scsi (the default value) or ata.

   Example:

   hd-idle -i 0 -c ata -a sda -i 300 -a sdb -i 1200 -c scsi
   

   This example sets the default idle time to 0 (meaning hd-idle will never try to spin down a disk) and the default api command to ata, then sets explicit idle times for disks which have the string sda or sdb in their device name and sets sdb to use scsi api command.

Understand the logs

By default hd-idle only logs to the standard output. You can find them in the syslog if the application starts via service.

If you set the log file (-l flag) then the application writes extra details to it. (Check the Configuration section).

Standard log

The standard log output registers two kinds of events:

• disk spin up
• disk spin down

Aug  8 00:14:55 enterprise hd-idle[9958]: sda spindown
Aug  8 00:14:55 enterprise hd-idle[9958]: sdb spindown
Aug  8 00:14:56 enterprise hd-idle[9958]: sdc spindown
Aug  8 00:17:55 enterprise hd-idle[9958]: sdb spinup
Aug  8 00:28:55 enterprise hd-idle[9958]: sdb spindown

Log file

You can enable the log file with the flag -l followed by the log path. (Check the Configuration section).

This is the kind of entry shown in the log file:

date: 2020-07-30, time: 05:28:01, disk: sdc, running: 601, stopped: 76654

Explanation:

• date and time when the disk spins up.
• disk involved.
• running seconds the device was running before it spun down the last time.
• stopped seconds since last spin down. This is the time the disk was asleep before spinning up.

Important Note:

The log file is written after a full cycle of running-stopped-wakeup.

A bit more on running explained with the above example:

timestamp	disk spin	event	new disk spin	running	stopped
2020-07-29 07:59:57	down	disk activity	up	?	?
2020-07-29 08:09:58	up	go to sleep	down	-	-
2020-07-30 05:28:01	down	disk activity	up	08:09:58 - 07:59:57 = 601s	2020-07-30 05:28:01 - 2020-07-29 08:09:58 = ~21h (76654s)

Explanation:

At 07:59:57 the disk is on standby and hd-idle detects disk activity.

At 08:09:58 the disk is active and hd-idle determines inactivity of the disk and spins it down.

At 05:28:01 on the next day the disk is on standby and hd-idle detects disk activity. It writes on the log file 601s of previous disk spin up and ~21h of standby.

Warning on spinning down disks

A word of caution: hard disks don't like spinning up too often. Laptop disks are more robust in this respect than desktop disks but if you set your disks to spin down after a few seconds you may damage the disk over time due to the stress the spin-up causes on the spindle motor and bearings. It seems that manufacturers recommend a minimum idle time of 3-5 minutes, the default in hd-idle is 10 minutes.

You have been warned...

Troubleshot

This section covers some usual issues that users face while using hd-idle.

Disks won't spin down

Unfortunately, it's not possible to get hd-idle working alongside disk monitoring tools like smartmontools. You have to disable those tools in order to get hd-idle working.

LUKS support

Using encrypted disk or partitions with LUKS is supported by the use of symlinks.

1. Run the following command with you're disk mounted: sudo lsblk /dev/sd* -o PATH,FSSIZE,LABEL,UUID,PARTLABEL,PARTUUID,MODEL,SIZE,SERIAL,TYPE,WWN

PATH                            FSSIZE LABEL UUID                                 PARTLABEL PARTUUID                             MODEL   SIZE SERIAL TYPE  WWN
/dev/sde                                                                                                                         ST400   3.7T ZGY0LB disk  0x5000c500a3d1d419
/dev/sde1                                    100e952e-0ffb-4b73-bb1a-8401d4fe56c0 dropbox   14a81aa8-c2c9-448e-967b-85d87dc9b488           1T        part  0x5000c500a3d1d419
/dev/sde1                                    100e952e-0ffb-4b73-bb1a-8401d4fe56c0 dropbox   14a81aa8-c2c9-448e-967b-85d87dc9b488           1T        part  0x5000c500a3d1d419
/dev/sde2                         2.6T three 175e2227-d24f-4ad0-9e42-2ddb8846682d           d2792423-3c07-44fe-ab6b-a1aca61c73a5         2.7T        part  0x5000c500a3d1d419
/dev/sde2                         2.6T three 175e2227-d24f-4ad0-9e42-2ddb8846682d           d2792423-3c07-44fe-ab6b-a1aca61c73a5         2.7T        part  0x5000c500a3d1d419
/dev/mapper/luks-100e952e-0ffb-4b73-bb1a-8401d4fe56c0
                               1007.8G dropbox
                                             649dd15e-6750-472c-8185-4d76bffc2490                                                       1024G        crypt 

You have to take symlinks that resolve to disk devices: /dev/sd*.

In the example above /dev/mapper/luks-100e952e-0ffb-4b73-bb1a-8401d4fe56c0 is the Path to the encrypted partition, which WWN is 0x5000c500a3d1d419.

2. Run the following command to see which devices the system has identified using by-id: sudo ls -lv /dev/disk/by-id/

Output:

lrwxrwxrwx 1 root root  9 Jul 18 15:56 ata-ST4000DM005-2DP166_ZGY0LBRB -> ../../sde
lrwxrwxrwx 1 root root 10 Jul 18 16:01 ata-ST4000DM005-2DP166_ZGY0LBRB-part1 -> ../../sde1
lrwxrwxrwx 1 root root 10 Jul 18 15:56 ata-ST4000DM005-2DP166_ZGY0LBRB-part2 -> ../../sde2
lrwxrwxrwx 1 root root  9 Jul 18 15:56 wwn-0x5000c500a3d1d419 -> ../../sde
lrwxrwxrwx 1 root root 10 Jul 18 16:01 wwn-0x5000c500a3d1d419-part1 -> ../../sde1
lrwxrwxrwx 1 root root 10 Jul 18 15:56 wwn-0x5000c500a3d1d419-part2 -> ../../sde2

Here we see that we can either use ata-ST4000DM005-2DP166_ZGY0LBRB or wwn-0x5000c500a3d1d419 as symlinks.

3. Edit /etc/default/hd-idle to use the symlink you prefer. In my case, I went with the symlink using WWN (unique storage identifier), yet I could have chosen MODEL (device identifier) instead.

HD_IDLE_OPTS='-i 0 -c ata -s 1 -l /var/log/hd-idle.log -a /dev/disk/by-id/wwn-0x5000c500a3d1d419 -i 600' Or HD_IDLE_OPTS='-i 0 -c ata -s 1 -l /var/log/hd-idle.log -a /dev/disk/by-id/ata-ST4000DM005-2DP166_ZGY0LBRB -i 600'

SCSI response not ok

You can find information about the issue here: SCSI-response-not-ok

License

GNU General Public License v3.0, see LICENSE.