From legacy to cloud serverless - Part 3

David WOGLO - Sep 4 - - Dev Community

Note: This article was originally published on
Dec 25, 2023 here. It has been republished here to reach a broader audience.

diagram

Welcome to the third part of this serie! In this segment, we dive into testings and pipeline configuration on Google Cloud, specifically focusing on continuous integration using Cloud Build, On-demand Vulnerability Scanner, and Artifact Registry. You can find the project repository here, or, if you prefer, you can bring your own project.

Let me walk you through the pipeline. With each push to the main branch, Cloud Build is triggered. First, it runs unit tests on the code. If the tests pass, it proceeds to build the image. After the image is built, Cloud Build invokes the image scanner to ensure it's free of vulnerabilities. If all is well, the image is sent and stored in the Artifact Registry, ready for deployment. But for this article, we'll focus solely on the CI part. Let's start with the tests.

Unittest

Here's the code we plan to test

import os
from flask import Flask
from pymongo import MongoClient
from flask import Flask, render_template, request, url_for, redirect
from bson.objectid import ObjectId
import mongomock



app = Flask(__name__, template_folder='templates')

if os.environ.get('TESTING'):
    client = mongomock.MongoClient()
else:
    client = MongoClient(os.environ['MONGO_URI'])


db = client.flask_db
todos = db.todos


@app.route('/', methods=('GET', 'POST'))
def index():
    if request.method=='POST':
        content = request.form['content']
        degree = request.form['degree']
        todos.insert_one({'content': content, 'degree': degree})
        return redirect(url_for('index'))

    all_todos = todos.find()
    return render_template('index.html', todos=all_todos)


@app.post('/<id>/delete/')
def delete(id):
    todos.delete_one({"_id": ObjectId(id)})
    return redirect(url_for('index'))
Enter fullscreen mode Exit fullscreen mode

For an explanation of the code, refer to the first article in this series.

Now, let's move on to the testing phase

The test is written using Python's built-in unittest module, which provides a framework for writing and running tests.

  1. Import necessary modules and create a mock MongoDB instance

    The test begins by importing the necessary modules. unittest is the testing framework, patch and MagicMock from unittest.mock are used to replace parts of the system that you're testing with mock objects, and ObjectId from bson.objectid is used to create unique identifiers. The app and todos are imported from the app.py file. mongomock is used to create a mock MongoDB instance for testing, and flask is used to manipulate the request context during testing.

    import unittest
    from unittest.mock import patch, MagicMock
    from bson.objectid import ObjectId
    from app import app, todos
    import mongomock
    import flask
    
    mock_db = mongomock.MongoClient().db
    
  2. Define the test case

    A test case is defined by creating a new class that inherits from unittest.TestCase. This class will contain methods that represent individual tests.

    class TestApp(unittest.TestCase):
    
  3. Set up the test environment

    The setUp method is a special method that is run before each test. Here, it's used to create a test client instance of the Flask app and enable testing mode.

    def setUp(self):
        self.app = app.test_client()
        self.app.testing = True
    
  4. Write the test

    The test_index_post method is the actual test. It tests the behavior of the app when a POST request is sent to the index route (/).

    def test_index_post(self):
    
  5. Mock the database operation

    The patch function is used to replace the insert_one method of todos with a MagicMock. This allows the test to simulate the behavior of the database operation without actually interacting with a real database.

    with patch('app.todos.insert_one', new_callable=MagicMock) as mock_insert_one:
    
  6. Create a test request context

    A test request context is created for the app using app.test_request_context. This allows the test to simulate a request to the app.

    with app.test_request_context('/'):
    
  7. Set the request method and form data

    The request method is set to 'POST' and the request form data is set to a dictionary with 'content' and 'degree' keys.

    flask.request.method = 'POST'
    flask.request.form = {'content': 'Test Content', 'degree': 'Test Degree'}
    
  8. Send a POST request to the app

    A POST request is sent to the app using self.app.post. The form data is passed as the data argument.

    result = self.app.post('/', data=flask.request.form)
    
  9. Assert the expected results

    The assertEqual method is used to check that the status code of the response is 302. The assert_called method is used to check that the insert_one method was called.

    self.assertEqual(result.status_code, 302)
    mock_insert_one.assert_called()
    

This test ensures that when a POST request is sent to the index route with the correct form data, the app responds with a 302 status code and inserts the data into the database.

