ansible/roles/prometheus/templates/ilo.rules.yml.j2
Alexandre Iooss 40d9108b37
All checks were successful
continuous-integration/drone/push Build is passing
continuous-integration/drone/pr Build is passing
prometheus: add iLO alert rules
2021-12-31 21:26:27 +01:00

58 lines
1.4 KiB
Django/Jinja

---
{{ ansible_managed | comment }}
{% macro raw(string) -%}
{{ "{{" }} {{ string }} {{ "}}" }}
{%- endmacro %}
groups:
- name: ilo.rules
rules:
- alert: IloBiosSelfTestDegraded
expr: cpqHeHWBiosCondition >= 3
for: 3m
labels:
severity: critical
annotations:
summary: >-
Une erreur a été détectée lors du POST du serveur
- alert: IloTemperatureSensorDegraded
expr: cpqHeTemperatureCondition >= 3
for: 3m
labels:
severity: critical
annotations:
summary: >-
Le capteur de température est dégradé
- alert: IloFanDegraded
expr: cpqHeFltTolFanCondition >= 3
for: 3m
labels:
severity: critical
annotations:
summary: >-
Le ventilateur est dégradé
- alert: IloPowerSupplyDegraded
expr: cpqHeFltTolPowerSupplyStatus >= 2
for: 3m
labels:
severity: critical
annotations:
summary: >-
L'alimentation est dégradée
- alert: IloOverrideSwitchState
expr: cpqSm2CntlriLOSecurityOverrideSwitchState == 2
for: 3m
labels:
severity: critical
annotations:
summary: >-
Le switch de réinitialisation n'est pas à l'état d'origine
...