mirror of
https://github.com/samber/awesome-prometheus-alerts.git
synced 2026-06-25 02:46:59 +08:00
70ac7d9cae
* smartctl_exporter publishes both drive_trip and current drive temperatures. Since most of the alerts are going to be permanent, it does not make sense to wait for the alert to be on for a certain time. Temperature sensors likewise vary, using the last sample is not sufficient to alert on potential issues. * Add an option to run GitHub Action manually * Add an option to force running the action for testing purposes * Set variables correctly * Set variables correctly * Publish * Clean up some more metrics * Publish * Minor bug fixes * Publish * Removed queries that throw errors when systems are upgraded. Also fixed and simplified a few Postgres queries. * Publish * Refined some more queries * Publish * PostgreSQL now has optimized autovacuum behavior * Publish * PostgreSQL now has optimized autovacuum behavior * Publish * Publish * Query fails if instance names are not unique across jobs. This fixes it. * Publish * Ruby is out of date --------- Co-authored-by: samber <samber@users.noreply.github.com>
77 lines
3.8 KiB
YAML
77 lines
3.8 KiB
YAML
groups:
|
|
|
|
- name: SmartctlExporter
|
|
|
|
rules:
|
|
|
|
- alert: SmartDeviceTemperatureWarning
|
|
expr: '(avg_over_time(smartctl_device_temperature{temperature_type="current"} [5m]) unless on (instance, device) smartctl_device_temperature{temperature_type="drive_trip"}) > 60'
|
|
for: 0m
|
|
labels:
|
|
severity: warning
|
|
annotations:
|
|
summary: SMART device temperature warning (instance {{ $labels.instance }})
|
|
description: "Device temperature warning on {{ $labels.instance }} drive {{ $labels.device }} over 60°C\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|
|
|
|
- alert: SmartDeviceTemperatureCritical
|
|
expr: '(max_over_time(smartctl_device_temperature{temperature_type="current"} [5m]) unless on (instance, device) smartctl_device_temperature{temperature_type="drive_trip"}) > 70'
|
|
for: 0m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: SMART device temperature critical (instance {{ $labels.instance }})
|
|
description: "Device temperature critical on {{ $labels.instance }} drive {{ $labels.device }} over 70°C\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|
|
|
|
- alert: SmartDeviceTemperatureOverTripValue
|
|
expr: 'max_over_time(smartctl_device_temperature{temperature_type="current"} [10m]) >= on(device, instance) smartctl_device_temperature{temperature_type="drive_trip"}'
|
|
for: 0m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: SMART device temperature over trip value (instance {{ $labels.instance }})
|
|
description: "Device temperature over trip value on {{ $labels.instance }} drive {{ $labels.device }})\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|
|
|
|
- alert: SmartDeviceTemperatureNearingTripValue
|
|
expr: 'max_over_time(smartctl_device_temperature{temperature_type="current"} [10m]) >= on(device, instance) (smartctl_device_temperature{temperature_type="drive_trip"} * .80)'
|
|
for: 0m
|
|
labels:
|
|
severity: warning
|
|
annotations:
|
|
summary: SMART device temperature nearing trip value (instance {{ $labels.instance }})
|
|
description: "Device temperature at 80% of trip value on {{ $labels.instance }} drive {{ $labels.device }})\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|
|
|
|
- alert: SmartStatus
|
|
expr: 'smartctl_device_smart_status != 1'
|
|
for: 0m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: SMART status (instance {{ $labels.instance }})
|
|
description: "Device has a SMART status failure on {{ $labels.instance }} drive {{ $labels.device }})\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|
|
|
|
- alert: SmartCriticalWarning
|
|
expr: 'smartctl_device_critical_warning > 0'
|
|
for: 0m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: SMART critical warning (instance {{ $labels.instance }})
|
|
description: "Disk controller has critical warning on {{ $labels.instance }} drive {{ $labels.device }})\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|
|
|
|
- alert: SmartMediaErrors
|
|
expr: 'smartctl_device_media_errors > 0'
|
|
for: 0m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: SMART media errors (instance {{ $labels.instance }})
|
|
description: "Disk controller detected media errors on {{ $labels.instance }} drive {{ $labels.device }})\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|
|
|
|
- alert: SmartWearoutIndicator
|
|
expr: 'smartctl_device_available_spare < smartctl_device_available_spare_threshold'
|
|
for: 0m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: SMART Wearout Indicator (instance {{ $labels.instance }})
|
|
description: "Device is wearing out on {{ $labels.instance }} drive {{ $labels.device }})\n VALUE = {{ $value }}\n LABELS = {{ $labels }}"
|