Auto scaling FAQs

Under what conditions will the cluster scale up or down?

• Scale-up: When insufficient resources prevent pod scheduling, resulting in pods being stuck in a pending state.
• Scale-down: When a node's resource utilization is low, and all pods on that node can be rescheduled to run on other nodes.

What is the auto scaling scan frequency?

30 seconds

What is the range for setting the maximum and minimum number of nodes during scaling?

The minimum node count must be ≥0, and the maximum node count must be less than the current cluster node quota (this quota can be increased by submitting a ticket). These limits define the range for auto-scaling but do not restrict manually initiated scaling.

How is node configuration determined during auto scale-up?

During auto scale-up, new nodes will be created based on the most common node configuration type in the cluster, which is the node type with the highest count in the current cluster.

How to prevent specific nodes from being scaled down and released?

The following types of nodes will not be scaled down or released:

• The pod on the node is constrained by the PodDisruptionBudget controller.
• The pod on the node is not managed by a controller, such as a deployment, replica set, job, or stateful set.
• Pod on the node is stored locally
• After a pod is evicted from a node, it has nowhere to go, meaning no other node can schedule this pod
• The node has the annotation: "cluster-autoscaler.kubernetes.io/scale-down-disabled": "true"