This project deploys a working infrastructure as code at AWS to deploy a service that accepts as input a text file, and produces as output a count of the words present in the file.

The code base is Python with FastAPI for the service and Terraform with AWS EKS for the infrastructure.

In the service folder is the complete service implementation with Python and its Dockerfile for testing. In the infra folder is the infrastructure as code using Terraform.

The service is build using FastAPI and there is only one endpoint /wc/ and it only accepts POST requests with a text file.

Create a virtualenv, install the dependencies from requirements.txt then run pytest:

$ cd service
$ python -m virtualenv venv
$ source venv/bin/activate
(venv) $ pip install -r requirements.txt
(venv) $ pytest
(venv) $ gunicorn app.main:app -w 4 -k uvicorn.workers.UvicornWorker

The above command to run using gunicorn exposes the default port, check the service in localhost:8000/docs.

$ sudo docker build -t word-count:v1 .

$ sudo docker run --name word_count --publish 80:80 word-count:v1
$ sudo docker exec -it word_count pytest

The above command run using an optimized config for gunicorn and exposes the port 80.

There is also an image in the dockerhub: zsinx6/word-count if needed.

The Docker image is based in the tiangolo/uvicorn-gunicorn-fastapi image, which is an optimized FastAPI server in a Docker container, and auto-tuned for the server based in the and number of CPU cores.

There is an automatic interactive API documentation (provided by Swagger UI) in the /docs. In this documentation is possible to attach a file and make a POST request against the API to manually test the service, and shows all the information about the API.

There is also the alternative automatic documentation (provided by ReDoc) in the /redoc.

First create a file with some text, then make a POST request in /wc/:

$ echo "two words" > testfile.txt
$ curl -X POST "http://localhost/wc/" -H  "accept: application/json" -H  "Content-Type: multipart/form-data" -F "[email protected];type=text/plain"

The response for the above request is:

{"word_count":2}

For the above test to work, the Docker container must be running, check here. If don't want to use Docker, then add the port 8000 in the curl command (e.g. http://localhost:8000/wc/).

The deploy of the infra is in two phases, first we deploy the Kubernetes cluster in the AWS EKS, along with all the needed AWS services (vpc, security groups), then we deploy the service inside the EKS cluster (using the Docker image build from the word-count service folder's Dockerfile).

In the deploy-cluster folder is the workspace for the EKS cluster:

• vpc.tf provisions a VPC, subnets and availability zones using the AWS VPC Module.
• security-groups.tf provisions the security groups used by the EKS cluster.
• eks-cluster.tf provisions all the resources (AutoScaling Groups, etc...) required to set up an EKS cluster using the AWS EKS Module.
• outputs.tf defines the output configuration.
• versions.tf sets the Terraform version to at least 0.14. It also sets versions for the providers used in this project.
• kubernetes.tf the Kubernetes provider is included in this file so the EKS module can complete successfully. Otherwise, it throws an error when creating kubernetes_config_map.aws_auth.

In the deploy-wc-kubernetes folder is the workspace for the service to run inside the EKS cluster:

• kubernetes.tf the Kubernetes provider, which deploys the wc service using the Docker image (zsinx6/word-count) created for the word-count service hosted here, there is also a LoadBalancer, which routes the traffic from the external load balancer to pods matching the selector (in this case, the service). The output is defined to return the external ip address of the service.

Warning: there could be charges when running this, since the deploy is at AWS

Since the project uses AWS, it is expected to have the aws-cli already installed and configured (see here for detailed information).

$ cd infra/deploy-cluster
$ terraform init

$ terraform apply

The above step should take approximately 20 minutes, and when finished the EKS cluster will be available and the Terraform will print the outputs defined in the outputs.tf.

$ aws eks --region $(terraform output -raw region) update-kubeconfig --name $(terraform output -raw cluster_name)

Now the kubectl tool should work.

$ cd infra/deploy-wc-kubernetes
$ terraform init

$ terraform apply

Now the deploy is complete, and the word-count service should already be available, the above command outputs the ip address for the service.

To check the running services (this commands also display the external-IP):

$ kubectl get services

There are 2 ways to upgrade the running image:

For both we need to change the service source code, rebuild the Docker image using an updated tag for the version and push the image to a registry (e.g. dockerhub). The first one is the recommended, since we can use the kubectl directly.

Then we can run the following command to update the deployment inside the EKS cluster:

$ kubectl set image deployments/scalable-word-count word-count=zsinx6/word-count:v2

Or change the infra/deploy-wc-kubernetes/kubernetes.tf to use the updated image, then apply the terraform:

resource "kubernetes_deployment" "wc" {
  # ...

  spec {
    # ...

    template {
      # ...

      spec {
        container {
          image = "zsinx6/word-count:v2"
          # ...
      }
    }
  }

  # ...
}

$ terraform apply

There are also 2 ways to scale, the first one is the recommended.

Just use the kubectl:

$ kubectl scale deployment/scalable-word-count --replicas=4

Or change the number of replicas inside the infra/deploy-wc-kubernetes/kubernetes.tf in the deployment resource and use terraform to apply:

resource "kubernetes_deployment" "wc" {
  # ...

  spec {
    replicas = 4

    # ...
  }

  # ...
}

Then update the AWS:

$ terraform apply

Since the EKS is at AWS, it has the AWS CloudWatch, and we can easily check the EC2 and the EKS cluster health state along with the metrics.

Terraform has the destroy command:

$ cd infra/deploy-wc-kubernetes
$ terraform destroy
$ cd infra/deploy-cluster
$ terraform destroy