Description

Will print time in python as displayed in the python console and output it to a Flask web application. Multiple deployment options from a local docker-compose all the way to GCP Terraform deployment with Helm and K8s and autoscaling prebuilt into it.

[App]

• Dockerfile -> lighter than a CokeZero with 0 vulnerabilities as scanned on Snyk, size 25mb
• requirements.txt -> python dependencies
• print_timestamp.py -> original code supplied
• main.py -> my Python code, this is a basic Flask Python app that launches a subprocess every time the function is called at the / of the webpage I included two other functions that return HTTP response 200 and 500 for the @app.route just so I can generate more replies for the integrated load tester and the Grafana dashboard.(more on that in just a sec!) Also introduced a library needed for exporting the metrics from the Flask application to a Prometheus instance.

[Generator]

• generate_events.py -> script that picks at random endpoints and keeps sending requests for generating load for testing autoscaling
• Dockerfile -> just another Dockerfile, 0 vulnerabilities, size 25mb
• requirements.txt -> python dependencies

[Grafana]

• grafana config & dashboard files

[Helmchart]

• Helm chart for deploying with Helm.

[Manual_gke_provisioning]

• Collection of yaml files that can be applied with kubectl -f . for a working cluster in GKE.

[Prometheus]

• Prometheus and its config file

[Terraform]

• Will create a separate VPC and separate subnets for the Kubernetes cluster, create the K8 cluster and deploy the Helm chart
• Will create a two node cluster with machine type: n1-standard-1
• Requires helm, terraform, gcloud SDK and a service account key (json) with enough privileges.

[LINKS]

The GKE cluster will expose the Grafana dashboard and the Flask application at these links:

• GRAFANA = http://lb-ip:3000/d/_eX4mpl3
• FLASK = http://lb-ip:5000/

[Notes]

• Prometheus nor Grafana have any persistent volume claims, I would attach an NFS share in production for this.
• All K8 deployments come with ConfigMaps defined
• Cluster monitoring is possible as well but wasn't mentioned under bonus objectives, at the same time a deployment on GKE has observability by default.
• Default namespace used
• If you want to spin it locally all you have to do is docker-compose up
• Prototyped everything on GCP
• HPA autoscaling is configured in all deployment scenarios (manual, helm or terraform)
• Edit terraform.tfvars for declaring your GCP project and region

[Pics]