Kubernetes commands
kubectl config get-clusters
to get all clusters in your .kube/config file
kubectl config get-contexts
to get contexts
kubectl config current-context
to print current context
In order to change it use the next command:
kubectl config use-conext <context_name>
To get list of all objects of special kind verb ‘get’ is used:
kubectl get pods
kubectl get deployments
kubectl get secrets
kubectl get services
kubectl get pdb
kubectl get hpa
Additional switches that can be added to the above commands:
-o wide - show more information
--watch or -w - watch for changes
Useful get command:
kubectl get <object_type> <object_name> -o yaml
to print YAML manifest of requested resource
For ex. kubectl get po my-test-pod -o yaml
The first troubleshooting verb is describe, it shows detailed info about resource and latest events related to this resource:
kubectl describe pod <pod_name>
kubectl describe deployment <deployment_name>
kubectl describe secret <secret_name>
kubectl describe service <service_name>
kubectl describe hpa <hpa_name>
kubectl describe pdb <pdb_name>
The second troubleshooting word is log, it shows pod logs:
kubeclt logs <pod_name> - to see all logs that were written by pod until now.
Useful flags:
-f or --follow – to catch new logs in real-time
--previous – to see logs of previous failed container (in case pod keeps failing)
--tail <number> – print
The third verb is exec:
kubectl exec -it <pod_name> bash – this creates new bash session inside container inside pod you specified and provides it for you as standard terminal.
Editing resources on the fly:
kubectl edit deploy <deploy_name>
to edit your deployment manifest on the fly. Can be useful to tests app behavior with special configs, like new /modified env variable, different amount of resource, etc.
kubectl edit hpa <hpa_name>
to edit HorizontalPodAutosacller resource. Can be useful to scale app to the maximum amount of pods.
In order to delete resource:
kubectl delete pod <pod_name>
(it is one of the approaches to restart your application, delete pods one by one, but waiting for the new one before deleting the next one)
kubectl delete deploy <deploy_name>
(in order to stop application on some specific environment) this can’t be restored automatically, only through CI/CD job or using backuped .yaml file.
Another approach to stop app on specific env significantly SAFELY (read as USE THIS OPTION) is:
kubectl scale deployment <deployment_name> --replicas=0
kubectl delete service <service_name>
(in order to delete k8s service together with corresponding LoadBalancer in cloud), traffic to app will be stopped.