Container Service for Kubernetes:Make CronHPA compatible with HPA

How it works

By default, both CronHPA and HPA set their scaleTargetRef to the same Deployment. Whichever acts last overwrites the other's result.

The fix: set CronHPA's scaleTargetRef to the HPA resource instead of the Deployment. With this setup:

• CronHPA reads HPA's minReplicas, maxReplicas, and desiredReplicas before acting.

• CronHPA adjusts HPA's minReplicas or maxReplicas based on its scheduled targetSize.

• HPA scales pods through the Deployment and its ReplicaSet.

CronHPA never changes the pod count directly, eliminating conflicts.

Configure CronHPA to target HPA

In a typical setup, CronHPA and HPA both point at the same Deployment:

Sample CronHPA template (before modification)

apiVersion: autoscaling.alibabacloud.com/v1beta1
kind: CronHorizontalPodAutoscaler
metadata:
  labels:
    controller-tools.k8s.io: "1.0"
  name: cronhpa-sample
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx-deployment-basic
  jobs:
  - name: "scale-down"
    schedule: "30 */1 * * * *"
    targetSize: 1
  - name: "scale-up"
    schedule: "0 */1 * * * *"
    targetSize: 11

Sample HPA template

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: hpa-sample
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx-deployment-basic
  minReplicas: 4
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50

To resolve the conflict, change scaleTargetRef in your CronHPA manifest to point at the HPA resource:

Example: CronHPA configured to scale business-hours traffic

Assume hpa-sample manages traffic from 09:00 to 17:00 (UTC+8). Configure CronHPA to scale out at 09:00 and in at 17:00:

apiVersion: autoscaling.alibabacloud.com/v1beta1
kind: CronHorizontalPodAutoscaler
metadata:
  labels:
    controller-tools.k8s.io: "1.0"
  name: cronhpa-sample
spec:
  scaleTargetRef:
    apiVersion: autoscaling/v2
    # Set the resource type to HPA.
    kind: HorizontalPodAutoscaler
    # Specify the name of the HPA resource.
    name: hpa-sample
  jobs:
  - name: "scale-up-9am"
    schedule: "0 0 9 * * * "
    targetSize: 20
    runOnce: false
  - name: "scale-down-5pm"
    schedule: "0 0 17 * * *"
    targetSize: 3
    runOnce: false

Compatibility rules

When CronHPA targets HPA, it acts as a dynamic floor for replicas: raising minReplicas to scale out and lowering minReplicas to scale in, without setting the pod count directly. HPA scales within those bounds based on metrics.

The following table shows how CronHPA and HPA interact under each scaling condition.

Table headers:

• HPA (min/max): HPA's minReplicas/maxReplicas.

• CronHPA desired: The CronHPA job's targetSize.

• Current replicas: Running pod count.

• Final replicas: Pod count after scaling.

Next steps

• Implement Horizontal Pod Autoscaler (HPA)

• Use CronHPA for scheduled horizontal scaling