devops/materials/02-k8s-objects-and-getting-started.md

id	title
mat-02-k8s-objects-and-getting-started	Kubernetes Objects & Getting Started Locally

Kubernetes objects and getting started

Declarative model (desired state)

Kubernetes is declarative:

• you declare the desired state in a manifest (YAML)
• you post it to the API server
• Kubernetes stores desired state
• controllers continuously reconcile current state to desired state

Example scenarios:

• desired replicas = 10, 2 fail → controller schedules 2 new replicas
• change replicas from 3 to 6 → controller schedules 3 more replicas
• update app version from 1.0 to 2.0 → controller rolls out 2.0 gradually and rolls back if needed

Pods

A Pod is the atomic unit of scheduling in Kubernetes.

• simplest use: run one container per pod
• advanced: multiple containers in one pod share network stack (same IP), volumes, IPC/shared memory

Important:

• scale by increasing the number of pods, not containers inside a pod
• pod lifecycle is “brutal”: pods are replaced (immutable); they do not “come back” after death

Deployments (controllers)

In practice you rarely deploy pods directly. You use higher-level controllers such as Deployments to get:

• self-healing
• scaling
• rollouts/rollbacks

Common controller names you may encounter:

• Deployment
• DaemonSet
• StatefulSet

Services

Pods are ephemeral (they can and will die). Services provide a stable facade:

• stable DNS name
• load balancing
• “service discovery” style routing to healthy pods

Getting Kubernetes locally: installers overview

Kubernetes is complex (control plane, etcd, kubelet, networking…). Installers help.

Examples of installers/distributions:

• K3s
• K3d (runs K3s in Docker)
• kind (Kubernetes in Docker)
• MicroK8s
• Minikube
• Docker Desktop (Windows/Mac)

kubectl

kubectl is the CLI tool for controlling clusters via the API server. Version rule of thumb:

• kubectl should be within one minor version of the cluster control plane. Config:
• uses local kubeconfig file: ~/.kube/config
• supports switching between clusters

K3d quick start

Create a cluster:

• k3d cluster create mycluster

Delete it:

• k3d cluster delete mycluster

Example verification commands:

• kubectl get nodes
• kubectl get pods --all-namespaces

K3d complex setup

Example: 3 server nodes and 5 agent nodes:

• k3d cluster create complex -s 3 -a 5

K3d also supports a config file (useful for versioning and sharing setup).