1-CloudNative

Cloud native workshop

Setup GCP

1. Enable following Google Cloud APIs in your GCP project. Use the Big Data project that we use during classes

* Compute Engine API
* Cloud Monitoring API
* Cloud Logging API
* Kubernetes Engine API
* Cloud Resource Manager API
* Identity and Access Management API  

2. Create serviceaccount for terraform in the project. Download json keys

Terraform

1. Clone this repo. Place downloaded json key file inside the 1-CloudNative/terraform dir. Run an interactive session into the container based on image you created in step 0. Mount your container home dir (should be VOLUME) to the directory that contains this repo

sudo docker run -it -v "<PATH_TO_REPO_ON_HOST>:<PATH_TO_CONTAINER_HOME>" cloud-native

2. Enter terraform folder. Initialize terraform. Examine terraform.tfvars and modify them. Make sure that the project_id corresponds to your GCP project ID! Run terraform plan with json keys.

cd 1-CloudNative/terraform
terraform init
GOOGLE_APPLICATION_CREDENTIALS=<path_to_your_json> terraform plan

3. Examine the plan created by terraform. Run terraform apply with json keys. Answer yes to the question - your cluster should be ready in approx 10 minutes.

Kubernetes - first steps

1. Copy the generated Kubernetes config to access the cluster

mkdir -p ~/.kube
cp config ~/.kube/config

2. Inspect the api-resources and objects deployed in the fresh cluster

kubectl api-resources
kubectl get namespaces
kubectl get pods -n kube-system

3. Create new namespace for out sample app

kubectl create namespace sample-app

4. Inspect the objects (deployment, service). See how they are meant to interact with each other

5. Apply the objects in the namespace. Inspect the running pods.

kubectl apply -f service.yaml -n sample-app
kubectl apply -f deployment.yaml -n sample-app

kubectl get pods -n sample-app

6. Inspect the service. What kind of IP address was assigned to it? Connect to the pod using port-forward. You should be able to curl inside

kubectl get services -n sample-app
kubectl get pods -n sample-app
kubectl port-forward pod/<pod_name> 8080:8080 -n sample-app &
curl localhost:8080

Kubernetes - operations

1. Remove the service and modify it to be LoadBalancer. Wait a moment for Google Cloud LoadBalancer to make its magic. What kind of IP address was assigned to it? Type the address in your browser. Can you access the app?

EDITOR=nano kubectl edit service/example -n sample-app
kubectl get services -n sample-app

2. Edit the running deployment. Increase number of replicas (eg, 10). Wait for the pods to spawn. Why it took so long? Mind the resources section. See the nodes of your cluster. What happened? Check, which pod is running on which node

EDITOR=<editor_of_your_choice> kubectl edit deployment/example -n sample-app
watch kubectl get pods -n sample-app
kubectl get nodes
kubectl get pods -o wide

3. Perform several maintenance operations: describe pods, get the yaml object dump, watch container logs, exec into the running container

kubectl get pods -n sample-app
kubectl describe pods/<pod_name> -n sample-app
kubectl get pods/<pod_name> -o yaml
kubectl logs <pod_name> -c <container_name> -f
kubectl exec -it <pod_name> -c <container_name> /bin/bash

4. Then, bump a version to 2.0. Watch how rolling update performs its work

EDITOR=<editor_of_your_choice> kubectl edit deployment/example -n sample-app
watch kubectl get pods -n sample-app

Helm

1. Enter the helm folder containing the KubeInvaders chaer

cd ~/1-CloudNative/helm

2. Create new namespace for KubeInvaders:

kubectl create ns kube-invaders

3. Inspect a helm chart. Install kube invaders from the helm chart, set config to point to your sample-app. Inspect the template to see what was generated

cd helm
helm install kubeinvaders . -n kube-invaders

4. Due to bug in kubeinvaders (or rather - our lack of domain) we have to wait for loadbalancer to establish new IP address, and then update the installment using this ip address. Edit the values.yaml file with the LoadBalancer IP and then upgrade the helm deployment

helm upgrade kubeinvaders . -n kube-invaders

5. Play kube-invaders. At the same time, observe how your application survives the fatality. Congratulations! You know how to deploy and manage a simple kubernetes app! Consider installing some helm/stable chart (wordpress? minecraft?) to test your cluster on software you know.

Cleanup

1. Move back to the terraform. Execute terraform destroy, and accept changes. All the changes we made to your GCP project during this workshop should disappear in a couple of minutes.

2. Stop the container, remove it. Keep the image and terraform files, we will use them in the next workshop

Extras and where to go next in Kubernetes world:

Three words for each:

• Istio
• Operators
• KNative
• Kubeflow