Under what conditions will the cluster scale up/down?

Scale-up: Due to the lack of resources, pod scheduling failed, that is, pod has always been in the Pending state. Scale-down: When the resource utilization of a node is low, and the pod existing on the node can be rescheduled to other nodes.

What is the scanning frequency of automatic scale-up/down?

30 seconds

What is the setting range of the maximum and minimum nodes of the scale-up/down?

The minimum node number shall be greater than or equal to 1, and the maximum node number shall be less than or equal to the current cluster node number limit (this limit can be promoted by submitting tickets). The maximum and minimum number of nodes will limit the scope of auto scale-up/down, but will not limit the user's manually initiated scale-up/down.

How to determine the configuration of creating nodes during auto scale-up?

During automatic scale-up, nodes will be created according to the most popular node configuration type in the current cluster, that is, the configuration type with the largest number of nodes in the current cluster.

How to prevent a specific node from being released by scale-down?

The following types of nodes are not released by scale-down:

• The pod on the node is limited by PodDisruptionBudget controller.
• The pod on the node is not created by the controller, such as deployment, replica set, job, stateful set.
• Pod on node uses local storage.
• There is no place to go after the pod is evicted, that is, no other node can call this pod.
• The node has an annotation: "cluster-autoscaler.kubernetes.io/scale-down-disabled": "true".