How to join worker nodes with a custom configuration¶

When creating a Canonical Kubernetes cluster you may need to join a worker node with a configuration that differs from the default. For example, the worker node may need to use alternative certificates for security reasons or the worker node may have specific networking requirements that must be configured at node creation. Passing extra command line arguments or a configuration file at cluster join allows you to modify the configuration of your worker node.

Prerequisites¶

This guide assumes the following:

A working Kubernetes cluster deployed with the k8s snap

Configuration options¶

Command line flags¶

To discover the configuration options available as command line arguments, on the control node run:

sudo k8s join-cluster --help

Configuration file¶

More configuration options are available when a configuration file is specified. Please consult the reference page for all of the available configuration options and their defaults.

Command line arguments¶

In this example, the name of the new worker node joining the cluster is specified through command line arguments, rather than defaulting to the hostname of the worker machine.

A join token must be generated on the control node of the cluster. Specify the join token is for a --worker node. The name given for the worker node in this example is custom-worker which is not the default hostname.

sudo k8s get-join-token custom-worker --worker

On the new worker machine, install the snap:

sudo snap install k8s --classic --channel=1.32-classic/stable

Join the cluster with the token generated from the output of the get-join-token command and specify the --name we want the worker node to be called. This must match the name used in the get-join-token command.

sudo k8s join-cluster --name custom-worker <JOIN-TOKEN>

After a few moments, the node should have joined the cluster with a success message. Verify the node has joined the cluster with the custom name by switching to the control node and running:

sudo k8s kubectl get nodes

The output should list the custom-worker node in a Ready state.

Configuration file¶

In this example, the configuration file provided at cluster join will set the proxy mode of the worker machine to ipvs.

A join token must be generated on the control plane node of the cluster. Specify the join token is for a --worker node. The name given for the worker node in this example is worker-machine which is the default hostname.

sudo k8s get-join-token worker-machine --worker

On the new worker machine, install the snap:

sudo snap install k8s --classic --channel=1.32-classic/stable

Create a custom_config.yaml file that sets the intended custom configurations.

cat <<EOF > custom_config.yaml
extra-node-kube-proxy-args:
    "--proxy-mode" : "ipvs"
EOF

Join the cluster with the token generated from the output of the get-join-token command and the custom_config.yaml file.

sudo k8s join-cluster --file path/to/custom_config.yaml <JOIN-TOKEN>

After a few moments, the node should have joined the cluster with a success message. Verify the node has joined the cluster by switching to the control node and running:

sudo k8s kubectl get nodes

The output should list the worker-machine node as in a Ready state.

Also verify the proxy mode configuration has been applied to the worker node by checking the logs of kube-proxy on the worker machine:

sudo journalctl -u snap.k8s.kube-proxy | grep ipvs

The output should show the proxy-mode is ipvs.