Your test code should look something like the following:

import unittest
from unittest.mock import patch, MagicMock
from bson.objectid import ObjectId
from app import app, todos
import mongomock
import flask

## Create a mock MongoDB instance
mock_db = mongomock.MongoClient().db

class TestApp(unittest.TestCase):
    def setUp(self):
        # Create a test client instance
        self.app = app.test_client()
        # Enable testing mode. Exceptions are propagated rather than handled by the the app's error handlers
        self.app.testing = True 

    def test_index_post(self):
        # Patch the insert_one method of todos with a MagicMock
        with patch('app.todos.insert_one', new_callable=MagicMock) as mock_insert_one:
            # Create a test request context for the app
            with app.test_request_context('/'):
                # Set the request method to 'POST'
                flask.request.method = 'POST'
                # Set the request form data
                flask.request.form = {'content': 'Test Content', 'degree': 'Test Degree'}
                # Send a POST request to the app
                result = self.app.post('/', data=flask.request.form)
                # Assert that the status code of the response is 302
                self.assertEqual(result.status_code, 302)
                # Assert that the insert_one method was called
                mock_insert_one.assert_called()
Enter fullscreen mode Exit fullscreen mode

Now, to execute the test, set the environment variable TESTING=true, Setting TESTING=True switches the application to use a mock MongoDB client for testing, instead of the real MongoDB database.

Now, if your test is successful, let's move on to configuring Cloud Build.

Cloud Build setup

Follow the guide to connect Cloud Build to your repository and this one for initial configurations.

Once that's done, let's move on to writing the Cloud Build configuration file, where we'll instruct it on how to execute the pipeline, the steps involved, dependencies, and so on.

Cloud Build config file

The Cloud Build Config file is written in YAML, a human-readable data serialization language.

Here are the main sections of our config file:

  1. Substitutions: These are user-defined variables that can be replaced in the Cloud Build configuration file. They are defined under the substitutions key. In this case, _REGION, _REPOSITORY, _IMAGE, and _SEVERITY are defined.

    substitutions:
      _REGION: us-central1
      _REPOSITORY: from-legacy-to-cloud
      _IMAGE: from-legacy-to-cloud
      _SEVERITY: '"CRITICAL|HIGH"'
    
  2. Steps: These are the operations that Cloud Build will perform. Each step is a separate action and they are executed in the order they are defined.

