The kubectl command cheat sheet

Quick reference

Inspect

$kubectl get pods -A -o wide
$kubectl describe pod api-7d9f
$kubectl get deploy api -o yaml
$kubectl get events -A --watch
$kubectl explain deployment.spec.strategy

Logs

$kubectl logs deploy/api --tail=100
$kubectl logs api-7d9f -f
$kubectl logs api-7d9f --previous
$kubectl logs -l app=api --prefix
$kubectl logs api-7d9f --since=1h

Debug

$kubectl exec -it api-7d9f -- sh
$kubectl port-forward svc/api 8080:80
$kubectl debug -it api-7d9f --image=busybox
$kubectl top pod api-7d9f
$kubectl run tmp --rm -it --image=busybox

Apply

$kubectl apply -f k8s/
$kubectl apply -k overlays/prod
$kubectl diff -f k8s/
$kubectl rollout status deploy/api
$kubectl rollout undo deploy/api

Explained

Inspect

kubectl get pods -A -o wide shows every pod with its node and IP. describe pod api-7d9f gives the events at the bottom, which is usually the reason. get deploy api -o yaml shows what the cluster holds, get events -A --watch follows a live incident, and explain settles a field name.

Logs

kubectl logs deploy/api --tail=100 is the first look. Add -f to follow, or --previous for the container that died. logs -l app=api --prefix tags every pod's lines with its name, and --since=1h cuts a noisy service to the window that matters.

Debug

kubectl exec -it api-7d9f -- sh gets a shell. port-forward svc/api 8080:80 reaches the service from your laptop. debug -it api-7d9f --image=busybox adds a container when the image has no shell. top pod shows usage, and run tmp --rm -it --image=busybox gives you a disposable pod inside the cluster network.

Apply

kubectl apply -f k8s/ or apply -k overlays/prod. diff -f k8s/ shows the change before you make it, rollout status blocks until it is done, and rollout undo is the way back.

In one line

Diff before apply

$kubectl diff -f k8s/