Create a cluster

You create a managed Kubernetes cluster from a single form: pick a name, a network, a storage pool, node sizes, and optionally a set of add-on bundles. Akasha provisions the nodes and installs the selected add-ons for you.

Before you start

You need a network and a storage pool in your project. See About networks; available storage pools are set up by your administrator.

Create the cluster

1. Go to Kubernetes → Clusters and click Create Cluster.
2. Enter a Cluster Name. Lowercase letters, numbers, and hyphens only.
3. Select the Network the cluster nodes attach to. The cluster's API endpoint is a virtual IP allocated from this network's uplink address ranges.
4. Select the Root Disk Pool — the storage pool used for the node root disks.
5. Optionally select Add-ons bundles (see below).
6. If storage is enabled, configure the worker Data Disk Size and Storage Type (see below).
7. Choose the control plane size (CP Size) and CP Replicas — 1 (default) or 3 for high availability.
8. Choose the Worker Size and Worker Count (0–99, default 0). You can add more worker pools later.
9. Click create. The cluster appears with status Provisioning.

Node flavors

Both control plane and worker sizes use flavors:

You can also enter a custom flavor in the format c<vCPU>-m<GB RAM>, for example c4-m16.

Add-on bundles

Bundles pre-install common components right after the cluster is ready:

metrics-server is always installed, regardless of your selection.

Selecting Database or CI & CD automatically enables Storage (those components need persistent volumes) and raises the default sizing to c4-m8 nodes with at least 2 workers.

You can install any of these components individually later from the cluster's Resource tab — see Add-ons.

Worker data disks

When storage is enabled, each worker gets a data disk in addition to its root disk:

• Storage Pool — the pool the data disks are created in.
• Data Disk Size — 20, 50 (default), 100, or 200 GiB, or a custom size such as 75GiB.
• Storage Type — OpenEBS-LVM (default) provisions LVM-backed local persistent volumes; Longhorn provisions distributed, replicated storage across workers. A third type, Raw (unmanaged), is available when adding worker pools and leaves the disk for you to manage.

Provisioning lifecycle

Cluster creation runs as an asynchronous operation. The cluster status moves from Provisioning to Provisioned as nodes boot and join; progress and errors are surfaced on the cluster page and in the operations list. Once the cluster is ready, any add-ons you selected are installed as queued operations — if one fails, the Resource tab shows a Retry install button.

When the status is Provisioned, you can download the kubeconfig and open the web terminal — see Access clusters.

Delete a cluster

Deleting a cluster destroys all of its nodes and every workload running on them. There is no undo.

1. Open the cluster and click Delete cluster in the header.
2. Type the cluster name to confirm.

The cluster status changes to Deleting while nodes and cluster resources are torn down.

Related

• About managed Kubernetes
• Access clusters
• Manage worker pools
• Add-ons