Step 01 of 06
Learn the concept
Kubernetes has a reputation for requiring a cloud account and a procurement meeting. Not for this course. kind runs the same API machinery locally, so every mistake costs seconds and disk space, not a surprise line item.
The ideas this is made of
A node is the machine Kubernetes can place work on
In a cloud cluster, a node is usually a virtual machine. In kind, a node is a Docker container running the pieces a Kubernetes node needs, including kubelet and a container runtime. That sounds fake until you remember the Kubernetes API is the same. For learning, it is perfect: real scheduling semantics, zero AWS bill.
The control plane decides; workers run
The API server receives requests. The scheduler chooses a node for unscheduled Pods. Controllers create and repair lower-level objects. kubelet, running on each node, makes the chosen containers exist. A single-node kind cluster may put these roles in one container, but the responsibilities are still separate and the words matter.
kubeconfig tells kubectl which cluster you mean
kubectl reads kubeconfig, usually from ~/.kube/config, to find clusters, users and contexts. A context combines a cluster, credentials and optional namespace. Many production mistakes start as kubectl pointed at the wrong context. Read kubectl config current-context before touching anything expensive. Here, nothing is expensive.
Describe and events are the first smoke trail
kubectl get tells you the headline. kubectl describe shows scheduling decisions, probe failures, image pull errors and recent Events attached to the object. Events are short-lived cluster notes such as FailedScheduling, Pulled, BackOff and Unhealthy. When something is wrong, read them before rewriting YAML.
kubectl config get-contexts
kubectl config use-context kind-beacon
kubectl config current-contextThe command does not create a cluster. It changes which API server future kubectl commands speak to. That difference has saved more jobs than bravery has.
The first commands to reach for
| Command | What it tells you | When to use it |
|---|---|---|
| Names, status, age | Start here |
| Spec, status, events | When status is not enough |
| Container stdout/stderr | When the process started |
| Recent cluster actions | When scheduling or probes fail |
| Target cluster | Before changing anything |
What these are called on the job
kind — Kubernetes in Docker; a local cluster whose nodes are containers.
Node — A machine, virtual machine or kind container that can run Pods.
Context — The kubeconfig entry selecting a cluster, user and optional namespace.
Event — A timestamped cluster observation attached to an object, often the fastest clue.