Kubernetes

Kubernetes

If you are using Kubernetes you have two options to install Pitchfork: with Helm, or by manually creating all of the individual Kubernetes resources.

Helm

To install Pitchfork with Helm you need to add a new repository for Expedia Group and then you can simply run a helm install.

helm repo add expediagroup http://expediagroup.github.io/helm-charts/repo
helm install pitchfork expediagroup/pitchfork

If you need to override the default options you can find the available settings here.

Individual resources

To install the individual Pitchfork resources you can use the following scripts.

Deployment

apiVersion: apps/v1
kind: Deployment
metadata:
  name: pitchfork
  labels:
    app: pitchfork
spec:
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
  replicas: 1
  selector:
    matchLabels:
      app: pitchfork
  template:
    metadata:
      labels:
        app: pitchfork
        release: pitchfork
      annotations:
        prometheus.io/scrape: "true"
        prometheus.io/port: "8081"
        prometheus.io/path: "/actuator/prometheus"
    spec:
      containers:
        # Please replace "latest" with the most recent version available at https://hub.docker.com/r/expediagroup/pitchfork/tags
        - name: pitchfork
          image: expediagroup/pitchfork:latest
          ports:
            - containerPort: 9411
            - containerPort: 8081
          env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - name: SERVER_PORT
              value: "9411"
            # If you are familiar with the JVM you can tune memory and settings here. If not, these should give you an overall decent experience.
            - name: JAVA_JVM_ARGS
              value: "-XX:MaxRAMPercentage=80.0"
            # You can enabled and configure more forwarders here.
            - name: PITCHFORK_FORWARDERS_LOGGING_ENABLED
              value: "true"
            # You can enable and configure more forwarders here.
            - name: PITCHFORK_FORWARDERS_HAYSTACK_KAFKA_ENABLED
              value: "true"
            - name: PITCHFORK_FORWARDERS_HAYSTACK_KAFKA_BOOTSTRAP_SERVERS
              value: "kafka-service:9092"
            # The following properties are use to tag metrics produced by Pitchfork with the app name and with the name of the pod.
            - name: MANAGEMENT_METRICS_TAGS_APP
              value: "pitchfork"
            - name: MANAGEMENT_METRICS_TAGS_INSTANCE
              value: $(POD_NAME)
            - name: MANAGEMENT_METRICS_EXPORT_GRAPHITE_TAGS_AS_PREFIX
              value: "app,instance"
            # Isolating actuator endpoints. This allows Pitchfork to handle healthchecks even when under extremely high load.
            - name: MANAGEMENT_SERVER_PORT
              value: "8081"
          resources:
            requests:
              memory: "1Gi"
              cpu: "1"
            limits:
              memory: "1Gi"
              cpu: "1"
          livenessProbe:
            httpGet:
              path: /info
              port: 8081
            initialDelaySeconds: 10
            timeoutSeconds: 1
            periodSeconds: 30
          readinessProbe:
            httpGet:
              path: /health
              port: 8081
            initialDelaySeconds: 10
            timeoutSeconds: 1
            periodSeconds: 30
          # Artificial sleep to allow connections to drain
          lifecycle:
            preStop:
              exec:
                command: [ "/bin/sleep", "20"]

Service

apiVersion: v1
kind: Service
metadata:
  name: pitchfork
  labels:
    app: pitchfork
spec:
  type: ClusterIP
  ports:
    - name: app-port
      port: 9411
      protocol: TCP
      targetPort: 9411

Horizontal Pod Autoscaler

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: pitchfork
  labels:
    app: pitchfork
spec:
  scaleTargetRef:
    apiVersion: extensions/v2beta1
    kind: Deployment
    name: pitchfork
  minReplicas: 1
  maxReplicas: 3
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 30