Autoscaling

Autoscaling is a feature in which the cluster is capable of increasing or decreasing the number of pods as the demand for service response need it.

Module

The Horizontal Pod Autoscaler automatically scales the number of pods in a replication controller, deployment or replica set based on observed CPU utilization.

Overview

At the end of this module, you will :

  • Learn the format of a YAML Autoscale file

  • Learn how to manage a Autoscale

  • Learn the composition of a Autoscale

Prerequisites

Create the directory data/autoscaling in your home folder to manage the YAML file needed in this module.

mkdir  ~/data/autoscaling
mkdir  ~/data/autoscaling/metrics-server

This module needs the metrics-server to be deployed on the cluster to get the monitoring values like CPU and memory. Ensure that the module is up and running before continuing.

Create

Looks up a Deployment, ReplicaSet, or ReplicationController by name and creates an autoscaler that uses the given resource as a reference. An autoscaler can automatically increase or decrease number of pods deployed within the system as needed.

Horizontal Pod Autoscaler automatically scales the number of pods in a deployment or replica set based on observed CPU, Memory or Custom Metrics utilization depending the API version used.

The Kubernetes basic autoscaling architecture can be schematized like this :

The create command can directly ask the API resource to create an HorizontalPodAutoscaler in command line or create an HorizontalPodAutoscaler object based on a yaml file definition.

Deploy metrics-server on Digital Ocean cluster

By default, metrics-server, the tool that will pull the metrics, is not installed on Digital Ocean cluster, to enable autoscaling, we need to install it.

---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: system:aggregated-metrics-reader
  labels:
    rbac.authorization.k8s.io/aggregate-to-view: "true"
    rbac.authorization.k8s.io/aggregate-to-edit: "true"
    rbac.authorization.k8s.io/aggregate-to-admin: "true"
rules:
- apiGroups: ["metrics.k8s.io"]
  resources: ["pods", "nodes"]
  verbs: ["get", "list", "watch"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: metrics-server:system:auth-delegator
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:auth-delegator
subjects:
- kind: ServiceAccount
  name: metrics-server
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: metrics-server-auth-reader
  namespace: kube-system
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: extension-apiserver-authentication-reader
subjects:
- kind: ServiceAccount
  name: metrics-server
  namespace: kube-system
---
apiVersion: apiregistration.k8s.io/v1beta1
kind: APIService
metadata:
  name: v1beta1.metrics.k8s.io
spec:
  service:
    name: metrics-server
    namespace: kube-system
  group: metrics.k8s.io
  version: v1beta1
  insecureSkipTLSVerify: true
  groupPriorityMinimum: 100
  versionPriority: 100
---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: metrics-server
  namespace: kube-system
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: metrics-server
  namespace: kube-system
  labels:
    k8s-app: metrics-server
spec:
  selector:
    matchLabels:
      k8s-app: metrics-server
  template:
    metadata:
      name: metrics-server
      labels:
        k8s-app: metrics-server
    spec:
      serviceAccountName: metrics-server
      volumes:
      # mount in tmp so we can safely use from-scratch images and/or read-only containers
      - name: tmp-dir
        emptyDir: {}
      containers:
      - name: metrics-server
        image: k8s.gcr.io/metrics-server-amd64:v0.3.6
        args:
          - --cert-dir=/tmp
          - --secure-port=4443
          - --kubelet-insecure-tls
          - --kubelet-preferred-address-types=InternalIP,ExternalIP,Hostname
        ports:
        - name: main-port
          containerPort: 4443
          protocol: TCP
        securityContext:
          readOnlyRootFilesystem: true
          runAsNonRoot: true
          runAsUser: 1000
        imagePullPolicy: Always
        volumeMounts:
        - name: tmp-dir
          mountPath: /tmp
      nodeSelector:
        beta.kubernetes.io/os: linux
---
apiVersion: v1
kind: Service
metadata:
  name: metrics-server
  namespace: kube-system
  labels:
    kubernetes.io/name: "Metrics-server"
    kubernetes.io/cluster-service: "true"
spec:
  selector:
    k8s-app: metrics-server
  ports:
  - port: 443
    protocol: TCP
    targetPort: main-port
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: system:metrics-server
rules:
- apiGroups:
  - ""
  resources:
  - pods
  - nodes
  - nodes/stats
  - namespaces
  - configmaps
  verbs:
  - get
  - list
  - watch
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: system:metrics-server
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:metrics-server
subjects:
- kind: ServiceAccount
  name: metrics-server
  namespace: kube-system
kubectl apply -f ~/data/autoscaling/metrics-server/metrics-server.yaml

Exercise n°1

  1. Run a sample app based on a webserver to expose it on port 80.

  2. Create an Horizontal Pod Autoscaler to automatically scale the Deployment if the CPU usage is above 50%.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: php-apache
  labels:
    app: php-apache
spec:
  replicas: 3

  selector:
    matchLabels:
      app: php-apache
  template:
    metadata:
      labels:
        app: php-apache
    spec:
      containers:
      - name: php-apache
        image: k8s.gcr.io/hpa-example
        resources:
          requests:
            cpu: "200m"
          limits:
            cpu: "300m"
        ports:
        - name: php-apache
          containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
  name: php-apache
spec:
  type: ClusterIP
  selector:
    app: php-apache
  ports:
    - name: php-apache
      port: 80
      protocol: TCP
      targetPort: php-apache
kubectl apply -f data/autoscaling/01_php-apache.yaml
# version kubectl <1.17
kubectl run php-apache --image=k8s.gcr.io/hpa-example --requests=cpu=200m --limits=cpu=300m --expose --port=80

