devops-pipeline

An example devops pipeline for prototyping. Author: Dylan Klein

What does this pipeline do?

The demo pipeline builds a simple "hit counter" web app using Python's Flask framework with Redis as a database. It then deploys the dockerised architecture to Google Kubernetes Engine (GKE).

Architecture

The following is a visual representation of the CICD process.

1. Code

Developers can develop the Flask app on their localhost in a containerised environment using Docker. If you wish to run the app on localhost (for testing purposes):

1. Modify app.py line 7:

cache = redis.Redis(host='my-redis-svc.default.svc.cluster.local', port=6379)

to:

cache = redis.Redis(host='redis', port=6379)

⚠️ Don't push this change to your git repository.

2. Run the command $ docker-compose up to build the app locally.

2. Authenticate to GCP

When the workflow is triggered (either via git push or otherwise), the first thing that GitHub Actions does is authenticate to GCP via Workload Identity Federation.

3. Build

The following docker image is then built via GitHub Actions (Continuous Integration):

• dylan-website (Flask frontend)

4. Publish

The latest docker image (titled dylan-website) is then published to GCP's Artifact Registry.

5. Deploy

Once docker build & publish are successful, GitHub Actions deploys the following 2 docker images as workloads to GKE (Continuous Deployment):

1. gke-test (Flask frontend)
2. redis-master (Redis backend)

GitHub Actions also deploys the following 2 services, such that the workloads mentioned above can be accessed:

1. gke-test-service (External Loadbalancer - i.e. internet facing)
2. my-redis-svc (ClusterIP)

Installation

1. Create a new GCP project and authorise GitHub Actions via IAM. Follow these instructions for help: https://github.com/google-github-actions/setup-gcloud/tree/master/example-workflows/gke-kustomize
2. Git clone this repository: $ git clone https://github.com/dylman123/devops-pipeline.git
3. Add your GCP project ID as a secret in your project repository, using GKE_PROJECT as the key.
4. Modify cicd-pipeline.yml with your specific env variables.
5. Modify cicd-pipeline.yml with your specific auth details for:

• workload_identity_provider and
• service_account

Run

To run, all you have to do is git push your code or manually click 'Re-run all jobs' in GitHub Actions.

Test

To ensure your deployment worked successfully:

1. Find the IP address of the gke-test-service service which you deployed in GCP.
2. Navigate to the IP address in your browser on port 80.