Check: 'DogFood' Testing

Kontroler: A Self-Testing Workflow Orchestration Tool

My latest project, Kontroler, is a workflow orchestration tool with a unique twist: self-testing. It not only orchestrates workflows but also uses its own capabilities to test its functionality, a process known as "dogfooding." This continuous self-testing ensures the tool works reliably while simultaneously improving itself.

Why Self-Testing?

My passion for automation drives my desire to create efficient and reliable tools. Kontroler isn't just another workflow runner; it's designed to prove its own functionality without external testing tools. This built-in stress test provides a strong foundation, and observing its self-imposed quality control is both practical and rewarding.

How Kontroler Tests Itself

Achieving self-testing required creating additional components:

• GitHub Webhook Listener: Handles incoming webhooks, processes the infrastructure directory, and generates the DAG and DAGRun.
• Kontroler Webhook Listener: Handles outgoing events from Kontroler, providing task completion status and linked task information.

A new .kontrolerci directory was added to the repository. This directory houses YAML files defining the DAG/CI pipeline. A simple Go builder example:

<code>build-server:
  image: golang:1.23.5-bullseye
  script: |
    echo "Navigating to server directory"
    cd server
    go mod tidy
    echo "Building server application"
    go build -o main cmd/main.go</code>

The complete directory structure is available on GitHub: https://www.php.cn/link/6968240043185f65f22097299b865f6b

What Does it Test?

The current pipeline tests:

• Package Installation (npm, Go)
• Go Linting
• Unit Testing

These tests build confidence in the system, and the successful CI pipeline execution provides end-to-end verification of Kontroler's task allocation and webhook generation.

GitHub's UI displays the results:

The Rewards (and Challenges) of Dogfooding

The first successful pipeline run was incredibly satisfying – seeing those green checkmarks exceeded my expectations. Kontroler was truly being used, even if only by me.

However, self-testing wasn't without its challenges. Debugging issues like client authentication failures and errors related to temporary files proved time-consuming. But, this process of dogfooding revealed and allowed me to address crucial flaws.

Future Plans for Kontroler

Dogfooding has identified several areas for improvement:

• Runtime PodTemplate Override: Enhanced flexibility for workflows like CI pipelines.
• PVC Creation within the DagRun CRD: Improved resource management.

We will continue to dogfood Kontroler to further refine and productionize the tool.

The above is the detailed content of Check: 'DogFood' Testing. For more information, please follow other related articles on the PHP Chinese website!