sbraz / check_smart

Nagios-like plugin to check S.M.A.R.T. data

Introduction

This plugin can be used to monitor disks and detect changes in S.M.A.R.T. metrics.

Features

• Checks the status of all smart metrics and whether they are in a failing state
• Checks increments in critical counters such as read errors
• Outputs performance data for all attributes

By default, the script checks all disks present on the system and retains values over the last 4 runs. This means that in case a critical metric sees an increase, the check will return a warning 4 times. After the fifth run, the increment will no longer be detected.

Requirements

The script requires:

• Python 3.7 or newer
• nagiosplugin version 1.2.4 or newer
• smartmontools 7.0 or newer (JSON output support)
• sudo and access to smartctl --json=s -x commands, see the related section
• read-write access to /var/tmp/ (where the state file is created)

Security considerations

In order to limit the attack surface as much as possible, it is recommended to only grant sudo access to the required smartctl commands.

For example, create /etc/sudoders.d/check_smart containing:

icinga ALL=(ALL) NOPASSWD: /usr/sbin/smartctl --json=s -x /dev/sd[a-z]
icinga ALL=(ALL) NOPASSWD: /usr/sbin/smartctl --json=s -x /dev/sd[a-z][a-z]
icinga ALL=(ALL) NOPASSWD: /usr/sbin/smartctl --json=s -x /dev/nvme[0-9]n[0-9]

This should be enough to make the check work with most configurations.

Usage

To check all disks:

./check_smart.py

To list all available disks:

./check_smart.py --list-devices

To limit to two disks:

./check_smart.py -D /dev/sda /dev/sdb

Symlinks are also resolved, so the following trick can be used to make sure we are opening the same disk across reboots:

./check_smart.py -D /dev/disk/by-id/ata-*<device serial>

To exclude a disk:

./check_smart.py -X /dev/sda

It is possible to change the number of check attempts before an increment in a checked counter stops being reported as an error. The following will cause the check to return an error only once. All subsequent runs will be fine.

./check_smart.py --max-attempts 1

Sometimes, it is desirable to exclude certain counters from alerts:

./check_smart.py --exclude-metric Raw_Read_Error_Rate

The list of checked and non-checked metrics for a certain device can be obtained with:

./check_smart.py --checked-metrics -D /dev/sda
./check_smart.py --non-checked-metrics -D /dev/sda

Check the help for a description of all avaliable options:

./check_smart.py -h

Integration with Icinga

An Icinga CheckCommand can be defined with:

object CheckCommand "smart_metrics" {
  command = [PluginDir + "/check_smart.py"]
  arguments = {
    "--devices" = {
      value = "$smart_metrics_devices$"
      repeat_key = false
    }
    "--exclude-devices" = {
      value = "$smart_metrics_exclude_devices$"
      repeat_key = false
    }
    "--skip-removable" = {
      set_if = "$smart_metrics_skip_removable$"
    }
    "--max-attempts" = "$max_check_attempts$"
    "--exclude-metrics" = {
      value = "$smart_metrics_exclude_metrics$"
      repeat_key = false
    }
    "--ignore-failing-commands" = {
      set_if = "$smart_metrics_ignore_failing_commands$"
    }
  }
  vars.smart_metrics_skip_removable = true
}