* **Step 0: Install test dependencies**: This step uses a Python 3.10 Docker image to install the test dependencies listed in `docker/requirements-test.txt`. The `entrypoint` is set to `/bin/bash`, which means that the command that follows will be executed in a bash shell. The `args` key specifies the command to be executed, which in this case is a pip install command. The `-c` flag tells bash to read commands from the following string. The `|` character allows us to write multiple commands, which will be executed in order.
Enter fullscreen mode Exit fullscreen mode
    ```yaml
    - name: 'python:3.10-slim'
      entrypoint: '/bin/bash'
      args:
        - '-c'
        - |
          pip install --user -r docker/requirements-test.txt
      id: 'install-test-dependencies'
    ```
Enter fullscreen mode Exit fullscreen mode
* **Step 1: Run unit tests**: This step also uses a Python 3.10 Docker image to run the unit tests defined in [`test.py`](http://test.py). The `export TESTING=True` command sets an environment variable `TESTING` to `True`, which can be used to change the behavior of the application during testing. The `cd docker` command changes the current directory to `docker`, where the test file is located. The `python -m unittest` [`test.py`](http://test.py) command runs the unit tests in [`test.py`](http://test.py).
Enter fullscreen mode Exit fullscreen mode
    ```yaml
    - name: 'python:3.10-slim'
      entrypoint: '/bin/bash'
      args:
        - '-c'
        - |
          export TESTING=True
          cd docker 
          python -m unittest test.py
      id: 'run-tests'
    ```
Enter fullscreen mode Exit fullscreen mode
* **Step 2: Build the Docker image**: This step uses the `docker` Cloud Builder to build a Docker image from the Dockerfile located in the `docker/` directory. The image is tagged with the commit SHA. The `waitFor` key is used to specify that this step should wait for the `run-tests` step to complete before it starts. The `args` key specifies the command to be executed, which in this case is a docker build command. The `-t` flag is used to name and optionally tag the image in the 'name:tag' format.
Enter fullscreen mode Exit fullscreen mode
    ```yaml
    - name: 'gcr.io/cloud-builders/docker'
      args: ['build', '-t', '$_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA', 'docker/']
      waitFor: ['run-tests']
      id: 'build-image'
    ```
Enter fullscreen mode Exit fullscreen mode
  • Step 3: Inspect the Docker image and write the digest to a file: This step uses the docker Cloud Builder to inspect the Docker image and write the image digest to a file. The image digest is a unique identifier for the image. The docker image inspect command retrieves detailed information about the Docker image. The --format option is used to format the output using Go templates. The {{index .RepoTags 0}}@{{.Id}} template retrieves the first tag of the image and the image ID. The > operator redirects the output to a file. The && operator is used to execute the cat command only if the previous command succeeded.

    - name: 'gcr.io/cloud-builders/docker'
      entrypoint: '/bin/bash'
      args:
        - '-c'
        - |
          docker image inspect $_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA --format '{{index .RepoTags 0}}@{{.Id}}' > /workspace/image-digest.txt &&
          cat /workspace/image-digest.txt
      id: 'inspect-image'
    
  • Step 4: Scan the Docker image for vulnerabilities: This step uses the cloud-sdk Cloud Builder to scan the Docker image for vulnerabilities. The scan ID is written to a file. The gcloud artifacts docker images scan command scans the Docker image for vulnerabilities. The --format='value(response.scan)' option is used to retrieve the scan ID from the response. The > operator redirects the output to a file.

    - id: scan
      name: gcr.io/google.com/cloudsdktool/cloud-sdk
      entrypoint: /bin/bash
      args:
      - -c
      - |
        gcloud artifacts docker images scan $_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA \
        --format='value(response.scan)' > /workspace/scan_id.txt
    
  • Step 5: Check the severity of any vulnerabilities found: This step uses the cloud-sdk Cloud Builder to list the vulnerabilities found in the Docker image and check their severity. If any vulnerabilities with a severity matching _SEVERITY are found, the build fails. The gcloud artifacts docker images list-vulnerabilities command lists the vulnerabilities found in the Docker image. The --format='value(vulnerability.effectiveSeverity)' option is used to retrieve the severity of each vulnerability. The grep -Exq $_SEVERITY command checks if any of the severities match _SEVERITY. The echo command prints a message and the exit 1 command terminates the build if a match is found.

    - id: severity check
      name: gcr.io/google.com/cloudsdktool/cloud-sdk
      entrypoint: /bin/bash
      args:
      - -c
      - |
        gcloud artifacts docker images list-vulnerabilities $(cat /workspace/scan_id.txt) \
        --format='value(vulnerability.effectiveSeverity)' | if grep -Exq $_SEVERITY; \
        then echo 'Failed vulnerability check' && exit 1; else exit 0; fi
    
  • Step 6: Push the Docker image to Google Cloud Artifact Registry: This step uses the docker Cloud Builder to push the Docker image to the Google Cloud Artifact Registry. The waitFor key is used to specify that this step should wait for the severity check step to complete before it starts. The docker push command pushes the Docker image to a repository.

