Automatic Knative Service Manipulation

Introduction

@Brief: A VS Code extension that helps developers automatically manipulate Knative services on the Kubernetes cluster efficiently (e.g., deploying Knative services, retrieving details about Knative Services, deleting Knative services).

@Author: Dongzhi Zhang

@LastUpdate: 2024/03/11

Prerequisites

Please ensure the following tools are set up in your local environment.

• Docker / Docker Desktop

• Kubernetes (Minikube, kubectl)

• Knative

Supported Commands

It currently supports the following commands:

• aksm.deployKnService: Deploy a Knative service to the Kubernetes cluster (specified by the input "registry name" and "image name").

• aksm.describeKnService: Print detailed information about a Knative service (specified by the input "service name").

• aksm.listKnServices: Describe All Knative Services: Print detailed information about all Knative services.

• aksm.deleteKnService: Delete a Knative Service (specified by the input "service name").

• aksm.testExtension: Test whether the extension is working properly.

Example Usage

0. Ensure that necessary tools and dependencies are installed and started.

   • Refer to "prerequisites" for required tools.

   • Make sure minikube start -p knative and minikube tunnel --profile knative are executed to start the Minikube tunnel for Knative.

1. Search for the extension "Automatic Knative Service Manipulation" in the VS Code marketplace and install it.

2. Open the folder containing the Flask app ready to run as the root folder.

   • The Python file should be named app.py, or the deployment would fail.

   • You need to provide a requirements.txt containing all dependencies that your Flask app needs.

   • Here I provide a quickstart under the folder "quickstart", which contains a Flask app app.py running a text classification ML task and a requirements.txt containing dependencies flask, torch, transformers. You can conduct a brief test on the extension under this folder.

3. In the VS Code palette, execute aksm.testExtension to test whether the extension is working properly.

   • Use Ctrl+Shift+P or Cmd+Shift+P for Mac to trigger the VS Code palette, which is totally different from simply clicking the search bar above.

4. To run the Flask app within a Knative service and deploy it to the Kubernetes cluster, execute aksm.deployKnService in the VS Code palette.

   • You will need to specify the name of the Docker image registry ("registry name") and the Docker image ("image name") in the palette.

   • Then, the deployment will start. You can check the log terminal for the latest progress. (Note that pushing the Docker image might take about 4~5 minutes.)

   • After successfully deploying the Knative service, the service name will be printed in the log terminal and you can use it with aksm.describeKnService or aksm.listKnServices to check the service details.

5. To get the detailed information about the Knative service we just deployed, execute aksm.describeKnService in the VS Code palette.

   • You will need to specify the name of the Knative service ("service name") in the palette.

   • You can also get the detailed information about all Knative services with aksm.listKnServices.

6. Several steps to test whether the deployment truly succeeds.

   • Check Docker Hub / execute docker images to check whether the specified Docker image is successfully pushed to the specified registry.

   • Execute kn service list or kubectl get ksvc to check whether the Knative service is successfully deployed.

   • Use Postman to conduct API tests on the Knative service's URL to check whether the Flask app is running properly.

7. To delete a Knative service, execute aksm.deleteKnService in the VS Code palette.

   • You will need to specify the name of the Knative service ("service name") in the palette.

Release Notes

1.0.0

Initial release of "Automatic Manipulation on Knative Service".

1.1.0

The generated Dockerfile now installs dependencies based on the dependency file requirements.txt provided by the user.

Rename commands Deploy Service -> Deploy Knative Service, Get Service Info -> Describe Knative Service, Delete Service -> Delete Knative Service, Test Extension -> Test AKSM.

Add a new command Describe All Knative Services.

Update quickstart.

1.2.0

Replace usages of kubectl with kn.

Rename commands Deploy Knative Service -> aksm.deployKnService, Describe Knative Service -> aksm.describeKnService, Delete Knative Service -> aksm.deleteKnService, Test AKSM -> aksm.testExtension.

Update logs information.

1.3.0

Automate the building and pushing of Docker images with BuildPack for non Apple-Silicon-chip users.

Issues

• For Mac Apple-Silicon-chip (M1, M2) users, their system cannot use Buildpack to automatically build and push a Docker image based on the source code.

  • We have failed on known builder images that seem to support ARM64 architecture, such as paketobuildpacks/builder and dashaun/builder-arm, which can only automate the image building upon source codes in Java/Go/... (no Python solution).

  • There are other solutions that can help automatically build and push a Docker image based on the source codes, but they still cannot achieve our goals.

    • For Knative Build, it's deprecated and no longer maintained for a long time. Hence, we do not consider using it.

    • For Tekton Pipelines, it can support the automatic image building and pushing based on source codes under a ARM64 architecture. However, it requires us to create and define YAML files for the Task and TaskRun.

  • For non Apple-Silicon-chip users, they can use pack build [IMAGE_NAME] --builder [BUILDER_NAME] --path [SOURCE_CODE_PATH] to directly build and push the source code's corresponding Docker image, bypassing the steps for creating, building, and pushing the Docker image.

Relevant Tools

• Kubernetes: https://kubernetes.io/

• Knative: https://knative.dev/

• Docker: https://www.docker.com/

• Flask: https://flask.palletsprojects.com/en/3.0.x/

• Yeoman Generator: https://yeoman.io/

• Hugging Face: https://huggingface.co/models