Merge pull request #259 from Yolean/metrics-kafka-minion

Kafka-Minion as alternative to Burrow for consumer lag monitoring

Author: solsson

README.md

@@ -59,6 +59,7 @@ One option is to keep kubernets-kafka as a git submodule and edit the relative p
 Have a look at:
  * [./prometheus](./prometheus/)
  * [./linkedin-burrow](./linkedin-burrow/)
+ * [./consumers-prometheus](./consumers-prometheus/)
  * [or plain JMX](https://github.com/Yolean/kubernetes-kafka/pull/96)
  * what's happening in the [monitoring](https://github.com/Yolean/kubernetes-kafka/labels/monitoring) label.
  * Note that this repo is intentionally light on [automation](https://github.com/Yolean/kubernetes-kafka/labels/automation). We think every SRE team must build the operational knowledge first.

consumers-prometheus/kafka-minion-service.yaml

@@ -0,0 +1,13 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: metrics-minion
+  namespace: kafka
+  labels: &labels
+    app: kafka-minion
+    type: openmetrics
+spec:
+  selector: *labels
+  ports:
+  - name: http
+    port: 8080

consumers-prometheus/kafka-minion.yaml

@@ -0,0 +1,49 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: metrics-minion
+  namespace: kafka
+  labels: &labels
+    app: kafka-minion
+    type: openmetrics
+spec:
+  replicas: 1
+  selector:
+    matchLabels: *labels
+  template:
+    metadata:
+      labels: *labels
+      annotations:
+        prometheus.io/scrape: "true"
+        prometheus.io/port: "8080"
+        prometheus.io/path: /metrics
+    spec:
+      containers:
+      - name: kafka-minion
+        image: solsson/kafka-consumers-prometheus@sha256:a005aa02581fa46884b4a4499f7b4d56e889770c64e7e84d0a9ad195935fa59c
+        env:
+        - name: TELEMETRY_HOST
+          value: 0.0.0.0
+        - name: TELEMETRY_PORT
+          value: "8080"
+        - name: EXPORTER_IGNORE_SYSTEM_TOPICS
+          value: "true"
+        - name: EXPORTER_METRICS_PREFIX
+          value: kafka_minion
+        - name: LOG_LEVEL
+          value: info
+        - name: KAFKA_BROKERS
+          value: kafka-0.broker:9092, kafka-1.broker:9092, kafka-2.broker:9092
+        - name: KAFKA_CONSUMER_OFFSETS_TOPIC_NAME
+          value: __consumer_offsets
+        ports:
+        - name: http
+          containerPort: 8080
+        readinessProbe:
+          httpGet:
+            port: http
+            path: /readycheck
+        livenessProbe:
+          httpGet:
+            port: http
+            path: /healthcheck

consumers-prometheus/kustomization.yaml

@@ -0,0 +1,3 @@
+resources:
+- kafka-minion-service.yaml
+- kafka-minion.yaml

prometheus/50-kafka-jmx-exporter-patch.yml

@@ -1,4 +1,4 @@
-# meant to be applied using
+# meant to be applied using kustomize, or with pre-1.14 kubectl:
 # kubectl --namespace kafka patch statefulset kafka --patch "$(cat prometheus/50-kafka-jmx-exporter-patch.yml )"
 apiVersion: apps/v1
 kind: StatefulSet

prometheus/README.md

@@ -1,10 +1,27 @@
 # Export metrics to Prometheus
 
-Kafka uses JMX to expose metrics, as is already [enabled](https://github.com/Yolean/kubernetes-kafka/pull/96) for broker pods. There's many ways to use JMX. For example [Kafka Manager](../yahoo-kafka-manager/) uses it to display current broker traffic.
+JMX is already [enabled](https://github.com/Yolean/kubernetes-kafka/pull/96) for broker pods (TODO extract to kustomization). There's many ways to use JMX. For example [Kafka Manager](../yahoo-kafka-manager/) uses it to display current broker traffic.
 
