トーク:S.M.A.R.T.

S.M.A.R.T. (Self-Monitoring, Analysis, and Reporting Technology) is a supplementary component build into many modern storage devices through which devices monitor, store, and analyze the health of their operation. Statistics are collected (temperature, number of reallocated sectors, seek errors...) which software can use to measure the health of a device, predict possible device failure, and provide notifications on unsafe values.

Smartmontools

The smartmontools package contains two utility programs (smartctl and smartd) to analyze and monitor storage devices. Install smartmontools from the official repositories.

Detect if device has SMART support

To check if the device has SMART capability (it may be necessary to add -d ata to specify it is an ATA derived device):

# smartctl -i /dev/<device>

(where <device> is sda, hda,...). This will give general information about the device, the last two lines will show if it is supported:

SMART support is: Available - device has SMART capability.
SMART support is: Enabled

If SMART is not enabled, it can be enabled by doing:

# smartctl -s on /dev/<device>

Test the device health

Three type of health tests that can be performed on the device (all are safe to user data):

1. Short (runs tests that have a high probability of detecting device problems)
2. Extended (or Long; a short check with complete disk surface examination)
3. Conveyance (identifies if damage incurred during transportation of the device)

To view the device's available tests and the time it will take to perform each test do:

# smartctl -c /dev/<device>

To run the tests do:

# smartctl -t short /dev/<device>
# smartctl -t long /dev/<device>
# smartctl -t conveyance /dev/<device>

ノート: Some disks (e.g. SSDs) may not support all types of test. You can see what your device supports with smartctl --capabilities /dev/<device>

Results

To view the test's overall health status (compiled from all tests):

# smartctl -H /dev/<device>

To view the test's result errors:

# smartctl -l selftest /dev/<device>

To view the test's detailed results:

# smartctl -a /dev/<device>

If no errors are reported the device is likely healthy. If there are a few errors this may or may not indicate a problem and should be investigated further. When a device starts to fail it is recommended to backup the data and replace it.

Monitor devices

Devices can be monitored in the background with use of the smartmontools daemon that will check devices periodically and optionally email any potential problems. To have devices monitored on boot, enable smartd service:

 systemctl enable smartd.service

The smart daemon can be edited for more exact configuration in /etc/smartd.conf.

ヒント: /etc/smartd.conf is configured with somewhat esoteric command-line style options. See the comments and examples in the file, and refer to the manpage for a full explanation. What follows are some examples of the monitoring options.

Define the devices to monitor

To monitor all attributes of all disks specify:

 DEVICESCAN

Alternatively, enable monitoring of all attributes on individual disks:

#DEVICESCAN
/dev/<first_device> -a
/dev/<second_device> -a

ヒント: If you want to specify different monitoring options for different disks, you'll need to define them separately rather than use DEVICESCAN.

Email potential problems

To have an email sent when a failure or new error occurs, use the -m option:

DEVICESCAN -m address@domain.com

To be able to send the email externally (i.e. not to the root mail account) a MTA (Mail Transport Agent) or a MUA (Mail User Agent) will need to be installed and configured. Common MTAs are MSMTP and SSMTP. Common MTUs are sendmail and Postfix. It's enough to simply configure S-nail if you don't want anything else.

Once the mail agent is setup the -M test option can be used to test if an email will be sent (restart the daemon immediately to discover):

DEVICESCAN -m address@domain.com -M test

Power management

If you use a computer under control of power management, you should instruct smartd how to handle disks in low power mode. Usually, in response to SMART commands issued by smartd, the disk platters are spun up. So if this option is not used, then a disk which is in a low-power mode may be spun up and put into a higher-power mode when it is periodically polled by smartd.

DEVICESCAN -n standby,15,q

More info on smartmontools wiki.

Schedule self-tests

smartd can tell disks to perform self-tests on a schedule. The following /etc/smartd.conf configuration will start a short self-test every day between 2-3am, and an extended self test weekly on Saturdays between 3-4am:

 DEVICESCAN -s (S/../.././02|L/../../6/03)

Alert on temperature changes

smartd can track disk temperatures and alert if they rise too quickly or hit a high limit. The following will log changes of 4 degrees or more, log when temp reaches 35 degrees, and log/email a warning when temp reaches 40:

 DEVICESCAN -W 4,35,40

ヒント: You can determine the current disk temperature with the command smartctl -A /dev/<device> | grep Temperature_Celsius

ヒント: If you have some disks that run a lot hotter/cooler than others, remove DEVICESCAN and define a separate configuration for each device with appropriate temperature settings.

Complete smartd.conf example

Putting together all of the above gives the following example configuration:

• DEVICESCAN (smartd scans for disks and monitors all it finds)
• -a (monitor all attributes)
• -o on (enable automatic online data collection)
• -S on (enable automatic attribute autosave)
• -n standby,q (don't check if disk is in standby, and supress log message to that effect so as not to cause a write to disk)
• -s ... (schedule short and long self-tests)
• -W ... (monitor temperature)
• -m ... (mail alerts)

 DEVICESCAN -a -o on -S on -n standby,q -s (S/../.././02|L/../../6/03) -W 4,35,40 -m <username or email>

Start/reload the smartd service and check status

 # systemctl start smartd

or

 # systemctl reload smartd

Check status:

 # systemctl status smartd

Full smartd log:

 # journalctl -u smartd

GUI Applications

• Gsmartcontrol — A GNOME frontend for the smartctl hard disk drive health inspection tool

http://gsmartcontrol.berlios.de/home/index.php/en/Home || gsmartcontrol

Resources

• Smartmontools Homepage
• Smartmontools on Ubuntu Wiki