Skip to main content

Node pools

A node pool is a group of identically-sized worker nodes. The cluster's Node pools tab lists them with Name, Flavor, Scale (min–max, auto?), Nodes (current count) and Status.

Node pools

The row menu (⋯) has View detail, Edit, Scale, View machines, Recover and Delete.

Add a pool​

New pool opens the pool builder — a derived read-only Name, a Type (Standard / Processor Optimized / Memory Optimized), a Flavor, optional GPU (model + count), Min / Max nodes, Storage (GiB) and an autoscaling toggle.

Create node pool

Edit a pool​

Edit changes the pool's name, description and scaling, and — importantly — its Labels and Taints. Labels and taints control which workloads schedule onto the pool.

Edit node pool

note

Reserved label prefixes (kubernetes.io/, k8s.io/) are flagged — use your own keys for scheduling labels. Taints take an effect of NoSchedule, PreferNoSchedule or NoExecute.

Scale a pool​

Scale sets the desired node count directly. (With autoscaling enabled the pool moves within its min–max range on its own; scaling manually sets a target inside that range.)

Scale node pool

View machines​

View machines lists the pool's nodes with their IP and Status. Each node has Drain (cordon and evict its pods so they reschedule elsewhere — the node stays in the pool) and Remove (type-to-confirm; permanently removes the node).

Pool machines

Draining and removing nodes disrupt workloads

Drain evicts running pods; Remove and Delete node pool (type-to-confirm) take the node(s) out of the cluster for good. Make sure your workloads have somewhere else to schedule before you drain or remove capacity.