> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ankra.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Create a Cluster

> Create a Kubernetes cluster on your own cloud account in the Ankra portal, from choosing the provider to the review checks and the first Online status.

This guide walks the create form in the Ankra portal, which is the same for every provider. You end with a running Kubernetes cluster in your own cloud account, connected to Ankra and shown as **Online**. Each provider guide holds only what differs: the credential, the sizes, the locations and the provider's own fields.

| Provider | Ankra managed | Provider managed |
| - | - | - |
| Hetzner | [Hetzner Clusters](https://docs.ankra.ai/guides/hetzner-clusters.md) | - |
| OVHcloud | [OVHcloud Clusters](https://docs.ankra.ai/guides/ovh-clusters.md) | [OVHcloud MKS](https://docs.ankra.ai/guides/ovh-mks-clusters.md) |
| UpCloud | [UpCloud Clusters](https://docs.ankra.ai/guides/upcloud-clusters.md) | [UpCloud UKS](https://docs.ankra.ai/guides/uks-clusters.md) |
| DigitalOcean | [DigitalOcean Clusters](https://docs.ankra.ai/guides/digitalocean-clusters.md) | [DigitalOcean DOKS](https://docs.ankra.ai/guides/doks-clusters.md) |
| AWS | [AWS Clusters](https://docs.ankra.ai/guides/aws-clusters.md) | [Amazon EKS](https://docs.ankra.ai/guides/eks-clusters.md) |

**Ankra managed** means Ankra creates the servers, installs Kubernetes and upgrades it when you ask. **Provider managed** means the cloud provider runs the control plane, and Ankra creates the cluster and its node pools and connects it. Either way the cluster is created in your account and your provider bills you for it. On Proxmox VE and HPE Morpheus it runs on your own infrastructure, with no provider bill. Want to try Ankra without a cloud account? Start a free [playground](https://docs.ankra.ai/guides/cluster-sandbox.md) instead. Already running Kubernetes? [Connect it](https://docs.ankra.ai/guides/import-cluster.md) instead.

## Before you start

* **The `clusters.create` permission.** Without it the page says "Your role does not allow creating clusters." Ask an organisation owner for it. See [Roles and access](https://docs.ankra.ai/guides/roles-and-access.md).
* **A payment method on the organisation.** The review step warns "A payment method is required to create this cluster. Free credit does not replace it." and the create is refused without one. See [Credit and payment method](https://docs.ankra.ai/platform/billing.md#credit-and-payment-method).
* **A credential for the provider**, with the permissions its credential page lists, for example [Hetzner](https://docs.ankra.ai/platform/credentials/hetzner.md) or [AWS](https://docs.ankra.ai/platform/credentials/aws.md). You can also add it from inside the form.
* **An SSH key credential** for the servers. You can add it from inside the form too. See [SSH Key Credentials](https://docs.ankra.ai/platform/credentials/ssh-key.md).
* **The provider enabled for your organisation.** A provider your organisation cannot use is not listed. Opened by its address, the form says "\<Provider> is not enabled for this organisation."

