Container Service for Kubernetes:Use Horizontal Pod Autoscaling (HPA)

Key timing parameters

For more details on the core algorithm and configurable behaviors, see the Kubernetes Horizontal Pod Autoscaling documentation.

Container Service for Kubernetes (ACK) provides several workload and node scaling solutions. For a comparison of available options, see Auto Scaling.

HPA configuration parameters

The following table describes the parameters available across all console entry points. The parameter labels vary slightly depending on which page you use.

If you specify both CPU and memory metrics, HPA triggers a scaling event when either metric exceeds its threshold.

Metric availability by entry point:

Option 1: Create HPA with a new application

This example uses a stateless Deployment. The steps are similar for other workload types.

1. Log on to the ACK console. In the left navigation pane, click Clusters.

2. On the Clusters page, click the name of the target cluster. In the left navigation pane, choose Workloads > Deployments.

3. On the Deployments page, click Create From Image.

4. On the Create page, configure the application: For complete configuration details, see Create a stateless workload (Deployment).

   • Basic Information: Set the application name, replica count, and other details.

   • Container Configuration: Set the container image and resource requests (CPU and memory).

   • Advanced Configuration > Scaling: Select HPA and click Enable, then configure the metric, target utilization, minimum replicas, and maximum replicas.

After the Deployment is created, click the Deployment name on the Deployments page and open the Pod Scaling tab. This tab shows HPA metrics (CPU and memory usage, replica range) and provides options to update or disable the policy.

Option 2: Add HPA to an existing application

1. Log on to the ACK console. In the left navigation pane, click Clusters.

2. On the Clusters page, click the name of the target cluster. In the left navigation pane, choose Workloads > Deployments.

3. On the Deployments page, click the target application name. Open the Pod Scaling tab and click Create in the HPA section.

4. In the Create dialog box, configure the HPA policy:

   • Name: Enter a policy name.

   • Metric: Click Add to select a metric (CPU Usage or Memory Usage) and set the Threshold (target utilization percentage).

   • Max. Containers: Set the maximum replica count.

   • Min. Containers: Set the minimum replica count.

After the policy is created, click the Deployment name and open the Pod Scaling tab to view HPA metrics and manage the policy.

Option 3: Use the Workload Scaling page

1. Log on to the ACK console. In the left navigation pane, click Clusters.

2. On the Clusters page, click the name of the target cluster. In the left navigation pane, click Workload Scaling.

3. In the upper-right corner, click Create Auto Scaling, then open the HPA and CronHPA tab.

4. Select the target workload. In the Configure Scaling Policy section, select the HPA checkbox and configure the policy:

   • Scaling Policy Name: Enter a policy name.

   • Min. Containers: Set the minimum replica count (integer >= 1).

   • Max. Containers: Set the maximum replica count (must be greater than the minimum).

   • Scaling Metric: Select one or more metric categories and configure the threshold for each:

     • Resource: CPU usage and memory usage. Available by default.

     • Custom: GPU memory usage, GPU utilization, and custom metrics. Requires ack-alibaba-cloud-metrics-adapter.

     • External: Nginx Ingress QPS and custom metrics. Requires ack-alibaba-cloud-metrics-adapter.

   Note

   If you select Custom or External metrics and ack-alibaba-cloud-metrics-adapter is not installed, the console displays an Install button. Click Install to deploy the adapter before configuring these metrics.

After the policy is created, view and manage it on the Workload Scaling page. The Actions column provides options to view metrics, update the configuration, or disable the policy.

Step 1: Deploy a sample application

Create a file named nginx.yml:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx
  labels:
    app: nginx
spec:
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: registry-cn-hangzhou-vpc.ack.aliyuncs.com/acs/nginx:1.27.0 # Replace the region ID with the region of your cluster.
        ports:
        - containerPort: 80
        resources:
          requests:         # Required. Without requests, HPA cannot calculate utilization.
            cpu: 500m

Apply the Deployment:

kubectl apply -f nginx.yml

Step 2: Create the HPA resource

Create a file named hpa.yml. The scaleTargetRef field points to the Deployment that HPA manages. The example below scales the Deployment between 1 and 10 replicas, targeting 50% average CPU utilization.

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: nginx-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: nginx-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50

To scale on both CPU and memory, add both resource types in the metrics field of a single HPA. Do not create separate HPAs for each metric.

metrics:
- type: Resource
  resource:
    name: cpu
    target:
      type: Utilization
      averageUtilization: 50
- type: Resource
  resource:
    name: memory
    target:
      type: Utilization
      averageUtilization: 50

Apply the HPA:

kubectl apply -f hpa.yml

Step 3: Verify the HPA

After applying the HPA, check its status:

kubectl get hpa nginx-hpa

During initial deployment, you may see warnings indicating that HPA is still collecting metrics:

Warning  FailedGetResourceMetric       2m (x6 over 4m)  horizontal-pod-autoscaler  missing request for cpu

Wait for HPA to start collecting metrics, then verify that it is operating normally:

kubectl describe hpa nginx-hpa

Expected output when HPA is running and the load is below the threshold:

Type    Reason             Age   From                       Message
----    ------             ----  ----                       -------
Normal  SuccessfulRescale  5m6s  horizontal-pod-autoscaler  New size: 1; reason: All metrics below target

Related topics

• Horizontally scale pods with Alibaba Cloud metrics -- Use External Metrics from Alibaba Cloud components to drive HPA.

• Horizontal pod autoscaling based on Prometheus metrics -- Convert Prometheus metrics into HPA-compatible metrics.

• Node autoscaling FAQ -- Troubleshoot common node scaling issues.

• Coordinate CronHPA and HPA -- Use scheduled and metric-based scaling together.

Other scaling solutions

• CronHPA for scheduled horizontal scaling -- Scale pods on a cron-like schedule for periodic traffic patterns.

• Predictive scaling (AHPA) -- Automatically identify business cycles from historical metrics and scale pods proactively.

• Vertical Pod Autoscaling (VPA) -- Automatically adjust resource requests and limits based on usage.

• Event-driven autoscaling -- Scale pods based on Kubernetes events, message queues, or custom triggers.

Combined solutions

Use HPA with node autoscaling to automatically add nodes when pod scaling exhausts available cluster resources.