Create a Managed Kubernetes Cluster from the Deploy Kubernetes page in my.cubepath.com. CubePath runs the control plane for you while your worker nodes run on CubePath VPS that you size and scale — you just download a kubeconfig and use kubectl like any other cluster.
NoteYour account must be verified before you can deploy.
What "managed" means here
CubePath runs the API server, scheduler, controller manager and etcd, keeps them patched and healthy, and hands you an endpoint. You run everything above that line: your workloads, your node pools, your add-ons.
Each worker is a real VPS in your account, billed like any other VPS. That's the cost model — you pay for the workers you run, plus the HA control plane if you enable it.
Deploy
- 1Name your clusterPick the project and a cluster name (lowercase letters, numbers and dashes).
- 2Choose version and control planeSelect the Kubernetes version (the default is pre-selected) and optionally enable a high-availability control plane — 2 control-plane replicas, +$20.00/mo. Worth it for production; leave it off for dev and test.
- 3Pick region and instance typeChoose a location in Miami (MIA), Houston (HOU) or Barcelona (BCN), then an instance type for the worker nodes.
- 4Configure node poolsSet each pool's name, plan and node count (1–100). Auto Scale is on by default; add more pools with Add Pool.
- 5DeployOptionally adjust Advanced Settings (IPv4/IPv6, pod/service CIDRs, private network), then click Deploy Cluster.
The cluster provisions in the background; its status moves through Provisioning to Active.
ImportantDisabling public IPv6 requires attaching a private network first, so worker nodes stay reachable.
Access
Clusters are accessed with kubectl, not SSH. Once the cluster is active, open it, click Cluster Management, then Download Kubeconfig to get the <cluster>-kubeconfig.yaml file. Point kubectl at it:
export KUBECONFIG=./<cluster>-kubeconfig.yaml
kubectl get nodes
From there it's a standard cluster: apply manifests, install Helm charts, run whatever you like.
WarningThe kubeconfig grants full administrative access to the cluster. Treat it like a private key — don't commit it, and don't paste it into a shared channel.
Expose your apps
A fresh cluster has no public entry point. To send internet traffic to your services:
- 1Install an ingress controllerPick one from the Addons tab.
- 2Put a load balancer in frontDeploy one from the Load Balancers tab, targeting the worker nodes.
- 3Point DNS at itCreate a record in DNS records for the load balancer's IP.
For internal-only clusters, attach a private network at deploy time and skip public worker IPs.
Protection and deletion
Enable Protection blocks accidental deletion of the cluster. Deleting is permanent: it removes the cluster, every node pool, and destroys all worker VPS instances. Disable protection first if it's on.
Limits and constraints
| Thing | Limit |
|---|---|
| Node pools per cluster | 1–10 |
| Nodes per pool | 1–100 |
| Labels per pool | 50 |
| Taints per pool | 50 |
- Pod and service CIDRs default to
10.42.0.0/16and10.43.0.0/16. Custom ranges must not overlap each other, the node network, or your private network. - No SSH to the control plane. You interact with the cluster only through the Kubernetes API.
- One location per cluster. Workers and the private network live in a single region.
The cluster tabs
Node Pools, Network, Addons, Load Balancers, Metrics and Activity.
NoteEverything here is also available over the CubePath API (
Authorization: Bearer <token>, scopeskubernetes:read/kubernetes:write; write requests also needX-Requested-With: XMLHttpRequest). See api.cubepath.com.