Obviously, the pipeline can be diveded into two parts - CI and CD. I decided to use Python to simpify the process of creating the pipeline (but implementation in shell scripts is also possible). The pipeline will be run on a RaspberryPI and the image will be deployed there as well. To reduce the memmory usage on the Raspberry, the pipeline will run directly on the host machine, not in a container. However, it is possible to run the pipeline in a container as well.
I created a sample Flask application which will be used in testing the pipeline which you can find here.
To check for changes, a GitHub webhook will be used. The webhook will send a POST request to the RaspberryPI with a payload containing information about the changes in the repository. The payload will be parsed and the pipeline will be run.
An designeted /webhook endpoint will be created to handle the POST requests.
When the app receives a POST request from the webhook, it will pull the new code from the repository.
When the new code is pulled, the tests will be run.
If the tests pass, the pipeline will build a Docker image from a Dockerfile included in the repository.
If the image is built successfully, it will be pushed to a Docker Hub repository with a new tag (in some way incremented).
The pipeline will check for changes in the Docker Hub repository for a new image.
If there is a new image, the previously running image will be stopped and removed.
When the previous image is stopped and removed, the new image will be pulled and run.
If any of the steps fails, a notification will be sent via email.
If the image is deployed successfully, a notification will be sent via email.
Each step will be logged and the logs will be stored in a file. The file is named after the date and time of the run and the run number.
The pipeline will not proceed with the next steps if any of the checks fails and will rollback to the previous deployment.