Container Service for Kubernetes:Enable node autoscaling

Prerequisites

Before you begin, ensure that you have:

• Activated the Auto Scaling service

• Read Node scaling to understand how node auto scaling works, its use cases, and usage notes

• Reviewed Usage notes to understand node scaling quotas and limitations

If your workloads use scheduling policies with known limitations, choose one of the following options before proceeding:

• Option 1: Switch to node instant scaling.

• Option 2: Deploy affected workloads in node pools with auto scaling disabled. For example, to deploy ack-node-local-dns-admission-controller in a node pool with auto scaling disabled, add the following node affinity rule to its configuration:

  nodeAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      nodeSelectorTerms:
      - matchExpressions:
        - key: "k8s.aliyun.com"
          operator: "NotIn"
          values: ["true"]

Enable node auto scaling for a cluster

Node pool auto scaling only takes effect after you enable it at the cluster level.

1. Log in 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 Nodes > Node Pools.

3. On the Node Pools page, click Enable next to Node Scaling.

   

4. If this is your first time enabling node auto scaling, activate the service and complete role authorization as prompted. Otherwise, skip this step.

   • ACK managed cluster: Authorize ACK to use the AliyunCSManagedAutoScalerRole to access your cloud resources.

   • ACK dedicated cluster: Authorize ACK to use the KubernetesWorkerRole and AliyunCSManagedAutoScalerRolePolicy for scaling management.

5. In the Node Scaling Configuration panel, set Node Scaling Method to Auto Scaling, configure the parameters below, then click OK.

   To modify Node Scaling Configuration later, switch to Instant Scaling and follow the prompts.

   Scale-out settings

   Parameter	Description	Default
   Node Pool Scale-out Policy	Determines which node pool to scale out when multiple eligible node pools exist. Random Policy: selects a node pool randomly. Default Policy: selects the node pool that wastes the least resources. Priority-based Policy: scales out node pools in priority order, defined by Node Pool Scale-out Priority.	—
   Node Pool Scale-out Priority	Scale-out priority for each node pool, expressed as an integer from 1 to 100. A higher number means higher priority. Only applies when Node Pool Scale-out Policy is set to Priority-based Policy. Click + Add, select a node pool with auto scaling enabled, and set a priority value. If no auto-scaling node pools are available yet, skip this parameter and configure it after completing Configure a node pool with auto scaling enabled.	—
   Scan Interval	How often the cluster is evaluated for scale-out or scale-in.	60s

   Scale-in settings

   Important

   Scale-in for Elastic Compute Service (ECS) nodes requires the Scale-in Threshold, Defer Scale-in For, and Cooldown conditions to all be met. Scale-in for GPU-accelerated nodes additionally requires the GPU Scale-in Threshold condition.

   Parameter	Description	Default
   Allow Scale-in	Enables scale-in. When disabled, all scale-in configuration is ignored.	—
   Scale-in Threshold	The maximum ratio of a node's resource requests to its capacity at which scale-in is triggered. Scale-in only occurs when both CPU and memory utilization fall below this threshold.	—
   GPU Scale-in Threshold	The scale-in threshold for GPU-accelerated nodes. Scale-in only occurs when CPU, memory, and GPU utilization all fall below this threshold.	—
   Defer Scale-in For	How long after a scale-in condition is detected before the scale-in is executed. Unit: minutes. Important The autoscaler performs scale-in activities only when Scale-in Threshold is configured and the Defer Scale-in For condition is met.	10
   Cooldown	After a scale-out, the autoscaler waits for this period before performing scale-in. During the cooldown, scale-in conditions are still checked, but no scale-in is executed. After the cooldown ends, nodes that have met the scale-in conditions for the full Defer Scale-in For duration are removed. For example, with Cooldown set to 10 minutes and Defer Scale-in For set to 5 minutes, nodes are removed 5 minutes after the cooldown ends.	—

   Advanced scale-in settings

   Parameter	Description
   Pod Termination Timeout	Maximum time to wait for pods on a node to terminate during scale-in. Unit: seconds.
   Minimum Number Of Replicated Pods	Minimum number of pods in a ReplicaSet before scale-in is allowed. Nodes are not scaled in if the ReplicaSet has fewer replicas than this value.
   Evict DaemonSet Pods	When enabled, DaemonSet pods are evicted during scale-in.
   Skip Nodes Hosting Kube-system Pods	When enabled, nodes running pods in the kube-system namespace are excluded from scale-in. Does not apply to mirror pods or DaemonSet pods.

Configure a node pool with auto scaling enabled

Node auto scaling only applies to node pools with Scaling Mode set to Auto. Either modify an existing node pool or create a new one. For full steps, see Create and manage a node pool.

The following table describes the key parameters to configure:

(Optional) Verify node auto scaling

After completing the configuration above, verify that auto scaling is active.

FAQ

What's next

For clusters with more than 100 nodes in an auto-scaling node pool, or more than 20 node pools with auto scaling enabled, switch to node instant scaling for faster provisioning and broader instance type support. See Enable node instant scaling.