Learn OpenShift

Memory-based autoscaling

Unlike autoscaling based on CPU utilization, memory-based autoscaling can only be enabled by creating a HorizontalPodAutoscaler from a raw YAML/JSON definition:

# cat hpa-memory.yml kind: HorizontalPodAutoscalerapiVersion: autoscaling/v1metadata:  name: hpa-httpd-memoryspec:  scaleTargetRef:    apiVersion: v1    kind: DeploymentConfig    name: httpd  minReplicas: 2  maxReplicas: 4  metrics:  - type: Resource    resource:      name: memory      targetAverageUtilization: 10

Let's enable autoscaling now:

# oc create -f hpa-memory.yml horizontalpodautoscaler "hpa-httpd-memory" created

Give it a minute or two to pick up the metrics from Heapster and you will be able to see how the current memory utilization is different from the target:

# oc get ...