kubernetes.yaml

# First, we setup a deployment, which is responsible for deploying pod sfrom
# a template.  The deployment is also responsible for doing application
# rolling-updates.  A deployment will automatically do a rolling update if
# its image URL is updated.
apiVersion: apps/v1beta1
kind: Deployment
metadata:
  annotations:
  labels:
    app: car-server
  name: car-server
  namespace: batman
spec:
  # We can set replicas here statically, or you can check out HPAs (horizontal
  # pod autoscaler) to automatically scale your application out with load.
  strategy:
    rollingUpdate:
      maxSurge: 1 # we can make one surplus pod while rolling out updates
      maxUnavailable: 1 # only one pod below the replicas count can be down
    type: RollingUpdate
  selector:
    matchLabels: # these labels must match the pod template used
      app: car-server
  template:
    metadata:
      labels: # these labels must match the select (matchLabels) of the deployment
        app: car-server
      namespace: batman
    spec:
      affinity: # many types of affinity and anti-affinity are availble - check the docs!
        podAntiAffinity: # avoid pods with settings below
          preferredDuringSchedulingIgnoredDuringExecution: # other options include disallowing scheduling entirely if two pods are on the same node
          - weight: 100 # only used if multiple rules are in place and conflict
            podAffinityTerm:
              topologyKey: kubernetes.io/hostname
              labelSelector:
                matchExpressions:
                - key: app # this is the key of the label on the pod we want to avoid
                  operator: In
                  values:
                  - car-server # this is the value of the label on the pod we want to avoid
      containers: # you may have multiple containers that share a hostname and network stack
      - name: temp
        image: jdowni000/temp:v2 # this is the URL to the docker image to use. docker hub is assumed by batman
        env:
        - name: ENV_VAR
          value: "env var value"
        livenessProbe: # if a liveness probe does not pass, the pod will be removed from the load balancer and eventually restarted
          failureThreshold: 3
          initialDelaySeconds: 2
          periodSeconds: 15
          successThreshold: 1
          tcpSocket:
            port: 8000
          timeoutSeconds: 1
        readinessProbe: # if a readiness probe does not pass, the pod will not be put into the load balancer
          failureThreshold: 3
          initialDelaySeconds: 2
          periodSeconds: 15
          successThreshold: 1
          tcpSocket:
            port: 8000
          timeoutSeconds: 1
        #args: ["--flag", "here"] # command line arguments
        imagePullPolicy: Always
        resources:
          requests:
            cpu: 50m # millicores are 1/1000 of a full core
            memory: 50Mi # Mi Gi GB and MB are all supported
      restartPolicy: Always
      terminationGracePeriodSeconds: 5 # if sent an interrupt, the pod is allowed up to this long to exit gracefully
---
apiVersion: policy/v1beta1
kind: PodDisruptionBudget
metadata:
  name: car-server-pdb
  namespace: batman
spec:
  minAvailable: 1 # only one pod that matches the selector below can be unavailable at any time
  selector:
    matchLabels:
      app: car-server
---
# This service will be available as car-server.batman (name . namespace)
apiVersion: v1
kind: Service
metadata:
  labels:
  annotations:
    # service.beta.kubernetes.io/aws-load-balancer-internal: "0.0.0.0/0" # use an internal load balancer, or remove for an external one
    external-dns.alpha.kubernetes.io/hostname: "example-app.alpha.k8s.aort.theplatform.com." # create a DNS entry here. Must be under the cluster's VPC subdomain
  name: car-server
  namespace: batman
spec:
  type: LoadBalancer # LoadBalancer indicates we want this service to have an external ELB.  By batman, services are available only within the cluster
  ports:
  - port: 80 # listen publicly on 80
    targetPort: 8000 # forward requests to healthy pods matching the below selector on this port
  selector:
    app: car-server
---
# The HorizontalPodAutoscaler (hpa) is responsible for adjusting the number of replicas in a deployment as load changes
apiVersion: autoscaling/v2beta1
kind: HorizontalPodAutoscaler
metadata:
  name: car-server
  namespace: batman
spec:
  scaleTargetRef:
    apiVersion: apps/v1beta1
    kind: Deployment
    name: car-server
  minReplicas: 2 # at least two replicas for redundancy (think of the PDB above here)
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      targetAverageUtilization: 90 # we want all our pods to operate with 50% of their requested millicores
# The following types can be used at the same time, or in replacement of, the CPU target use metric above
#  - type: Pods
#    pods:
#      metricName: packets-per-second
#      targetAverageValue: 1k
#  - type: Object
#    object:
#      metricName: requests-per-second
#      target:
#        apiVersion: extensions/v1beta1
#        kind: Ingress
#        name: main-route
#      targetValue: 10k