-At Yolean we use Prometheus. This folder adds a sidecar to the broker pods that exports selected JMX metrics over HTTP in Prometheus format. To add a container to an existing pod we must use the `patch`command:
+This folder adds a sidecar to the broker pods that exports selected JMX metrics over HTTP in Prometheus format. To add a container to an existing pod we must use the `patch`command:
+
+Using kubectl 1.14+
+
+```
+kubectl --namespace kafka apply -k prometheus/
+```
+
+Using pre-1.14 kubectl:
 
 ```
 kubectl --namespace kafka apply -f prometheus/10-metrics-config.yml 
 kubectl --namespace kafka patch statefulset kafka --patch "$(cat prometheus/50-kafka-jmx-exporter-patch.yml )"
 ```
+
+## Consumer lag monitoring
+
+See [Burrow](../linkedin-burrow)
+or [Kafka Minion](../consumers-prometheus/)
+
+## Prometheus Operator
+
+Use the [prometheus-operator](../variants/prometheus-operator/) kustomization.

prometheus/kustomization.yaml

@@ -0,0 +1,9 @@
+bases:
+# This kustomization needs to depend on one with kafka in it, to add the sidecar,
+# but it needs to be the kafka from the chosen variant, as ../kafka here would override other kustomizations
+#- ../kafka
+#- ../variants/scale-1
+resources:
+- 10-metrics-config.yml
+patchesStrategicMerge:
+- 50-kafka-jmx-exporter-patch.yml

variants/prometheus-operator-example/additional-scrape-configs.yaml

@@ -0,0 +1,12 @@
+
+apiVersion: monitoring.coreos.com/v1
+kind: Prometheus
+metadata:
+  name: k8s
+  namespace: monitoring
+spec:
+  additionalScrapeConfigs:
+    # github.com/kubernetes-sigs/kustomize/blob/master/examples/kvSourceGoPlugin.md is clearly WIP
+    name: additional-scrape-configs-5m4c7m6mc9
+    # See https://github.com/prometheus/prometheus/pull/4131, and upon disagreement see https://github.com/prometheus/prometheus/issues/4484
+    key: pods-discovery-by-prometheus-io-annotations.yaml

variants/prometheus-operator-example/alertmanager-main-scale-1.yaml

@@ -0,0 +1,7 @@
+apiVersion: monitoring.coreos.com/v1
+kind: Alertmanager
+metadata:
+  name: main
+  namespace: monitoring
+spec:
+  replicas: 1

variants/prometheus-operator-example/k8s-kafka-rbac.yaml

@@ -0,0 +1,32 @@
+# Allows the "k8s" prometheus from Prometheus Operator contrib to do service discovery iin the kafka namespace
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: prometheus-k8s
+  namespace: kafka
+rules:
+- apiGroups:
+  - ""
+  resources:
+  - services
+  - endpoints
+  - pods
+  verbs:
+  - get
+  - list
+  - watch
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: prometheus-k8s
+  namespace: kafka
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: prometheus-k8s
+subjects:
+- kind: ServiceAccount
+  name: prometheus-k8s
+  namespace: monitoring

variants/prometheus-operator-example/k8s-kafka-servicemonitor.yaml

@@ -0,0 +1,38 @@
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: broker-monitoring
+  namespace: kafka
+  labels:
+    app: kafka
+spec:
+  publishNotReadyAddresses: true
+  ports:
+  - name: fromjmx
+    port: 5556
+  selector:
+    app: kafka
+---
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: kafka
+  namespace: monitoring
+  labels:
+    k8s-app: kafka
+spec:
+  namespaceSelector:
+    matchNames:
+    - kafka
+  selector:
+    matchLabels:
+      app: kafka
+  endpoints:
+  # https://github.com/coreos/prometheus-operator/blob/master/Documentation/api.md#endpoint
+  - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+    interval: 120s
+    scrapeTimeout: 119s
+    port: fromjmx
+    scheme: http
+    path: /metrics

variants/prometheus-operator-example/k8s-minion-servicemonitor.yaml

@@ -0,0 +1,22 @@
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: kafka-metrics-minion
+  namespace: monitoring
+  labels:
+    k8s-app: kafka-metrics-minion
+spec:
+  namespaceSelector:
+    matchNames:
+    - kafka
+  selector:
+    matchLabels:
+      app: kafka-minion
+      type: openmetrics
+  endpoints:
+  - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+    interval: 30s
+    scrapeTimeout: 30s
+    port: http
+    scheme: http
+    path: /metrics