    - name: 'gcr.io/cloud-builders/docker'
      args: ['push', '$_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA']
      id: 'push-image'
      waitFor: ['severity check']
    
  • Images: This key specifies the Docker images that Cloud Build should build and push to the Google Cloud Artifact Registry. In this case, it's the Docker image built in Step 2.

    images:
    - '$_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA'
    

This cloudbuild.yaml file defines a complete CI/CD pipeline for our application. It installs test dependencies, runs unit tests, builds a Docker image, inspects the image, scans the image for vulnerabilities, checks the severity of any vulnerabilities found, and pushes the image to the Google Cloud Artifact Registry. This pipeline ensures that the application is tested, secure, and ready for deployment.

The complete config file should look like this:

substitutions:
  _REGION: us-central1
  _REPOSITORY: from-legacy-to-cloud
  _IMAGE: from-legacy-to-cloud
  _SEVERITY: '"CRITICAL|HIGH"'

steps:
# Step 0: Install test dependencies
- id: 'install-test-dependencies'
  name: 'python:3.10-slim'
  entrypoint: '/bin/bash'
  args:
    - '-c'
    - |
      pip install --user -r docker/requirements-test.txt

# Step 1: Run unit tests
- id: 'run-tests'
  name: 'python:3.10-slim'
  entrypoint: '/bin/bash'
  args:
    - '-c'
    - |
      export TESTING=True
      cd docker 
      python -m unittest test.py

# Step 2: Build the Docker image
- id: 'build-image'
  name: 'gcr.io/cloud-builders/docker'
  args: ['build', '-t', '$_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA', 'docker/']
  waitFor: ['run-tests']

# Step 3: Inspect the Docker image and write the digest to a file.
- id: 'inspect-image'
  name: 'gcr.io/cloud-builders/docker'
  entrypoint: '/bin/bash'
  args:
    - '-c'
    - |
      docker image inspect $_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA --format '{{index .RepoTags 0}}@{{.Id}}' > /workspace/image-digest.txt &&
      cat /workspace/image-digest.txt

# Step 4: Scan the Docker image for vulnerabilities
- id: scan
  name: gcr.io/google.com/cloudsdktool/cloud-sdk
  entrypoint: /bin/bash
  args:
  - -c
  - |
    gcloud artifacts docker images scan $_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA \
    --format='value(response.scan)' > /workspace/scan_id.txt

# Step 5: Check the severity of any vulnerabilities found
- id: severity check
  name: gcr.io/google.com/cloudsdktool/cloud-sdk
  entrypoint: /bin/bash
  args:
  - -c
  - |
    gcloud artifacts docker images list-vulnerabilities $(cat /workspace/scan_id.txt) \
    --format='value(vulnerability.effectiveSeverity)' | if grep -Exq $_SEVERITY; \
    then echo 'Failed vulnerability check' && exit 1; else exit 0; fi

# Step 6: Push the Docker image to Google Cloud Artifact Registry
- id: 'push-image'
  name: 'gcr.io/cloud-builders/docker'
  args: ['push', '$_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA']
  waitFor: ['severity check']

images:
- '$_REGION-docker.pkg.dev/$PROJECT_ID/$_REPOSITORY/$_IMAGE:$COMMIT_SHA'
Enter fullscreen mode Exit fullscreen mode

View build results

Now, commit and push your changes. If the Cloud Build triggers are configured correctly, the build should be triggered. Connect to the Google Cloud Console, go to Cloud Build > History to view your builds.

build result

If it fails, click on it to see the error messages and troubleshoot to resolve the issues. Once the build succeeds, you can access the Artifact Registry and see the stored image, ready for use.

buid result

What next ?

Well, that wraps up this article. In the next one, we'll delve into automating deployments—the CD part. After vulnerability scanning of container images, we'll be putting security policies in place through Binary Authorization, allowing only approved/trusted images to be deployed on Cloud Run. But before that, we'll migrate our Mongo database to Google Firestore. After that, we'll deploy our app on Cloud Run and connect it to Firestore to make it fully operational.

See you in the next article. Until then, I'm available on social media (I'm more active on LinkedIn) for any information or additional suggestions. Thanks for reading!

. . . . . . . . . . .
Terabox Video Player