# Create an Horizontal Pod Autoscaler based on the CPU usage
kubectl autoscale deployment php-apache --cpu-percent=50 --min=3 --max=10

Exercise n°2

  1. Run a sample nginx application exposing port 8080

  2. Create an Horizontal Pod Autoscaler to automatically scale the Deployment if the memory is above 80%.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx
  labels:
    app: nginx
spec:
  replicas: 1

  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: nginx
        resources:
          requests:
            memory: "500Mi"
          limits:
            memory: "1Gi"
        ports:
        - name: nginx
          containerPort: 8080
---
apiVersion: v1
kind: Service
metadata:
  name: nginx
spec:
  type: ClusterIP
  selector:
    app: nginx
  ports:
    - name: nginx
      port: 8080
      protocol: TCP
      targetPort: nginx
kubectl apply -f data/autoscaling/02_deployment.yaml
# version kubectl <1.17
kubectl run nginx --image=nginx --requests=memory=500m --limits=memory=1G --expose --port=8080
apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: nginx
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx
  minReplicas: 1
  maxReplicas: 5
  metrics:
  - type: Resource
    resource:
      name: memory
      target:
        type: Utilization
        averageUtilization: 80
kubectl apply -f data/autoscaling/01_hpa.yaml

Get

The get command list the object asked. It could be a single object or a list of multiple objects comma separated. This command is useful to get the status of each object. The output can be formatted to only display some information based on some json search or external tools like tr, sort, uniq.

The default output display some useful information about each services :

  • Name : the name for the newly created object

  • Reference : the object managed by the autoscaler, like Pod name, a Deployment name ...

  • Targets : the metrics defined to autoscale the referenced resource

  • Minpods : the lower limit for the number of pods that can be set by the autoscaler

  • Maxpods : the upper limit for the number of pods that can be set by the autoscaler

  • Replicas : the current replicas number

  • Age : the age of the object from his creation

Exercise n°1

Get the current HorizontalPodAutoscaler resources in the default namespace.

kubectl get hpa

Exercise n°2

Stress the Pod created in the previous section and check the HorizontalPoMindAutoscaler associated.

# Connect to the Pod
kubectl run -it load-generator --image=busybox /bin/sh

# Run a loop bash command in the container to stress the CPU
while true; do wget -q -O- http://php-apache.default.svc.cluster.local; done

# Check the Horizontal Pod Autoscaler status
kubectl get hpa

Scale the load-generator if you want to stress the php-apache Pods quickly.

Longer Execution Times

The autoscaling can take more than 2 minutes to run. Please be patient. Do not close the window or cancel the operation.

Exercise n°3

Stop to stress the Pod previously created and check that the autoscaler come back to normal.

kubectl get hpa

Describe

Once an object is running, it is inevitably a need to debug problems or check the configuration deployed.

The describe command display a lot of configuration information about the Horizontal Pod Autoscaler (labels, annotations, etc.) and the scale policy (selector, type, number of pods, ...).

This command is really useful to introspect and debug an object deployed in a cluster.

Exercise n°1

Describe one of the existing Autoscaler in the default namespace.

kubectl describe horizontalpodautoscaler php-apache

Explain

Kubernetes come with a lot of documentation about his objects and the available options in each one. Those information can be fin easily in command line or in the official Kubernetes documentation.

The explain command allows to directly ask the API resource via the command line tools to display information about each Kubernetes objects and their architecture.

Exercise n°1

Get the documentation of a specific field of a resource.

kubectl explain hpa.spec

Add the --recursive flag to display all of the fields at once without descriptions.

Delete

The delete command delete resources by filenames, stdin, resources and names, or by resources and label selector.

Be careful on the deletion of an autoscaling object, this can have effects in the availability of the services associated.

Note that the delete command does NOT do resource version checks, so if someone submits an update to a resource right when you submit a delete, their update will be lost along with the rest of the resource.

Exercise n°1

Delete the previous autoscaling group created in command line.

# Delete the HorizontalPodAutoscaler
kubectl delete hpa php-apache

# Delete the Pods
kubectl delete deployment php-apache load-generator

# Delete the Services
kubectl delete service php-apache

Module exercise

The purpose of this section is to manage each steps of the lifecycle of an application to better understand each concepts of the Kubernetes course.

The main objective in this module is to understand how to dynamically and automatically manage the number of Pods needed to handle the workload.

For more information about the application used all along the course, please refer to the Exercise App > Voting App link in the left panel.

Based on the principles explain in this module, try by your own to handle this steps. The development of a yaml file is recommended.

The file developed has to be stored in this directory : ~/data/votingapp/10_autoscaling

  1. Manage the HorizontalPodAutoscaler of the worker Pods to :

    1. Ensure that the worker has minimum one Pods

    2. Ensure that the worker has maximum five Pods

    3. Ensure that the Pods is autoscaled when the CPU is above 80%.

External documentation

Those documentations can help you to go further in this topic :

PreviousQuotas & LimitsNextNetworks

Last updated