flex.planeflex.plane
Kubernetes

Scale a Cluster

Add or remove worker nodes.

Scaling a Kubernetes cluster in flex.plane means changing the size of its node pools. You can add nodes to handle more load, remove nodes to save resources, or reconfigure pools entirely.

Scale up

To add worker nodes, navigate to the cluster's detail page and adjust the node pool sizes. The scale operation is declarative — you specify the desired state of all node pools, and the platform reconciles to match.

The cluster detail page shows each node pool with its current size, status, and individual nodes.

New nodes are provisioned as VMs, boot up, install Kubernetes components, and join the cluster. The process typically takes a few minutes per node.

Scaling is declarative. You specify the desired state of all node pools, and the platform reconciles the current state to match. Include all pools in every scale operation, not just the ones you are changing.
You can also add entirely new node pools by including them in the nodePools array. This is how you expand into new zones or add pools with different compute profiles.

Scale down

To remove nodes, decrease the size of the node pool. The platform gracefully drains and removes the excess nodes.

When scaling down:

  1. The platform selects which nodes to remove.
  2. Selected nodes are cordoned (no new pods scheduled).
  3. Existing pods are drained (gracefully evicted).
  4. The underlying VMs are deleted.
Scaling down to 0 nodes is allowed but means no pods can run in that pool. Make sure your workloads have somewhere else to go, or you will have pending pods.

To remove a node pool entirely, simply omit it from the nodePools array in the next scale operation.

Monitor scaling progress

After a scale operation, the cluster enters the Scaling state. Monitor progress on the cluster detail page.

Scaling is complete when:

  • The cluster state returns to Running.
  • Each pool's readyReplicas matches its size.
  • unavailableReplicas is 0.
If scaling seems stuck, check the state.reason fields. Common issues include insufficient host capacity in the target zone, storage provisioning failures, or network configuration problems.

Manage Node Pools

Configure and monitor node pools.

Access a Cluster

Retrieve the kubeconfig and connect.