FAQ

Do I need a Cloud Compute service first?

Yes. Managed Kubernetes runs on top of your Cloud Compute service — clusters draw their CPU, memory and storage from it. If you don't have one, the activation page prompts you to activate Cloud Compute first.

How do I connect with kubectl?

Use Download KubeConfig on any cluster page, or generate one under the cluster's Settings where you can choose the client-certificate lifetime (1–90 days). The kubeconfig is available once the cluster is Active.

What's the difference between Nodes and Instances?

Nodes are the Kubernetes nodes as the cluster sees them (name, role, internal IP, Ready state). Instances are the underlying compute VMs that back those nodes (flavor, image, IPs, volumes) — where you'd power-cycle or inspect a specific VM.

How does auto-scaling work?

Enable autoscaling on a node pool and set its Min / Max nodes. The cluster autoscaler adds nodes up to the maximum when workloads need capacity and removes them when idle. You can also Scale a pool to a specific desired count within its range.

Can I run GPU workloads?

Yes — when creating or adding a node pool, enable GPU acceleration and pick a GPU model and count (on GPU-capable flavors).

Why are Cancel subscription / Delete service greyed out?

Both are blocked while the service still has clusters. Delete every cluster first, then the service-level actions become available — see Service settings.

Are the add-ons managed after creation?

Add-ons chosen at creation (ingress, metrics-server, storage provisioner, monitoring, etc.) are installed automatically. The cluster Overview shows which are enabled.