connect-dashboard.yaml

{{- if and .Values.connect.enabled .Values.monitoring.enabled -}}
apiVersion: v1
kind: ConfigMap
metadata:
  name: connect-dashboard
  labels:
    grafana-dashboard: "true"
    app: {{ template "name" . }}
    release: {{ .Release.Name }}
    heritage: {{ .Release.Service }}
    chart: {{ template "chart" . }}
{{- with .Values.annotations }}
  annotations:
    {{ toYaml . }}
{{- end }}
data:
{{ (.Files.Glob "dashboards/strimzi-connect.json").AsConfig | indent 2 }}
{{- end }}

connect-service-monitor.yaml

{{- if and .Values.monitoring.enabled .Values.connect.enabled -}}
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: {{ template "name" . }}-kafka-connect
  labels:
    app: {{ template "name" . }}-kafka-connect
    chart: {{ template "chart" . }}
    release: {{ $.Release.Name }}
    heritage: {{ $.Release.Service }}
    wek8s-monitoring: "true"
    team: {{ .Values.monitoring.team }}
spec:
  jobLabel: {{ template "name" . }}
  selector:
    matchLabels:
      strimzi.io/cluster: {{ template "name" . }}
      strimzi.io/kind: KafkaConnect
      strimzi.io/name: {{ template "name" . }}-connect-api
  targetLabels:
    - app
    - env
    - strimzi.io/cluster
    - strimzi.io/kind
    - strimzi.io/name
  endpoints:
    - port: metrics
      interval: 5s
{{- end }}

kafka-dashboard.yaml

{{- if and .Values.kafka.install .Values.monitoring.enabled -}}
apiVersion: v1
kind: ConfigMap
metadata:
  name: kafka-dashboard
  labels:
    grafana-dashboard: "true"
    app: {{ template "name" . }}
    release: {{ .Release.Name }}
    heritage: {{ .Release.Service }}
    chart: {{ template "chart" . }}
{{- with .Values.annotations }}
  annotations:
    {{ toYaml . }}
{{- end }}
data:
{{ (.Files.Glob "dashboards/strimzi-kafka.json").AsConfig | indent 2 }}
{{ (.Files.Glob "dashboards/strimzi-zookeeper.json").AsConfig | indent 2 }}
{{- end }}

