infrastructure/clusters/cl01tl/manifests/kube-prometheus-stack/PrometheusRule-kube-prometheus-stack-prometheus-operator.yaml

---
# Source: kube-prometheus-stack/charts/kube-prometheus-stack/templates/prometheus/rules-1.14/prometheus-operator.yaml
apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: kube-prometheus-stack-prometheus-operator
  namespace: kube-prometheus-stack
  labels:
    app: kube-prometheus-stack
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/instance: kube-prometheus-stack
    app.kubernetes.io/version: "79.11.0"
    app.kubernetes.io/part-of: kube-prometheus-stack
    chart: kube-prometheus-stack-79.11.0
    release: "kube-prometheus-stack"
    heritage: "Helm"
spec:
  groups:
    - name: prometheus-operator
      rules:
        - alert: PrometheusOperatorListErrors
          annotations:
            description: Errors while performing List operations in controller {{$labels.controller}} in {{$labels.namespace}} namespace.
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatorlisterrors
            summary: Errors while performing list operations in controller.
          expr: (sum by (cluster,controller,namespace) (rate(prometheus_operator_list_operations_failed_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[10m])) / sum by (cluster,controller,namespace) (rate(prometheus_operator_list_operations_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[10m]))) > 0.4
          for: 15m
          labels:
            severity: warning
        - alert: PrometheusOperatorWatchErrors
          annotations:
            description: Errors while performing watch operations in controller {{$labels.controller}} in {{$labels.namespace}} namespace.
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatorwatcherrors
            summary: Errors while performing watch operations in controller.
          expr: (sum by (cluster,controller,namespace) (rate(prometheus_operator_watch_operations_failed_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m])) / sum by (cluster,controller,namespace) (rate(prometheus_operator_watch_operations_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]))) > 0.4
          for: 15m
          labels:
            severity: warning
        - alert: PrometheusOperatorSyncFailed
          annotations:
            description: Controller {{ $labels.controller }} in {{ $labels.namespace }} namespace fails to reconcile {{ $value }} objects.
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatorsyncfailed
            summary: Last controller reconciliation failed
          expr: min_over_time(prometheus_operator_syncs{status="failed",job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]) > 0
          for: 10m
          labels:
            severity: warning
        - alert: PrometheusOperatorReconcileErrors
          annotations:
            description: '{{ $value | humanizePercentage }} of reconciling operations failed for {{ $labels.controller }} controller in {{ $labels.namespace }} namespace.'
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatorreconcileerrors
            summary: Errors while reconciling objects.
          expr: (sum by (cluster,controller,namespace) (rate(prometheus_operator_reconcile_errors_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]))) / (sum by (cluster,controller,namespace) (rate(prometheus_operator_reconcile_operations_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]))) > 0.1
          for: 10m
          labels:
            severity: warning
        - alert: PrometheusOperatorStatusUpdateErrors
          annotations:
            description: '{{ $value | humanizePercentage }} of status update operations failed for {{ $labels.controller }} controller in {{ $labels.namespace }} namespace.'
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatorstatusupdateerrors
            summary: Errors while updating objects status.
          expr: (sum by (cluster,controller,namespace) (rate(prometheus_operator_status_update_errors_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]))) / (sum by (cluster,controller,namespace) (rate(prometheus_operator_status_update_operations_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]))) > 0.1
          for: 10m
          labels:
            severity: warning
        - alert: PrometheusOperatorNodeLookupErrors
          annotations:
            description: Errors while reconciling Prometheus in {{ $labels.namespace }} Namespace.
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatornodelookuperrors
            summary: Errors while reconciling Prometheus.
          expr: rate(prometheus_operator_node_address_lookup_errors_total{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]) > 0.1
          for: 10m
          labels:
            severity: warning
        - alert: PrometheusOperatorNotReady
          annotations:
            description: Prometheus operator in {{ $labels.namespace }} namespace isn't ready to reconcile {{ $labels.controller }} resources.
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatornotready
            summary: Prometheus operator not ready
          expr: min by (cluster,controller,namespace) (max_over_time(prometheus_operator_ready{job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]) == 0)
          for: 5m
          labels:
            severity: warning
        - alert: PrometheusOperatorRejectedResources
          annotations:
            description: Prometheus operator in {{ $labels.namespace }} namespace rejected {{ printf "%0.0f" $value }} {{ $labels.controller }}/{{ $labels.resource }} resources.
            runbook_url: https://runbooks.prometheus-operator.dev/runbooks/prometheus-operator/prometheusoperatorrejectedresources
            summary: Resources rejected by Prometheus operator
          expr: min_over_time(prometheus_operator_managed_resources{state="rejected",job="kube-prometheus-stack-operator",namespace="kube-prometheus-stack"}[5m]) > 0
          for: 5m
          labels:
            severity: warning