Skip to main content

Kubernetes

The Kubernetes page, under START in the sidebar, makes a local Kubernetes cluster when you ask for one, and only then. Until you create a cluster, nothing runs and no memory is used, and a stopped cluster uses none either.

A cluster is k3s, the lightweight Kubernetes, running in containers on your engine: a server node, plus worker nodes if you want them. It starts in under a minute.

Creating a cluster​

Click Create cluster… (or New cluster once you have one). The dialog asks for:

  • Name, e.g. dev.
  • Kubernetes version: 1.37, 1.36 or 1.35.
  • Worker nodes: 0 to 3. The server runs your pods too, so 0 is fine for most work. Each node uses about 500 MB of memory while it runs.
  • Ingress port: the port on your computer for the cluster's port 80, where its ingress (Traefik) answers. 8080 unless that's taken.
  • API port: the port of the Kubernetes API, 6550 unless that's taken.
  • Use it with kubectl (on by default): adds the context ferrodock-<name> to your kubeconfig and switches kubectl to it.

Ferrodock downloads k3s if it isn't on your machine yet, starts the nodes and waits until they're ready. The cluster opens when it's up.

Using kubectl and other tools​

With Use it with kubectl on, kubectl, Helm, k9s, Lens and your IDE on this computer reach the cluster as they would any other:

kubectl config use-context ferrodock-dev   # switch back to it later
kubectl get nodes

Ferrodock writes to the file in KUBECONFIG, or ~/.kube/config. It only adds the cluster's own context, user and cluster entries, and keeps a copy of the file as it was next to it (config.ferrodock-backup). Removing the cluster takes its context out again.

Inside a cluster​

Open shows the cluster:

  • Overview: its nodes, and every pod with its state, restarts and node, updated every few seconds. Your own pods are listed first.
  • Terminal: a shell on the server node, with kubectl ready.
  • Load image: copies an image from Docker into every node, so pods can use images you built without pushing them to a registry. Give the pod imagePullPolicy: IfNotPresent (or Never).

The header has the kubectl context (with Copy for the command that switches to it), the ingress address, and the API address.

Reaching your apps​

Port 80 of the cluster, where its Traefik ingress controller answers, is on the ingress port of your computer. Give your app an Ingress and it's at http://localhost:8080 (or the port you picked):

kubectl create deployment web --image=nginx:alpine
kubectl expose deployment web --port=80
kubectl create ingress web --rule="/*=web:80"
# then open http://localhost:8080

Stopping and removing​

Stop (the square on its row) stops every node and frees its memory; Start brings it back as it was, with your workloads and loaded images. Remove (the ✕) deletes the nodes, their network and storage, and the kubectl context, after asking.

tip

A cluster's nodes are containers named <name>-server and <name>-worker-1…, so they show up on the Containers page too.