Kubernetes Cheatsheet

For instance, we’ll consider placeholder name as revoir-service, format could be json|yaml|wide|name

Installation

create a new deployment/service from yaml definition

kubectl create -f revoir-service.yaml

post creation you can watch pods coming up

kubectl get pods -w

To get deployment details as yaml

kubectl get deploy/revoir-service -o yaml

Debugging

To view logs, port-forward you don’t need specific pod, you can do the same with deployment, it’ll use a random pod

• --timestamps useful as you want to associate logs with time
• --tail=lastN or --since=10s/10m/10h

  kubectl logs -f deploy/revoir-service --timestamps --tail=1
  

You can filter resources like pod/deployment/service/… with label

kubectl get pods -l 'release=revoir-service'

Forward a port from k8s pod 8080 to localhost 8090

kubectl port-forward deploy/revoir-service 8090:8080

Admin Commands

get top [nodes|pods] resource consumption

• cpu|memory can be passed for sort-by

  kubectl top nodes --sort-by='cpu'
  

  if you’re on k8s version older than 1.18, to deal with sort issue pipe to sort

  kubectl top nodes --sort-by='cpu' | sort --reverse --key 3 --numeric
  

  once you’ve found top nodes, to get pods specific to a node

  kubectl get pods --all-namespaces -o wide --field-selector spec.nodeName=gke-cluster-node-name
  

  scale replicas for a deployment

  kubectl scale deployment/revoir-service --replicas=4
  

  if max replicas limit is reached then this command wouldn’t have any effects beyond the limit

rolling restart of all pods in deployment

kubectl rollout restart deploy/revoir-service

Official k8s cheatsheet