- Kubernetes Techniques
- Looking at Log files
- Executing commands in a container
Sometimes things don't work as they should in your deployments, and you'd like to take a closer look to debug issues or understand what's going on. There are 3 techniques I use from a day to day basis when I work with kubernetes. Let me show you what these are.
When things are not deploying as expected, or things seem to be taking a while, I describe the deployments and pods associated with the deployments to look for errors.
Let's run the helloworld application that is bundled with this section by typing kubectl create -f helloworld-with-bad-pod.yaml.
As it's starting up, we can run a kubectl get deployments and a kubectl describe deployment bad-helloworld-deployment.
We notice that we have 0 available pods in the deployment that signals that there is something going on with the pod.
If we introspect pods with a kubectl get pods, we see that the bad-helloworld-deployment pod is in an image pull backoff state and isn't ready.
Describing the pod with kubectl describe pod bad-helloworld-deployment-7bb4b7466-f6nkm, will show me that kubernetes is having trouble pull the pod from the repository, either because it doesn't exist, or because we're missing the repository credentials.
Another technique I end up using a lot to track pod progress is looking at the log files for a pod. If you write your logs to standard out, you can get to them by the command kubectl logs <pod_name>. This will return the log statements that are being written by your application in the pod.
Finally, sometimes it is necessary to exec into the actual container running the pod to look for errors, or state. To do this, run the exec command kubectl exec -it <pod-name> -c <container-name> /bin/bash where -it is an interactive terminal and -c is the flag to specify the container name. Finally we want a bash style terminal.
This drops us into the container, and we can introspect into the details of our application.