ansible/roles/prometheus/templates/ilo.rules.yml.j2
Alexandre Iooss a320907047
All checks were successful
continuous-integration/drone/push Build is passing
continuous-integration/drone/pr Build is passing
prometheus: fix typo
2022-01-01 17:55:32 +01:00

83 lines
2.5 KiB
Django/Jinja

---
{{ ansible_managed | comment }}
{% macro raw(string) -%}
{{ "{{" }} {{ string }} {{ "}}" }}
{%- endmacro %}
groups:
- name: ilo.rules
rules:
- alert: IloResilientMemoryDegraded
expr: cpqHeResilientMemCondition{cpqHeResilientMemCondition!~"ok|other"} == 1
for: 3m
labels:
severity: warning
annotations:
summary: >-
La mémoire vive n'est plus résiliente
({{ raw('$labels.cpqHeResilientMemCondition') }})
- alert: IloBiosSelfTestDegraded
expr: cpqHeHWBiosCondition{cpqHeHWBiosCondition!~"ok|other"} == 1
for: 3m
labels:
severity: critical
annotations:
summary: >-
Une erreur a été détectée lors du POST du serveur
({{ raw('$labels.cpqHeHWBiosCondition') }})
- alert: IloBatteryDegraded
expr: cpqHeSysBatteryCondition{cpqHeSysBatteryCondition!~"ok|other"} == 1
for: 3m
labels:
severity: warning
annotations:
summary: >-
La batterie est dégradée
({{ raw('$labels.cpqHeSysBatteryCondition') }})
- alert: IloTemperatureSensorDegraded
expr: cpqHeTemperatureCondition{cpqHeTemperatureCondition!~"ok|other"} == 1
for: 3m
labels:
severity: critical
annotations:
summary: >-
Le capteur de température est dégradé
({{ raw('$labels.cpqHeTemperatureCondition') }})
- alert: IloFanDegraded
expr: cpqHeFltTolFanCondition{cpqHeFltTolFanCondition!~"ok|other"} == 1
for: 3m
labels:
severity: critical
annotations:
summary: >-
Le ventilateur est dégradé
({{ raw('$labels.cpqHeFltTolFanCondition') }})
- alert: IloPowerSupplyDegraded
expr: cpqHeFltTolPowerSupplyStatus{cpqHeFltTolPowerSupplyStatus!="noError"} == 1
for: 3m
labels:
severity: critical
annotations:
summary: >-
L'alimentation est dégradée
({{ raw('$labels.cpqHeFltTolPowerSupplyStatus') }})
- alert: IloOverrideSwitchState
expr: cpqSm2CntlriLOSecurityOverrideSwitchState{cpqSm2CntlriLOSecurityOverrideSwitchState="set"} == 1
for: 3m
labels:
severity: critical
annotations:
summary: >-
Le switch de réinitialisation n'est pas à l'état d'origine,
l'authentification est bypassée
...