kubectl reference guide

5 min read | by Jordi Prats

Generic kubectl usage:

kubectl config view: Create an empty kubeconfig for Pods
kubectl logs previous: Using kubectl logs previous we can get the logs of the already restarted container
kubectl explain: man for Kubernetes: With kubectl we have a sort of man to get the details of default Kubernetes objects
kubectl rollout history: By adding the record flag we can automatically set the change-cause annotation
kubectl set image: With kubectl set image we can update a container's image imperativelly
kubectl: create a yaml base template: By combining dry-run and yaml output we can use kubectl to create yaml templates for creating new resources on the cluster
retrieve yaml manifest for a given kubernetes object: Using kubectl we can retrieve the yaml manifest of any kubernetes object
kubectl top: kubectl top can be use to get a gist of what's going on on our cluster, resource-wise

sorting or filtering

Retrieve objects from a specific apiGroup: When we have the same "kind" on different API we can use the apiGroup to query them
Partially apply a manifest filtering by label: Using a label we can selectively apply a manifest using kubectl
Using labels for filtering pods with kubectl: The selector option for kubectl commands helps filtering pods based on it's labels
sort kubernetes objects using kubectl: kubectl can sort objects by specifying any numeric or string field using it's JSON path

ComponentStatus deprecated in v1.19+: ComponentStatus have been deprecated, we'll need to start using livez and readyz endpoints
status of the API server: Using kubectl get --raw we can get the status of Kubernete's API server
Manually evict a Pod from a node: Using kubectl drain we can evict a Pod from a node
Kubernetes remove taint: Using kubectl taint we can also remove taints
kubernetes: uncordon node: Using kubectl uncordon remove the SchedulingDisabled from a node
kubernetes: drain node: Safely put a kubernetes node into maintenance mode

kubectl patch file: kubectl patch can be used with a patch file to update an existing kubernetes object
kubectl patch: Using kubectl patch we can modify Kubernetes objects in batch mode

kubectl debug: Using ephemeral containers: Using ephemeral containers we'll be able to add containers on the existing Pods for troubleshooting
kubectl port-forward: Using kubectl port-forward we can reach services that are not exposed using other methods
kubectl diff usage: kubectl diff can be used to see the actual changes that will be performed by applying a given yaml file

kubectl convert plugin: Using kubectl convert we can convert manifests between API version
kubectl plugins: Write a custom plugin: We can write a simple kubectl plugin to act as an example for more complex actions

kubectl plugins: krew blame: Using kubectl blame we can identify when (and how) a field has been updated
Check cluster for deprecated objects: Using the krew deprecations we can identify which objects need to be updated before updating the cluster
kubectl get-all plugin: Using the krew plugin get-all we can list all the objects on a Kubernetes cluster
kubectl tree: Using kubectl tree we can visualize relationships between kubernetes objects
kubectl plugins: krew - unused-volumes plugin: kubectl plugin: unused-volumes to list PVC that are not assigned to any pod in the cluster

Autoscale a deployment base on it's Pod CPU metrics: A Kubernetes application can be autoscaled base on it's CPU usage using an HPA
kubectl rollout undo for rolling back releases: Kubernetes: Using kubectl rollout undo we can rollback to a previous state
Update on a deployment triggers a rolling update: When we update a deployment object Kubernetes will trigger by default a rolling update
Scaling deployments using kubectl scale: To be able to scale kubernetes deployments we don't need to edit it's yaml, we can use kubectl scale

kubectl run to create a pod: Let's see how to run a pod on kubernetes and it's equivalent docker commands
Interactive shell in a kubernetes pod: Troubleshooting k8s can be hard if we don't have the right tools, having an interactive shell with the right tools can really speedup incident resolution

How to get events from multiple kubernetes objects: kubectl get events can help us to better understand how the events happen across multiple objects

Set the default namespace for kubectl: If you need to work on a specific namespace you can set the it as your default namespace instead of repeating yourself on each kubectl command
Change kubectl context to use another k8s cluster: By switching kubectl context we can start working on a different cluster

Managing a secret on kubernetes: Let's take a look into how to create and retrieve a secret on kubernetes