Skip to content

4. Deploy the cluster

The profile is the recipe. Now you point it at the machine.

Create the cluster

  1. Left-hand menu → Clusters.
  2. Click Add New Cluster.
  3. Choose Edge Native as the environment, then Start Edge Native Configuration.

  4. Basic information:

    Field Value
    Cluster name dane-first-cluster
    Description optional
    Tags optional

  5. Next.

Pick your profile

  1. Select dane-agent-basic from the list.
  2. Confirm the version shown is the one you just saved.
  3. Next.

Confirm the variables

Palette now shows the profile's layers, and — because you declared them — a panel for the profile variables, pre-filled with your defaults:

Variable Value
Pod CIDR 100.64.0.0/18
Service CIDR 100.64.64.0/18

This is the payoff from step 3. Someone deploying this profile at a customer with a conflicting network changes the value here, once, and both packs follow — no YAML editing.

Leave the defaults as they are. Next.

screenshot — cluster creation showing the Pod CIDR / Service CIDR variable prompts

Assign the node

  1. On the node configuration step, you will see a node pool — a group of machines with the same role. You need one.
  2. Name it master-pool and make sure it is marked as a control plane pool.

  3. Under Edge Hosts, click to add a host and select edge-06396f804bc511eb80001ceec94be53d.

    Check the ID character by character — two other unclaimed hosts sit in this project and neither will build a working cluster. See step 1.

    Nothing in the list?

    Either you are in the wrong project, or the profile's cloud type is not Edge Native, or the node is already claimed by another cluster. Re-check step 1.

  4. Next.

One node or three — never two

Agent Mode supports a single node, or three or more. A two-node cluster is not a supported configuration: Kubernetes needs an odd number of control-plane members to break ties when deciding what is true, and two machines cannot. You have one node, which is fine.

Review and deploy

  1. Check the summary: three layers, one node pool, one host, your two variables.
  2. Click Finish Configuration / Deploy.

The cluster is now building. Nothing more to click.

What happens next

Expect 15–25 minutes. Palette will:

  1. Push the profile down to the agent on your node
  2. Install Kubernetes
  3. Install Cilium
  4. Report the cluster healthy

The status will move through Provisioning → Configuring → Running. Seeing errors scroll past in the event feed during this window is normal — components start in a dependency order and retry until what they need exists.

Do not intervene early

The single most common mistake is deciding it has failed at minute six and deleting it. Give it the full 25 minutes before concluding anything. Retry loops are how this is supposed to look.

Next: confirm it worked →