kubernetes-hands-on/05-pods

The base building block: pod

Introduction

In this section we will learn what is a pod, deploy your first container, configure Kubernetes, and interact with Kubernetes in the command line.

The base job of Kubernetes is to schedule pods. Kubernetes will choose how and where to schedule them. You can also see a pod as an object that requests some CPU and RAM. Kubernetes will take those requirements into account in its scheduling.

But it has a base assumption that a pod can be killed whenever it wants to. So keep in mind that a pod is mortal and it will be destroyed at some point.

First pod

Let's start to deploy the docker image mhausenblas/simpleservice. It's a stateless python JSON API that answers on multiple endpoints. In this hands-on we will only use the /health.

Here is our first manifest for Kubernetes:

apiVersion: v1
kind: Pod
metadata:
  name: simple-pod
spec:
  containers:
  - name: simple-pod
    image: mhausenblas/simpleservice:0.5.0

The manifest of Kubernetes represents a desired state. We do not write the steps to go to this state. It's Kubernetes who will handle it for us. Let's have a look a the fields:

• apiVersion: the version of the Kubernetes API we will be using, v1 here
• kind: what resource this object represents
• metadata: some metadata about this pod, more on it later
• spec: specification of the desired behavior of this pod
  • containers: the list of containers to start in this pod
    • name: the name of the container
    • image: which image to start

Let's apply this manifest to Kubernetes. This will tell Kubernetes to create the pod and run it.

$ kubectl apply -f 05-pods/01-simple-pod.yml
pod "simple-pod" created

We also could have used the kubectl create -f .... But it's better to have a declarative approach in Kubernetes rather than an imperative one, see.

kubectl get

Now list all the pods running in Kubernetes. get is the ls of Kubernetes.

$ kubectl get pod
NAME             READY     STATUS    RESTARTS   AGE
simple-pod   1/1       Running   0          4s

Logs

Let's have a look at the logs of this pod:

$ kubectl logs simple-pod
2018-10-01T09:21:59 INFO This is simple service in version v0.5.0 listening on port 9876 [at line 142]
2018-10-01T09:23:21 INFO /info serving from 172.17.0.4:9876 has been invoked from 172.17.0.1 [at line 101]
2018-10-01T09:23:21 INFO 200 GET /info (172.17.0.1) 1.38ms [at line 1946]

kubectl describe

Our first pod is now running. Now describe it. describe is a get on steroid, with more information.

$ kubectl describe pod simple-pod
[a lot of stuff]

IP: 172.17.0.1

[more stuff]

Look at the information provided. Get the field IP, it's the internal ip for this pod.

Connect to the cluster, and try to curl this ip - 172.17.0.4 in the example.

$ minikube ssh
                         _             _
            _         _ ( )           ( )
  ___ ___  (_)  ___  (_)| |/')  _   _ | |_      __
/' _ ` _ `\| |/' _ `\| || , <  ( ) ( )| '_`\  /'__`\
| ( ) ( ) || || ( ) || || |\`\ | (_) || |_) )(  ___/
(_) (_) (_)(_)(_) (_)(_)(_) (_)`\___/'(_,__/'`\____)

$ curl 172.17.0.4:9876/info
{"host": "172.17.0.4:9876", "version": "0.5.0", "from": "172.17.0.1"}

Kubernetes has a useful add-on, a web dashboard. It's included by default in minikube. You can start it with:

minikube dashboard

Exercises

1. Deploy a pod containing nginx. The image name is nginx, see: https://hub.docker.com/_/nginx/.
2. Do you think you can access the pod simple-service from outside of Kubernetes, without changing the manifest?

Clean up

kubectl delete pod --all

Answers

For 2), no, the pod is only visible from the inside of the cluster.

Links

• https://kubernetes.io/docs/concepts/workloads/pods/pod-overview/
• https://kubernetes.io/docs/concepts/workloads/pods/pod/
• https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle/