Scale a Kubernetes node pool
Scaling a AME Kubernetes node pool up or down
In your cluster’s detail page, you have an overview of node pools.
Each node pool can be individually modified without impacting other pools, or the cluster as a whole. The following information can be modified:
- Node Type (i.e. size of each virtual machine provisioned)
- Amount of nodes within a pool
- Kubernetes Node Labels
- Kubernetes Node Annotations
Examples
Scaling up/down
When changing the node count for a node pool, AME Kubernetes will do the following:
- When scaling down, it will drain each node that needs to be deprovisioned one by one.
- When scaling up, it will create the missing amount of VMs in parallel and automatically join these to the cluster.
Modifying node type
Changing the node type within a pool results in a full node pool replacement. Any missing nodes will first be provisioned before replacing existing nodes, similarly to how the scaling up works, if you modify both the node type and count within the same action.
Node Labels/Annotations
Labels and annotations will not result in a node replacement and will be configured within a couple of minutes on each node within a pool. Be careful with modifying existing labels when using these for affinity rules within Kubernetes, as this could result in pods being moved around the cluster.
Labels starting with the following prefix cannot be configured:
volumes.kubernetes.io/node.kubernetes.io/windows-buildalpha.kubernetes.io/provided-node-ipkubeadm.alpha.kubernetes.io/cri-socketkubernetes.io/archkubernetes.io/oskubernetes.io/hostnamek8s.avisi.cloud/