## Create the cluster

> [tab] Portal
>
> **1. Open the create form**
>
> Go to **Infrastructure → Clusters** and click **Add a cluster** (`/next/clusters/new`).
>
> The page **Add a cluster** opens, with four steps across the top: **Infrastructure**, **Configuration**, **Review** and **Provision**.
>
> **2. Choose the provider**
>
> Under **Where should the cluster run?** (**Or create a cluster with a provider** when a free playground is offered), pick the provider card. A card with the **GPU nodes** badge offers GPU sizes.
>
> Where the provider offers both, **Who runs Kubernetes on \<provider>?** follows. Pick **Ankra managed** or **Provider managed**. A provider with only one of them goes straight to the form.
>
> The form **Add a \<provider> cluster** opens. An **Estimated monthly cost** card on the right fills in once you choose a credential and a location.
>
> **3. Name and purpose**
>
> Under **Name and purpose**, enter the **Cluster name**: 3 to 53 lowercase letters, digits and hyphens, starting with a letter. Some provider-managed services allow other characters or a shorter name; the hint under the field says which.
>
> Then answer **What will this cluster run?**: **Production**, **Staging**, **Development**, **Testing**, **Local** or **Something else**. Nothing is chosen for you, and the form will not go to review until you pick one. With **Something else**, name it under **What it runs** and choose **How important is this cluster?**. The answer changes defaults:
>
> * **Development**, **Testing** and **Local** start with one control plane node; **Production** and **Staging** start with three.
> * Ankra never stops a production cluster at night.
> * On UpCloud, the ingress load balancer plan starts larger for a production cluster.
>
> **Description** is optional.
>
> **4. Where it runs**
>
> Under **Where it runs**, choose the **\<provider> credential** - the account Ankra creates the cluster in. With none saved, click **Add a credential**; **Add another credential** adds one more.
>
> Choose the **SSH key**. It is installed on every server, so you can reach it over SSH through the management server. With none saved, click **Add an SSH key**.
>
> Choose the location (called **Location**, **Region** or **Zone**, depending on the provider). Some providers ask more questions here, such as the size of the internet gateway; the provider guide explains them.
>
> The **Control plane**, **Worker nodes** and **Kubernetes** sections appear once the location's sizes are read.
>
> **5. Control plane**
>
> A **Provider managed** form has no control plane or management server to size: the provider runs them. Its provider guide covers the fields it has instead.
>
> Under **Control plane nodes**, pick **Single node** (lower cost, cluster changes pause if it goes down) or **High availability** (three nodes, recommended for production). Choose the **Control plane size**.
>
> Expand the **Management server** row (**Gateway** on OVHcloud) to change its size. It is the cluster's SSH entry point: Ankra and your SSH sessions reach the other servers through it, and they take no SSH from outside the cluster's network.
>
> Each size field offers **Sizes** (**Small**, **Medium**, **Large**, **X-large**) or **Exact server type**, the provider's full list with prices. GPU types are listed in **Exact server type** only.
>
> **6. Worker nodes**
>
> Under **Worker nodes**, each node group is a set of nodes of the same size. Set its **Group name**, **Nodes** and size. **Add a node group** adds another, up to the provider's limit.
>
> * **Placement, labels and taints** adds node labels (**Add node label**) and taints (**Add taint**) and, on OVHcloud, UpCloud and AWS, **Pin to availability zone**.
> * **Scale this group automatically** sets **At least** and **At most** (up to 100) nodes. See [Cluster autoscaling](https://docs.ankra.ai/guides/cluster-autoscaling.md). The box is not offered for Proxmox VE, HPE Morpheus or UpCloud Managed Kubernetes groups: UpCloud Managed Kubernetes refuses autoscaling on a node pool.
>
> **7. Kubernetes**
>
> Under **Kubernetes**, keep **kubeadm** (upstream Kubernetes with the Cilium network plugin, recommended for production) or pick **k3s** (lighter, for small and development clusters). With k3s, a **Network plugin** field appears. Choose the **Kubernetes version**.
>
> The distribution, the network plugin and the etcd layout cannot be changed after create. See [Choices fixed at create time](https://docs.ankra.ai/guides/operate-ankra-managed-clusters.md#choices-fixed-at-create-time).
>
> **8. Cluster configuration in Git (optional)**
>
> Under **Cluster configuration in Git**, tick **Keep the configuration in a Git repository** and choose the **Git connection**, the **Repository** and the **Branch**. Ankra then keeps the cluster's Stacks and add-ons as files in that repository. Only GitHub and Bitbucket Cloud connections are listed.
>
> With no Git connection, the section says "No Git connection is set up, so Ankra keeps the configuration itself." You can add a repository later in the cluster's settings. See [GitOps](https://docs.ankra.ai/concepts/gitops.md).
>
> **9. Advanced configuration (optional)**
>
> Expand **Advanced configuration** for:
>
> * **Private network range** - leave it empty for the provider's default.
> * **Addresses allowed to reach the management server** - IPv4 addresses or ranges, separated by commas. Leave it empty to allow SSH from any address. On Scaleway the field is **Addresses allowed to reach the gateway**, and empty keeps the Scaleway default. On Ankra Cloud the cluster the form creates is IPv6-only, so these IPv4 entries do not apply; see [Ankra Cloud Clusters](https://docs.ankra.ai/guides/ankra-cloud-clusters.md).
> * **Run etcd on separate nodes** (kubeadm only) - pick **3 nodes** or **5 nodes** and the **etcd node size**.
> * **Install the networking stack** - an ingress controller and load balancer, so applications can receive traffic from the internet. On by default.
> * **Add the Kubernetes Gateway API** - shown with the networking stack. See [Gateway API](https://docs.ankra.ai/guides/gateway-api.md).
> * **Give the cluster a public domain** - a domain under `ankra.cc` with automatic DNS records and TLS certificates. On by default. See [Use your own domain](https://docs.ankra.ai/guides/use-your-own-domain.md) for a domain of yours.
> * Provider sections, such as **AWS network** or **UpCloud zone placement**, described in the provider guide.
>
> **10. Review the cluster**
>
> Click **Review cluster**. A field that still needs an answer is listed at the top of the form, for example "Choose what this cluster will run."
>
> The page **Review \<cluster name>** lists the **Settings**, the server count and the cost estimate. Under **Checks before creating**, Ankra runs its checks right away. Where the provider supports it, Ankra asks the provider whether the account can create this cluster.
>
> * "All checks passed." means you can go on.
> * A failed check names what to change. Click **Back to configuration**, fix it, then review again. **Run the checks again** reruns checks that could not run.
> * A missing payment method shows as a warning here; the create itself is refused.
>
> **11. Create the cluster**
>
> Click **Create cluster**.
>
> The page **Creating \<cluster name>** shows three stages: **Create infrastructure**, **Set up Kubernetes** and **Connect to Ankra**. You can leave the page; setup continues. **View operation** opens the operation with every task.
>
> When setup finishes, the page reads **\<cluster name> is ready**. Click **Open the cluster**.

