Last modified May 30, 2023

'gsctl create cluster' command reference

gsctl and the REST API are being phased out. We recommend to familiarize yourself with our Management API and the kubectl gs plugin as a future-proof replacement. We provide a migration page to help you with the transition.

The command gsctl create cluster allows to create new Kubernetes clusters.

You can use the command with very simple syntax to create a cluster using default values. Note that several cluster specification details cannot be changed after creation.

In order to configure all details of the cluster according to your requirements, you’ll create a cluster definition first and pass it to the command.

Command line examples

The first and rather trivial example shows how to create a cluster for organization myorg and specifying the cluster name, while leaving all other settings to defaults:

gsctl create cluster --owner myorg --name "Test cluster"

The second example shows how to create a cluster where most or even all configurable details are specified, using a definition file:

gsctl create cluster --file prod_cluster_definition.yaml

Note that command line flags take precedence over values in the definition. This way you can, for example, define a default cluster name in the definition, but set a specific one via the --name flag when applying the same definition several times.

Full argument reference

  • --file, -f: Definition file path. See cluster definition reference for details. The value - means that the definition will be read from standard input.
  • --owner, -o: Name of the owner organization. Overwrites a name given per definition file.
  • --name: Name of the cluster. Overwrites name given in definition file.
  • --release, -r: Specific the workload cluster release version to use. Defaults to the latest active version. See list releases for details on listing available workload cluster releases.
  • --create-default-nodepool: Where node pools are supported (AWS since workload cluster release v10.0.0 and Azure since workload cluster release v13.0.0 ), setting this to false allows to suppress the creation of a default node pool. A default node pools would otherwise be created automatically if no cluster definition is given specifying any node pools details, to get you started quickly.
  • --master-ha: Where supported, this is true by default, which means that the cluster will have three control plane nodes. Available on AWS since workload cluster release v11.4.0 . Set this to false to have only one control plane node in the cluster (recommended only for test clusters).
  • --output: By specifying this flag with value json, the output can be printed in JSON format. This is convenient for use in automation. See JSON output for examples.

Passing the cluster definition via standard input

The --file (short -f) flag accepts the special value - to read a definition from standard input. Depending on your shell you can make use of that in various ways.

For example, you can use a pipe like this:

cat my-cluster.yaml | gsctl create cluster -f -

At least in bash, this syntax allows to include the entire definition in the command:

gsctl create cluster --file - <<EOF
owner: acme
name: Dev cluster
release: 7.1.1

JSON output

Passing flag --output with value json to gsctl create cluster changes the printed output to be formatted as a JSON object.

Example success output:

  "result": "created",
  "id": "f01r4"

When requesting cluster creation with workload cluster release v10.0.0 on AWS, or 13.0.0 on Azure, it is possible to recieve created-with-errors as value for the result key. This indicates some problem with either node pool creation or label attachment.

Example error output:

  "result": "error",
  "error": {
    "kind": "unknown",
    "annotation": "Unauthorized",
    "stack": [
        "file": "/go/src/giantswarm/gsctl/commands/create/cluster/v5.go",
        "line": 140
        "file": "/go/src/giantswarm/gsctl/commands/create/cluster/command.go",
        "line": 632