variants/prometheus-operator-example/kustomization.yaml

@@ -0,0 +1,31 @@
+bases:
+# With prometheus-operator only you must add your own Prometheus and Alertmanager resources
+#- github.com/coreos/prometheus-operator?ref=de9a6e1
+- github.com/coreos/kube-prometheus?ref=3a64636
+- ../../consumers-prometheus
+# The ../../prometheus base must be edited to point to the chosen kafka base
+# Actually to apply the sidecar with apply -k it has to be included with the kafka variant; can't be its own kustomization because you'll get
+# either "failed to find an object with apps_v1_StatefulSet|kafka to apply the patch" or "id 'apps_v1_StatefulSet|kafka|~P|zoo|~S' already used"
+#- ../../prometheus
+resources:
+- k8s-kafka-rbac.yaml
+# with base ../../prometheus
+#- k8s-kafka-servicemonitor.yaml
+# with base ../../consumers-prometheus
+- k8s-minion-servicemonitor.yaml
+patchesStrategicMerge:
+- prometheus-k8s-scale-1.yaml
+- prometheus-k8s-2.9.2.yaml
+- alertmanager-main-scale-1.yaml
+- prometheus-k8s-nodeport.yaml
+- additional-scrape-configs.yaml
+secretGenerator:
+- name: additional-scrape-configs
+  namespace: monitoring
+  #kvSources:
+  #- name: pods-discovery-by-prometheus-io-annotations.yaml
+    #pluginType: builtin
+    #args:
+    #- scrape-configs/pods-discovery-by-prometheus-io-annotations.yaml
+  files:
+  - scrape-configs/pods-discovery-by-prometheus-io-annotations.yaml

variants/prometheus-operator-example/prometheus-k8s-2.9.2.yaml

@@ -0,0 +1,8 @@
+apiVersion: monitoring.coreos.com/v1
+kind: Prometheus
+metadata:
+  name: k8s
+  namespace: monitoring
+spec:
+  baseImage: quay.io/prometheus/prometheus
+  version: v2.7.2

variants/prometheus-operator-example/prometheus-k8s-nodeport.yaml

@@ -0,0 +1,10 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: prometheus-k8s
+  namespace: monitoring
+spec:
+  type: NodePort
+  ports:
+  - port: 9090
+    nodePort: 32490

variants/prometheus-operator-example/prometheus-k8s-scale-1.yaml

@@ -0,0 +1,7 @@
+apiVersion: monitoring.coreos.com/v1
+kind: Prometheus
+metadata:
+  name: k8s
+  namespace: monitoring
+spec:
+  replicas: 1

variants/prometheus-operator-example/scrape-configs/pods-discovery-by-prometheus-io-annotations.yaml

@@ -0,0 +1,35 @@
+# Example scrape config for pods
+#
+# The relabeling allows the actual pod scrape endpoint to be configured via the
+# following annotations:
+#
+# * `prometheus.io/scrape`: Only scrape pods that have a value of `true`
+# * `prometheus.io/path`: If the metrics path is not `/metrics` override this.
+# * `prometheus.io/port`: Scrape the pod on the indicated port instead of the
+# pod's declared ports (default is a port-free target if none are declared).
+- job_name: 'kubernetes-pods'
+
+  kubernetes_sd_configs:
+  - role: pod
+
+  relabel_configs:
+  - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_scrape]
+    action: keep
+    regex: true
+  - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_path]
+    action: replace
+    target_label: __metrics_path__
+    regex: (.+)
+  - source_labels: [__address__, __meta_kubernetes_pod_annotation_prometheus_io_port]
+    action: replace
+    regex: ([^:]+)(?::\d+)?;(\d+)
+    replacement: $1:$2
+    target_label: __address__
+  - action: labelmap
+    regex: __meta_kubernetes_pod_label_(.+)
+  - source_labels: [__meta_kubernetes_namespace]
+    action: replace
+    target_label: kubernetes_namespace
+  - source_labels: [__meta_kubernetes_pod_name]
+    action: replace
+    target_label: kubernetes_pod_name