DevOps, CI/CD, Infrastructure as Code, Cloud

Shopping list CRUD Application.

We will build a Shopping list CRUD Application.The back-end server uses Node.js + Express for REST APIs. MongoDB is used for the persistence layer.

List of all the work performed (briefly, describing features and bonus tasks).

• Develop CRUD app: create, update, retrieve, delete an item
• CI/CD pipeline using: github, gitlab, gitlab runner, terraform, aws ec2 instance
• Iac using Ansible, Vagrant and Gitlab [work in progress...]
• Kubernetes cluster
• Istio [to do]

Instruction

Download and install MongoDB Community Server.

Clone the repository

git clone git@gitlab.com:bngom/shopping-list.git

From the root directory of the project run:

npm install

Lauch tha application

npm start

• Usage

Create an item in the shopping list

curl --location --request POST 'http://localhost:8080/api/item' \
--header 'Content-Type: application/json' \
--data-raw '{
    "name": "Pasta",
    "quantity": 2,
    "description": "Panzani"
}'

Get all created items

curl --location --request GET 'http://localhost:8080/api/item'

Get item by id

curl --location --request GET 'http://localhost:8080/api/item/<ID>'

Update an item

curl --location --request PUT 'http://localhost:8080/api/item/<ID>' \
--header 'Content-Type: application/json' \
--data-raw '{
    "name": "Pasta",
    "quantity": 3,
    "description": "Adja****"
}'

Delete one item

curl --location --request DELETE 'http://localhost:8080/api/item/<ID>'

Delete all items

curl --location --request DELETE 'http://localhost:8080/api/item/'

Testing

npm test

CI/CD pipeline

Github -> GitLab <- AWS EC2 Instance (Gitlab Runner)<- Terraform

Prerequisites

Install Terraform

Install aws-cli

Configure aws access keys and secrets by running aws configure.

Here how to get access key and secret from aws.

CI/CD Pipeline

Review the pipeline configuration file .gitlab-ci.yml. In this file we defined the different stages:

test: when develop branch is pushed it trigger the unittest execution.

deploy: when we push changes on uat branch it triggers the deployment of the app stack to the ec2 instance where the runner is installed.

Review the file docker-compose.yml

Deploy an EC2 instance with terraform (T2.micro free tier)

Move to the folder iac. The instance will be deployed on us-esat-2 region. Feel free to change the region in the file gitlab-runner-instance.tf.

The bootstrap script ./iac/user_data.sh will automatically deploy packages (docker, docker-compose, gitlab-runner...) and register the runner. in that script replace the token in the Register gitlab-runner step by your own. To get the token, go on your project in gitlab:

1. Go to Settings > CI/CD and expand the Runners section.
2. Note the token.

• In case you change the region make sure to change the ami (ami-07efac79022b86107). It differ from one region to another.

• We deploy the instance using an existing aws key-pair named ec2-p2, generate your own on aws (Service > EC2 > Key pairs > Create Key pair) and update the ./iac/gitlab-runner-instance.tf file.The key must be created in the same region of deployment.

Initialize

terraform init

Check the deployment plan

terraform plan

Provision an EC2 instance.

terraform apply

Check on AWS if the instance is correctely provisionned

Note the public ip in our case 18.216.168.169

ssh connect to the EC2 instance

ssh -i "ec2-p2.pem" ubuntu@EC2_PUBLIC_IP

Complete the gitlab-runner installation by adding gitlab-runner ALL=(ALL) NOPASSWD: ALL at the end of the sudoers file:

sudo nano /etc/sudoers

Go to the uat branch

git checkout uat

Trigger the deployment on the ec2 instance

git push

Connect throught ssh on the ec2 instance and check the running containers

Provision the VM with Ansible and Vagrant

• Provisioning servers with Vagrant and VirtualBox.
• Provisioning applications with Ansible Playbooks.

Disable-WindowsOptionalFeature -Online -FeatureName Microsoft-Hyper-V-All

vagrant box add ubuntu/trusty64

vagrant plugin install vagrant-vbguest

move to iac folder

...

Run the vagrant up command.

Server is running at 20.20.20.2:8080

You can connect through ssh ssh vagrant@127.0.0.1 -p 2222. The password is vagrant

Build Docker image

Review the Dockerfile

Build the docker image

docker build -t shopping-list-v1 .

Push docker image to docker registry

docker tag shopping-list-v1 230984/shopping-list-v1:1.0.0
docker push 230984/shopping-list-v1:1.0.0

Docker compose

Review the docker-compose.yml file

docker-compose up

Open a browser on http://localhost:8080 to see the application

docker ps

• shopping-list_web: which represents our application
• mongo: which represents the persistence layer docker

Test the local deployment

Container orchestration using Kubernetes

The application tier

webapp.yml

Our application component will consist of a Deployment and a Service

• Deploymnent: Creates and runs the webapp Pod.

• Service: Allow Pods be accessible to other Pods or users outside the cluster.

The database tier

Review the resource definitions mongo.yml. The database component consists of a PersistentVolumeClaim, a Service and Deployment definitions.

• PersistentVolumeClaim: Allows obtaining persistent storage volume.

• Service: Without a type field, Kubernetes assigns it the default type ClusterIP. Thus the Pod accessible from within the cluster and not from outside.

• Deployment: In the deployment we define a storage volume named storage, which references the PersistentVolumeClaim. The volume is referenced from the volumeMounts field which mounts the referenced volume at the path (/data/db). This is the path where MongoDB saves the data.

Deploy the application

From the root directory of the project run:

kubectl apply -f k8s

Check the deployments

kubectl get deployments

Check the Pods

kubectl get pods

Check the services

kubectl get services

Note that the webapp is exposed at port 32082

Get the ip of the cluster run minikube ip command

Open a browser at http://172.19.7.240:32082

Now test the api

In case we delete the mongo pod the data are not lost due to the persistent volume claim.

You can try to delete a pod kubectl delete pod <POD_NAME>.

Test the api with a get request. The data still remain

Service mesh using Istio

Clean up

minikube delete

minikube start

Deploy istio stack

kubectl apply -f istio/1-istio-init.yaml

kubectl apply -f istio/2-istio-minikube.yaml

Setup a username and passphrase for kiali.

kubectl apply -f istio/3-kiali-secret.yaml

Look at the file and notice that the username and password for kiali is just a base64 encoded string. The login is admin and the password admin

$ echo YWRtaW4= | base64 -d
admin

To allow istio inject sidecars foeach pod deployed, label the default namespace

kubectl apply -f istio/4-label-default-namespace.yaml

Deploy the application

kubectl apply -f k8s/mongo-pvc.yaml
kubectl apply -f k8s/mongo-service.yaml
kubectl apply -f k8s/mongo.yaml

Canaries Releases

kubectl apply -f istio/5-webapp-canary.yaml

kubectl apply -f istio/6-istio-rules.yaml

Kiali is running on port 31000

get the ip of your cluster

minikube ip

Open a browser at http://<YOUR-CLUSTER-IP>:31000

Canaries Release with KIALI UI

Distributed tracing with JAEGER

Comming soon...Need to add to some logic to propagate header x-request-id over the data plane.

Some metrics with graphana

Open a browser at http://<YOUR-CLUSTER-IP>:31002

Author

Barthelemy NGOM

barthe.ngom@gmail.com