CoScale Kubernetes Autoscaler

The CoScale Autoscaler retrieves metric data from the CoScale platform, checks whether the metric data is within the provided bounds and scales a Kubernetes Deployment accordingly.

Configuration

The autoscaler reads the configuration (in JSON format) from the SCALER_CONFIG environment variable. For example:

[
  {
    "namespace_name": "production",
    "deployment_name": "apache",
    "metric": {
      "name": "Docker container cpu usage percent from limit",
      "low_value": 20,
      "high_value": 80,
      "avg_interval_sec": 300
    },
    "scale_backoff_sec": 600,
    "min_replicas": 3,
    "max_replicas": 10
  }
]

This configuration checks the Docker container cpu usage percent from limit metric for the apache deployment in the production namespace. The average metric value for all containers in the deployment is calculated over a period of 300 seconds. If the average value is lower than 20, the number of replicas in the deployment is decreased by 1, but not below the minimum number of replicas 3. If the average value is higher than 80, the number of replicas in the deployment is increased by 1, but not above the maximum number of replicas 10. The metric values are checked every minute, scaling will only take place if 600 seconds have elapsed since the last scaling.

To connect to the CoScale platform the following environment variables are used

• API_URL: The base url for the CoScale Platform (eg. https://api.coscale.com)
• APP_ID: The application UUID
• ACCESS_TOKEN: A valid access token (UUID) for the provided APP_ID

Both the APP_ID and ACCESS_TOKEN can be found on the "Users & Teams > Access Tokens" page in the CoScale UI.

Deployment

The CoScale Autoscaler should be deployed on the Kubernetes cluster that you want to autoscale. The container needs a ServiceAccount and ClusterRoleBinding to communicate with the Kubernetes API.

cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: ServiceAccount
metadata:
  name: coscale-autoscaler
  namespace: coscale
---
apiVersion: rbac.authorization.k8s.io/v1beta1
kind: ClusterRoleBinding
metadata:
  annotations:
    rbac.authorization.kubernetes.io/autoupdate: "true"
  labels:
    kubernetes.io/bootstrapping: rbac-defaults
  name: coscale-autoscaler
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: cluster-admin
subjects:
- kind: ServiceAccount
  name: coscale-autoscaler
  namespace: coscale
---
apiVersion: extensions/v1beta1
kind: ReplicaSet
metadata:
  labels:
    name: coscale-autoscaler
  name: coscale-autoscaler
  namespace: coscale
spec:
  replicas: 1
  template:
    metadata:
      labels:
        name: coscale-autoscaler
    spec:
      serviceAccountName: coscale-autoscaler
      containers:
      - image: coscale/coscale-kubernetes-autoscaler:1.1.0
        name: autoscaler
        env:
        - name: API_URL
          value: "https://api.coscale.com"
        - name: APP_ID
          value: "00000000-0000-0000-0000-000000000000"
        - name: ACCESS_TOKEN
          value: "00000000-0000-0000-0000-000000000000"
        - name: SCALER_CONFIG
          value: '[{"namespace_name":"production","deployment_name":"apache","metric":{"name":"Docker container cpu usage percent from limit","low_value":20,"high_value":80,"avg_interval_sec":300},"scale_backoff_sec":600,"min_replicas":3,"max_replicas":10}]'
EOF

Don't forget to update API_URL, APP_ID and ACCESS_TOKEN in the example above.

OpenShift Deployment Configs

The CoScale Autoscaler has support for both Kubernetes Deployments and OpenShift Deployment configs (https://docs.openshift.com/enterprise/3.0/dev_guide/deployments.html).

Add the deployment_type field with value Deployment configs to the configuration to use the Autoscaler with OpenShift Deployment configs:

[
  {
    "namespace_name": "production",
    "deployment_type": "Deployment configs",
    "deployment_name": "apache",
    "metric": {
      "name": "Docker container cpu usage percent from limit",
      "low_value": 20,
      "high_value": 80,
      "avg_interval_sec": 300
    },
    "scale_backoff_sec": 600,
    "min_replicas": 3,
    "max_replicas": 10
  }
]

Deploying the CoScale autoscaler on OpenShift can be done using the following commands:

oc project coscale
oc create serviceaccount coscale
oadm policy add-cluster-role-to-user cluster-admin system:serviceaccount:coscale:coscale

cat << EOF | oc apply -f -
apiVersion: extensions/v1beta1
kind: ReplicaSet
metadata:
  labels:
    name: coscale-autoscaler
  name: coscale-autoscaler
  namespace: coscale
spec:
  replicas: 1
  template:
    metadata:
      labels:
        name: coscale-autoscaler
    spec:
      serviceAccountName: coscale-autoscaler
      containers:
      - image: coscale/coscale-kubernetes-autoscaler:1.1.0
        name: autoscaler
        env:
        - name: API_URL
          value: "https://api.coscale.com"
        - name: APP_ID
          value: "00000000-0000-0000-0000-000000000000"
        - name: ACCESS_TOKEN
          value: "00000000-0000-0000-0000-000000000000"
        - name: SCALER_CONFIG
          value: '[{"namespace_name":"production","deployment_type":"Deployment configs","deployment_name":"apache","metric":{"name":"Docker container cpu usage percent from limit","low_value":20,"high_value":80,"avg_interval_sec":300},"scale_backoff_sec":600,"min_replicas":3,"max_replicas":10}]'
EOF