ansible/roles/prometheus/templates/alert.rules.yml.j2

---
{{ ansible_managed | comment }}

{% macro raw(string) -%}
{{ "{{" }} {{ string }} {{ "}}" }}
{%- endmacro %}

groups:
  - name: alert.rules
    rules:

      - alert: InstanceDown
        expr: up == 0
        for: 3m
        labels:
          severity: critical
        annotations:
          summary: >-
            Invisible depuis plus de 3 minutes

      - alert: OutOfMemory
        expr: >-
          (
            node_memory_MemFree_bytes
            + node_memory_Cached_bytes
            + node_memory_Buffers_bytes
          ) / node_memory_MemTotal_bytes * 100 < 10
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: >-
            {{ raw('$value | printf "%.1f"') }}% de mémoire
            libre

      - alert: OutOfDiskSpace
        expr: >-
          node_filesystem_free_bytes / node_filesystem_size_bytes * 100 < 10
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: >-
            {{ raw('$value | printf "%.1f"') }}% d'espace libre pour
            {{ raw('$labels.mountpoint') }}

      - alert: OutOfInodes
        expr: node_filesystem_files_free / node_filesystem_files * 100 < 10
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: >-
            {{ raw('$value | printf "%.1f"') }}% d'inodes
            restants pour {{ raw('$labels.mountpoint') }}

      - alert: CpuUsage
        expr: >-
          (
            100 - avg by (instance) (
              irate(node_cpu_seconds_total{mode="idle"}[5m])
            ) * 100
          ) > 75
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: >-
            CPU à {{ raw('$value | printf "%.1f"') }}%

      - alert: SystemdServiceFailed
        expr: node_systemd_unit_state{state="failed"} == 1
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: >-
            {{ raw('$labels.name') }} a échoué

      - alert: LoadUsage
        expr: node_load1 > 5
        for: 2m
        labels:
          severity: warning
        annotations:
          summary: >
            Charge à {{ raw('$value') }}
      
      - alert: UpsOutputSourceChanged
        expr: upsOutputSource != 3
        for: 1m
        labels:
          severity: critical
        annotations:
          summary: >-
            Source d'alimentation changée

      - alert: UpsBatteryStatus
        expr: upsBatteryStatus == 3
        for: 2m
        labels:
          severity: warning
        annotations:
          summary: >-
            État de la batterie faible

      - alert: UpsBatteryStatus
        expr: upsBatteryStatus == 4
        for: 10m
        labels:
          severity: critical
        annotations:
          summary: >-
            État de la batterie critique

      - alert: UpsHighLoad
        expr: upsOutputPercentLoad > 70
        for: 5m
        labels:
          severity: critical
        annotations:
          summary: >-
            Charge de {{ raw('$value | printf "%.1f"') }}%

      - alert: UpsWrongInputVoltage
        expr: (upsInputVoltage < 210) or (upsInputVoltage > 250)
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: >-
            Tension d'entrée de {{ raw('$value') }}V

      - alert: UpsWrongOutputVoltage
        expr: >-
          abs(upsInputVoltage - avg_over_time(upsOutputVoltage[1d]))
            < 3 * stddev_over_time(upsOutputVoltage[1d])
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: >-
            Tension de sortie de {{ raw('$value') }}V

      - alert: UpsTimeRemaining
        expr: upsEstimatedMinutesRemaining < 8
        for: 1m
        labels:
          severity: warning
        annotations:
          summary: >-
            Autonomie restante de {{ raw('$value') }} min

      - alert: UpsTimeRemaining
        expr: upsEstimatedMinutesRemaining < 5
        for: 1m
        labels:
          severity: critical
        annotations:
          summary: >-
            Autonomie restante de {{ raw('$value') }} min
...