Skip to main content

Set up the management cluster

The rest of this will be run from your admin machine. This can be macOS or Linux, basically wherever you can run docker (although we have not tested on Windows, but should be fine).

Kind​

The Pi cluster will be created by CAPI. We'll have the management cluster on your admin machine. You can use a more permanent cluster if you like, but here we are using kind.

Install:

Create a new kind cluster.

kind create cluster --name lm-lab-manage

CAPI and CAPMVM​

The Liquid Metal infra provider is Cluster API Provider MicroVM (CAPMVM).

Create a cluster-api directory under home:

mkdir -p ~/.cluster-api

Set the CAPMVM version in your environment. First check for the version which is compatible with the version of flintlockd your boards are running. In my case I am using Flintlock v0.4.0, so the latest compatible CAPMVM version is v0.7.0.

export CAPMVM_VERSION=v0.7.0
danger

When checking your flintlockd version, be aware of this bug. If you ran the provision.sh without setting either the FLINTLOCK env var or the --version flag on the flintlock subcommand, then you will have the latest version installed.

Write the installation information to a clusterctl config file in the cluster-api repo.

cat << EOF >~/.cluster-api/clusterctl.yaml
providers:
  - name: "microvm"
    url: "https://github.com/liquidmetal-dev/cluster-api-provider-microvm/releases/$CAPMVM_VERSION/infrastructure-components.yaml"
    type: "InfrastructureProvider"
EOF
tip

If you already have a ~/.cluster-api/clusterctl.yaml file, move it to a backup first: mv ~/.cluster-api/clusterctl.yaml ~/.cluster-api/clusterctl.yaml.bak

Use clusterctl to initialise the management cluster.

export EXP_CLUSTER_RESOURCE_SET=true # required for the CNI addon we want deployed to the MicroVM cluster

clusterctl init --infrastructure microvm
Output
Fetching providers
Installing cert-manager Version="v1.5.3"
Waiting for cert-manager to be available...
Installing Provider="cluster-api" Version="v1.2.2" TargetNamespace="capi-system"
Installing Provider="bootstrap-kubeadm" Version="v1.2.2" TargetNamespace="capi-kubeadm-bootstrap-system"
I0927 13:51:15.765771  815920 request.go:665] Waited for 1.023726916s due to client-side throttling, not priority and fairness, request: GET:https://127.0.0.1:38035/apis/bootstrap.cluster.x-k8s.io/v1beta1?timeout=30s
Installing Provider="control-plane-kubeadm" Version="v1.2.2" TargetNamespace="capi-kubeadm-control-plane-system"
Installing Provider="infrastructure-microvm" Version="v0.8.0" TargetNamespace="capmvm-system"

Your management cluster has been initialized successfully!

You can now create your first workload cluster by running the following:

  clusterctl generate cluster [name] --kubernetes-version [version] | kubectl apply -f -
tip

If you see an error containing the words "401 Bad Credentials", make sure you don't have any Github auth methods set in your environment. GITHUB_TOKEN is a common culprit and can be removed with unset GITHUB_TOKEN.