> [tab] CLI
>
> Each provider has its own `create` command, with flags for most fields of the form. OVHcloud's internet gateway size and UpCloud's NAT gateway and load balancer plans have no `create` flag: they are changed after create. The provider guide shows it in full. For example:
>
> ```bash theme={null}
> ankra cluster hetzner create \
>   --name my-cluster \
>   --credential-id <hetzner-credential-id> \
>   --ssh-key-credential-ids <ssh-key-credential-id> \
>   --location fsn1 \
>   --environment development \
>   --control-plane-count 1 \
>   --worker-count 2
> ```
>
> `--environment` takes `production`, `staging`, `development`, `testing`, `local` or a name of your own, and defaults to `development`. See [`ankra cluster`](https://docs.ankra.ai/reference/cli/cluster) for every provider's flags.

> [tab] MCP
>
> The MCP server has one create tool per provider Ankra builds on: `create_hetzner_cluster`, `create_ovh_cluster`, `create_upcloud_cluster`, `create_digitalocean_cluster`, `create_scaleway_cluster`, `create_aws_cluster`, `create_ankracloud_cluster`, `create_proxmox_cluster` and `create_morpheus_cluster`. A provider-managed cluster is created with `create_managed_cluster`. Each needs a token with the `mcp:write` scope and the `clusters.create` permission. See [MCP tools](https://docs.ankra.ai/platform/mcp-tools.md).

## Verify

1. In **Infrastructure → Clusters**, the cluster's status reads **Online**. While it is being set up, it reads **In progress**, and the badge beside its name on its own pages reads **Provisioning**.
2. Point `kubectl` at it through Ankra and check the nodes. This needs a [Cluster Access](https://docs.ankra.ai/guides/cluster-access.md) grant.

   ```bash theme={null}
   ankra cluster kubeconfig add my-cluster --use
   kubectl get nodes
   ```

   Every control plane and worker node is `Ready`.

## What the form does not set

These choices have no field in the form. Use the CLI or the API when you need them at create time:

| Choice | What the form does | Where to set it |
| - | - | - |
| Cloud-init user data for a node group | Sends none | At create time only the API sets it (`node_groups[].user_data`). The CLI sets it on a group added later, with `--user-data-file` on `ankra cluster node-group add`. See [Node group user data](https://docs.ankra.ai/guides/node-group-user-data.md) |
| Cluster-wide availability zones on OVHcloud | Lets Ankra choose; a node group can still be pinned to a zone | `--availability-zones` on `ankra cluster ovh create` |
| Ubuntu series, root volume size and volume retention on AWS | Sends Ubuntu 24.04, a 40 GiB root volume and `retain` | `--ubuntu-series`, `--root-volume-gib` and `--retention-policy` on `ankra cluster aws create` |

## What can go wrong

| What you see | What to do |
| - | - |
| "Your role does not allow creating clusters." | Ask an organisation owner for the `clusters.create` permission |
| "\<Provider> is not enabled for this organisation." | The provider is not available to your organisation. Click **Choose another provider** |
| "\<Provider> locations could not be read with this credential. Check the credential, then try again." | The credential is wrong or lacks permissions. Fix it under **Integrations & alerts → Credentials** |
| "The credential was added, but it is not in the list yet" | Click **Read the list again**. Adding it a second time makes a second copy |
| **The cluster was not created** with "Add a payment method to create this cluster." | Click **Open Plan and billing**, add a payment method, then click **Create cluster** again |
| A message that starts "The 30-vCPU trial limit would be exceeded." | On the Free or Homebuilder plan, until an organisation owner approves paid usage in **Plan and billing**, Ankra refuses to create or grow a cluster it builds when the nodes that run workloads, across all the organisation's clusters, would then have more than 30 vCPUs. Ask an owner to approve paid usage, or choose fewer or smaller worker nodes |
| **Check your clusters before trying again** | The request may have been accepted. Look for the cluster in **Infrastructure → Clusters** before you create it again |
| **Could not finish setting up \<cluster name>** | Follow **How to fix this** on the page, then click **Try again**. Without the permission to run setup again, ask an organisation owner |

## Next steps

* [Operate Ankra Managed clusters](https://docs.ankra.ai/guides/operate-ankra-managed-clusters.md): node groups, control plane, bastion, upgrades, stop and start, terminate.
* [Stacks](https://docs.ankra.ai/concepts/stacks.md): deploy your first add-ons to the cluster.
* [Accessing Clusters with kubectl](https://docs.ankra.ai/guides/kubeconfig.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.