Skip to main content

Scale in specific nodes

Control which nodes a manual scale-down removes first by setting node priority.

Nodes have a priority from 1 (lowest) to 3 (highest); all nodes default to priority 3. When you scale down (via MachineDeployment or MachineSet), the Machine Controller Manager (MCM) deletes the lowest-priority worker nodes first. Lower a specific node's priority to 1 to make it the preferred target for future scale-downs.

Set node priority and scale down​

  1. Annotate the node you want removed first:

    kubectl annotate no <node-name> worker.fptcloud.com/nodepriority=1

    kubectl annotate command setting node priority

  2. On the portal, manually scale down the Worker Group by adjusting its min/max node count.

    Worker Group min-max node count fields

    Worker Group scale-down confirmation

The system scales in exactly the designated node over a few minutes; the cluster status shows Processing and continues operating normally.

Cluster status showing Processing during scale-in

Cluster status after scale-in completes

Annotate multiple nodes with worker.fptcloud.com/nodepriority=1 to target several specific nodes at once.

note

This feature doesn't apply when the cluster autoscaler scales in worker nodes.