kafka-prometheus-rules.yaml

{{- if and .Values.monitoring.enabled .Values.kafka.install -}}
apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: {{ template "name" . }}
  labels:
    wek8s-monitoring: "true"
    app: {{ template "name" . }}
    release: {{ .Release.Name }}
    heritage: {{ .Release.Service }}
    chart: {{ template "chart" . }}
{{- with .Values.annotations }}
  annotations:
    {{ toYaml . }}
{{- end }}
spec:
  groups:
    # from https://github.com/strimzi/strimzi-kafka-operator/blob/master/metrics/examples/prometheus/alerting-rules.yaml
    - name: kafka-alerts
      rules:
        - alert: KafkaUnderReplicatedPartitions
          expr: kafka_server_replicamanager_underreplicatedpartitions > 0
          for: 10s
          labels:
            severity: warning
            namespace: {{ .Release.Namespace }}
            alert_source: {{ .Release.Name }}
          annotations:
            summary: 'Kafka under replicated partitions'
            description: 'There are {{ "{{" }} $value {{ "}}" }} under replicated partitions on {{ "{{" }} $labels.kubernetes_pod_name {{ "}}" }}'
        - alert: KafkaAbnormalControllerState
          expr: sum(kafka_controller_kafkacontroller_activecontrollercount) != 1
          for: 10s
          labels:
            severity: warning
            namespace: {{ .Release.Namespace }}
            alert_source: {{ .Release.Name }}
          annotations:
            summary: 'Kafka abnormal controller state'
            description: 'There are  {{ "{{" }} $value {{ "}}" }}  active controllers in the cluster'
        - alert: KafkaUnderMinIsrPartitionCount
          expr: kafka_server_replicamanager_underminisrpartitioncount > 0
          for: 10s
          labels:
            severity: warning
            namespace: {{ .Release.Namespace }}
            alert_source: {{ .Release.Name }}
          annotations:
            summary: 'Kafka under min ISR partitions'
            description: 'There are {{ "{{" }} $value {{ "}}" }} partitions under the min ISR on {{ "{{" }} $labels.kubernetes_pod_name {{ "}}" }}'
        - alert: KafkaOfflineLogDirectoryCount
          expr: kafka_log_logmanager_offlinelogdirectorycount > 0
          for: 10s
          labels:
            severity: warning
            namespace: {{ .Release.Namespace }}
            alert_source: {{ .Release.Name }}
          annotations:
            summary: 'Kafka offline log directories'
            description: 'There are {{ "{{" }} $value {{ "}}" }} offline log directoris on {{ "{{" }} $labels.kubernetes_pod_name {{ "}}" }}'
    - name: zookeeper-alerts
      rules:
        - alert: ZookeeperAvgRequestLatency
          expr: zookeeper_avgrequestlatency > 10
          for: 10s
          labels:
            severity: warning
            namespace: {{ .Release.Namespace }}
            alert_source: {{ .Release.Name }}
          annotations:
            summary: 'Zookeeper average request latency'
            description: 'The average request latency is {{ "{{" }} $value {{ "}}" }} on {{ "{{" }} $labels.kubernetes_pod_name {{ "}}" }}'
        - alert: ZookeeperOutstandingRequests
          expr: zookeeper_outstandingrequests > 10
          for: 10s
          labels:
            severity: warning
            namespace: {{ .Release.Namespace }}
            alert_source: {{ .Release.Name }}            
          annotations:
            summary: 'Zookeeper outstanding requests'
            description: 'There are {{ "{{" }} $value {{ "}}" }} outstanding requests on {{ "{{" }} $labels.kubernetes_pod_name {{ "}}" }}'
{{- end }}

kafka-service-monitor.yaml

{{- if and .Values.monitoring.enabled .Values.kafka.install -}}
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: {{ template "name" . }}-kafka
  labels:
    app: {{ template "name" . }}-kafka
    chart: {{ template "chart" . }}
    release: {{ $.Release.Name }}
    heritage: {{ $.Release.Service }}
    wek8s-monitoring: "true"
    team: {{ .Values.monitoring.team }}
spec:
  jobLabel: {{ template "name" . }}
  selector:
    matchLabels:
      strimzi.io/cluster: {{ template "name" . }}
      strimzi.io/kind: Kafka
      strimzi.io/name: {{ template "name" . }}-kafka-bootstrap
  scrapeTimeout: 15
  targetLabels:
    - app
    - strimzi.io/cluster
    - strimzi.io/kind
    - strimzi.io/name
    # podTargetLabels:
  endpoints:
    - port: metrics
      interval: 10s
    # relabelings:
    #   - action: labelmap
    #     regex: __meta_kubernetes_pod_label_(.+)
    #     separator: ;
    #     replacement: $1
---
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: {{ template "name" . }}-zookeeper
  labels:
    app: {{ template "name" . }}-zookeeper
    chart: {{ template "chart" . }}
    release: {{ $.Release.Name }}
    heritage: {{ $.Release.Service }}
    wek8s-monitoring: "true"
    team: {{ .Values.monitoring.team }}
spec:
  jobLabel: {{ template "name" . }}
  selector:
    matchLabels:
      strimzi.io/cluster: {{ template "name" . }}
      strimzi.io/kind: Kafka
      strimzi.io/name: {{ template "name" . }}-zookeeper-client
  targetLabels:
    - app
    - strimzi.io/cluster
    - strimzi.io/kind
    - strimzi.io/name
  endpoints:
  - port: metrics
    interval: 5s
{{- end }}