Skip to content

Latest commit

 

History

History
100 lines (61 loc) · 3.12 KB

README.md

File metadata and controls

100 lines (61 loc) · 3.12 KB
Raw

ML Pipeline Test Infrastructure

This folder contains the integration and end-to-end (E2E) tests for the ML pipeline. Tests are executed using Kind (Kubernetes IN Docker) to simulate a Kubernetes cluster locally. GitHub Actions (GHAs) handle automated testing on pull requests.

At a high level, a typical test workflow will:

  • Build images for all components.
  • Create a dedicated test namespace in the cluster.
  • Deploy the ML pipeline using the newly built components.
  • Run the tests.
  • Clean up the namespace and temporary resources.

These steps are performed in the same Kubernetes cluster.

Running Tests Locally with Kind

To run tests locally, set up a Kind cluster and follow the same steps as the GitHub Actions workflows. This section details the process.

Setup

  1. Install Prerequisites:

  2. Set Up a Kind Cluster: Create a configuration file for your Kind cluster (optional):

    kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
  - role: control-plane
  - role: worker

    Create the cluster:

    kind create cluster --name kfp-test-cluster --config kind-config.yaml

    Verify the cluster:

    kubectl cluster-info --context kind-kfp-test-cluster

  3. Prepare the Test Environment:

    • Install Python test dependencies:

      pip install -r test/requirements.txt

    • Deploy Kubeflow Pipelines to the Kind cluster:

      kubectl apply -k manifests/

  4. Run the Tests: Execute the desired test suite:

    pytest test/kfp-functional-test/

For additional guidance on deploying Kubeflow Pipelines in Kind, refer to:

Automated Testing with GitHub Actions

Tests are automatically triggered on GitHub when:

  • A pull request is opened or updated.

GitHub Actions workflows are defined in the .github/workflows/ directory.

Reproducing CI Steps Locally

To replicate the steps locally:

  1. Clone the Kubeflow Pipelines repository:

    git clone https://github.com/kubeflow/pipelines.git cd pipelines

  2. Follow the steps outlined in the Running Tests Locally with Kind section.

  3. To mimic the GitHub Actions environment, export any required environment variables found in the workflow files.

Troubleshooting

Q: Why is my test taking so long?

  • The first run downloads many container images. Subsequent runs will be faster due to caching.
  • If you experience high latency, ensure the local system running Kind has sufficient resources (CPU, memory).

Q: How do I clean up the Kind cluster?

  • Delete the Kind cluster: kind delete cluster --name kfp-test-cluster