Run apps on Kubernetes

$ run as a normal user# run as root (or with sudo)

Skills in this chapter

Kubernetes — 8

kubectl — talk to the cluster

kubectl creates, reads, changes and deletes everything in Kubernetes.

$kubectl get pods -n shop -o wide
$kubectl apply -f deploy.yaml
$kubectl rollout status deploy/shop -n shop
$kubectl logs deploy/shop -n shop --since=10m
$kubectl describe pod shop-7d9c-abcde -n shop

Good practice:

Changes go through Git (GitOps); kubectl on production is for reading and emergencies.
Requests and limits, readiness and liveness probes on every container.

5.2

Helm — packages for Kubernetes

helm installs an app from a chart with your values, and upgrades or rolls it back as one release.

$helm repo add bitnami https://charts.bitnami.com/bitnami
$helm upgrade --install shop ./charts/shop -n shop -f values-prod.yaml
$helm history shop -n shop
$helm rollback shop 3 -n shop

5.3

Kustomize — one base, many environments

kustomize changes plain YAML per environment with overlays — no templates.

$kubectl kustomize deploy/overlays/prod
$kubectl apply -k deploy/overlays/prod
$kustomize edit set image shop=ghcr.io/acme/shop:1.4.0

5.4

minikube, kind and k3s — clusters to practise

A real cluster on your laptop or a small VM, made in a minute and thrown away.

$minikube start --addons ingress,metrics-server
$kind create cluster --name lab
$curl -sfL https://get.k3s.io | sh -

Tool	Best for
`minikube`	learning, add-ons, one laptop
`kind`	CI tests, many clusters, fast
`k3s`	small real servers, home lab, edge

5.5

k9s and stern — find problems fast

k9s is a live terminal UI for the cluster; stern follows the logs of many pods at once.

$k9s -n shop --readonly
$stern -n shop deployment/shop --since 15m -i ERROR

5.6

Debug a pod that does not start

The same 5 steps for every broken pod.

kubectl get pods — the STATUS: Pending, ImagePullBackOff, CrashLoopBackOff, OOMKilled?
kubectl describe pod — the Events at the bottom tell why.
kubectl logs POD --previous — the logs of the container before it crashed.
kubectl get events --sort-by=.lastTimestamp — what happened around it.
kubectl exec -it POD -- sh or kubectl debug — look inside.

Remember

Pending = scheduling (resources, taints); ImagePullBackOff = image name or registry login; CrashLoopBackOff = the app itself or its config.

All commands in this chapter