Ahoy

An exercise in docker and kubernetes

Introduction

The point of this workshop is to familiarize yourself with docker and kubernetes
This repository includes a small golang chat server, with an embedded frontend.
Your goal is to get this server to run in a cloud environment.
If you feel lost or confused, don't hesitate to ask or look for solutions online.

Prerequisites

• docker brew install docker
• kubectl brew install kubernetes-cli
• aws brew install aws-cli

Fork this repository, and work from there

0. Getting aboard (Running locally)

The chat server relies on a connection to a redis server in order to send messages.
We can use the public docker image for redis to test out the server locally: https://hub.docker.com/_/redis

TASK: Start a redis server locally and expose it's port, so you can try out the chat server

you can start the chat server with go run main.go

1. Untying the lines (Building a docker image)

Our next order of business is to get our go server to run in a docker container.
For that we will need to build a docker image, so that we have something to base our container on.

TASK: Write a dockerfile, that builds the go server

You can build a docker image from a Dockerfile with docker build .

2. Pulling the anchor (Running in docker compose)

A solution for orchestration when running docker containers locally is docker compose
We can use docker compose to define and run multiple containers in a single file rather than having to run complicated docker run commands all the time.

TASK: Write a docker-compose.yml file and run the chat and redis servers with docker compose up

BONUS: What do the following terms mean in docker: image, container, repository, tag and registry

3. Hoisting the flag (Running in kubernetes)

Intro to kubernetes

Kubernetes is a great solution for orchestration in the cloud.
Similarly to docker compose, it allows us to define a set of containers we want to run.
The difference is that we don't have to manually tell kubernetes to start the containers.
We simply give kubernetes our manifest, and kubernetes will continuously try to make sure that the containers are running somewhere on some maching.
Kubernetes does not operate directly on containers, though. Instead it manages what it calls Pods.
Pods are a collection of containers along with some metadata attached to them.

Getting access to the ahoy cluster

replace TEAMNAME with your team names in the commands below

$ lpass show Shared-WISEflow-Development/ahoy/TEAMNAME.csv --notes | aws configure import --csv file:///dev/stdin
Successfully imported 1 profile(s)
$ aws eks update-kubeconfig --name ahoy --alias ahoy --profile TEAMNAME --region eu-west-1

You can verify that you have access by running

$ kubectl get namespaces

Examples

A simple manifest with a pod could look like this:

# manifest.yaml
apiVersion: v1
kind: Pod
metadata:
  name: my-pod
  namespace: my-namespace
spec:
  containers:
    - name: alpine
      image: alpine:3.16
      command:
        - echo
        - "hello, mom"

In order to give it to kubernetes we can use kubectl like so

$ kubectl apply -f manifest.yaml
pod/my-pod created

And kubernetes will then schedule a pod on a machine in the cluster.

We can get the status of our pod with

$ kubectl get pod --namespace my-namespace
NAME      READY   STATUS             RESTARTS     AGE
my-pod    0/1     CrashLoopBackOff   1 (8s ago)   11s

Notice that the pod has the CrashLoopBackOff status. This happens when kubernetes continuously restarts the containers in the pod, but they keep crashing.
The reason the container crashes is of course the entrypoint of the container finishes immediately, so the container dies.

We can get the logs of the pod with

$ kubectl logs my-pod --namespace my-namespace
hello, mom

Where we can see that the container indeed did as we asked.

Exercise

TASK: Write a pod manifest containing a redis and a chat server, and apply it to kubernetes and see that the pod is running

Tip: You can test the application by port-forwarding to the pod with kubectl port-forward [POD] -n [NAMESPACE] localport:containerport

You'll have to push your docker image to a docker registry. You can use ttl.sh simply with

docker build . -t ttl.sh/my-image-with-some-id:2h
docker push ttl.sh/my-image-with-some-id:2h

and now that image will be available to other clients to pull with

docker pull ttl.sh/my-image-with-some-id:2h

Note that you only have write access to the namespace of your teams name

4. Setting sail (Getting resilient)

Scaling up

When we apply a pod directly to kubernetes. It will be scheduled to a node, and it cannot be updated or moved to a different node after that.
So let's say that the node in our cluster fails for some reason. Kubernetes can't reschedule the pod, and the server will not be run.
For that we will need another resource that can schedule new pods. One of these resources is called a Deployment.

An example deployment can look like this

apiVersion: apps/v1
kind: Deployment
metadata:
  name: my-deployment
  namespace: my-namespace
spec:
  replicas: 2
  selector:
    matchLabels:
      app: my-deployment
  template:
    metadata:
      labels:
        app: my-deployment
    spec:
      containers:
        - name: alpine
          image: alpine:3.16
          command:
            - echo
            - ${GREETING}
          resources:
            limits:
              memory: "128Mi"
              cpu: "500m"
          env:
            - name: GREETING
              value: "Hello, dad"

Kubernetes will take that deployment and try to maintain 2 replicas of the pod specification defined in it's template. We can also update the pod spec, now and kubernetes will roll out the new version.

TASK: Run your pods as a deployment

The redis question

The problem with including redis in every pod is that the servers cannot share their data source.

TASK: Run redis and the chat-server as seperate deployments, and connect the chat-server to redis through a kubernetes service

BONUS: Run multiple chat-servers on the same redis

5. AHOY! (Exposing the application)

TASK: Expose to the internet with an ingress

Nginx ingress controller, cert-manager and external-dns are available in the cluster. external-dns has access to ahoy.wiseflow.io and all it's subdomains