Node Pools

Open the cluster → Node Pools tab in my.cubepath.com.

A node pool is a group of identical workers — same plan, same settings. Splitting workers into pools is how you mix machine sizes in one cluster and steer workloads at the right hardware.

Why more than one pool

A single pool is fine to start. You add pools when workloads stop being interchangeable:

• A small general pool for web services plus a large memory pool for a database or cache.
• A pool of dedicated CPU workers for latency-sensitive services, alongside cheaper shared workers for batch jobs.
• A pool you can drain and replace independently, so a risky change doesn't touch everything.

You target a pool with labels (nodeSelector / nodeAffinity) and keep unwanted pods off it with taints.

Create a pool

Click Create Pool and set the Pool Name, the plan and the worker count. The pool provisions its VPS workers and joins them to the cluster automatically.

Scale a pool

Change the worker count to scale manually, or leave Auto Scale on and set min and max.

Note

With autoscaling enabled, the cluster autoscaler adds nodes when pods can't be scheduled, and removes them when they're underutilized — between your min and max. Set the minimum to what you need at your quietest hour, not your busiest: that's the floor you always pay for.

Each pool row lists its workers with their VPS status, K8s status, private IP and floating IP, so you can tell "the VPS is down" apart from "the VPS is up but the kubelet isn't ready".

Labels

Click Labels to edit the pool's node labels, one per line in key=value format. Labels are applied to every node in the pool, and are what your nodeSelector rules match against. Up to 50 per pool.

Auto-repair

Pools show an Auto-repair setting. When a worker goes unhealthy and stays that way, it's replaced automatically rather than sitting in the cluster failing scheduling.

Tip

Auto-repair replaces a node — it doesn't preserve anything stored on it. Anything that must survive belongs in a persistent volume, not on the node's local disk.

Remove a worker

Removing an individual worker drains and destroys it: pods are evicted and rescheduled elsewhere first, then the VPS is deleted. This can't be undone.

Warning

Make sure the rest of the pool has room for the evicted pods. Draining a node in a pool that's already at capacity leaves pods pending indefinitely.

Delete a pool

Deleting a pool removes it and destroys all its workers. It can't be undone. Move or reschedule anything pinned to that pool's labels first, or those pods will have nowhere to go.