aws-delivlib

aws-delivlib is a fabulous library for defining continuous pipelines for building, testing and publishing code libraries through AWS CodeBuild and AWS CodePipeline.

aws-delivlib is used by the AWS Cloud Development Kit and was designed to support simultaneous delivery of the AWS CDK in multiple programming languages packaged via jsii.

Pipeline Structure

A delivlib pipeline consists of the following sequential stages. Each stage will execute all tasks concurrently:

+-----------+     +-----------+     +-----------+     +----------------+
|  Source   +---->+   Build   +---->+   Test    +---->+    Publish     |
+-----------+     +-----------+     +-----+-----+     +-------+--------+
                                          |                   |
                                          v                   v
                                    +-----+-----+     +-------+-------+
                                    |   Test1   |     |      npm      |
                                    +-----------+     +---------------+
                                    |   Test2   |     |     NuGet     |
                                    +-----------+     +---------------+
                                    |   Test3   |     | Maven Central |
                                    +-----------+     +---------------+
                                    |    ...    |     |     PyPI      |
                                    +-----------+     +---------------+
                                                      |  GitHub Pages |
                                                      +---------------+
                                                      |GitHub Releases|
                                                      +---------------+

The following sections describe each stage and the configuration options available:

Source
- CodeCommit
- GitHub
Pull Request Builds
Build
Tests
Publish
- npm.js
- NuGet
- Maven Central
- PyPi
- GitHub Releases
- GitHub Pages
Metrics
Automatic Bumps and Pull Request Builds
Failure Notifications
ECR Mirror

Installation

To install, use npm / yarn:

$ npm i aws-delivlib

or:

$ yarn add aws-delivlib

and import the library to your project:

import delivlib = require('aws-delivlib');

The next step is to add a pipeline to your app. When you define a pipeline, the minimum requirement is to specify the source repository. All other settings are optional.

const pipeline = new delivlib.Pipeline(this, 'MyPipeline', {
  // options
});

The following sections will describe the various options available in your pipeline.

You can also take a look at the pipeline definition releasing the delivlib library itself for a real-world, working example.

Source

The only required option when defining a pipeline is to specify a source repository for your project.

`repo`: Source Repository (required)

The repo option specifies your source code repository for your project. You could use either CodeCommit or GitHub.

CodeCommit

To use an existing repository:

import codecommit = require('@aws-cdk/aws-codecommit');

// import an existing repository
const myRepo = codecommit.Repository.fromRepositoryName(this, 'TestRepo',
  'delivlib-test-repo');

// ...or define a new repository (probably not what you want)
const myRepo = new codecommit.Repository(this, 'TestRepo');

// create a delivlib pipeline associated with this codebuild repo
new delivlib.Pipeline(this, 'MyPipeline', {
  repo: new delivlib.CodeCommitRepo(myRepo),
  // ...
});

GitHub

To connect to GitHub, you will need to store a Personal GitHub Access Token as an SSM Parameter and provide the name of the SSM parameter.

import cdk = require('@aws-cdk/core');

new delivlib.Pipeline(this, 'MyPipeline', {
  repo: new delivlib.GitHubRepo({
    repository: 'awslabs/aws-delivlib',
    token: cdk.SecretValue.secretsManager('my-github-token'),
  }),
  // ...
})

`branch`: Source Control Branch (optional)

The branch option can be used to specify the git branch to build from. The default is master.

new delivlib.Pipeline(this, 'MyPipeline', {
  repo: // ...
  branch: 'dev',
})

Pull Request Builds

Pull Request Builds can be used to validate if changes submitted via a pull request successfully build and pass tests. They are triggered automatically by GitHub or CodeCommit when pull requests are submitted or updated.

Known in delivlib as AutoBuild, they can be enabled on the Pipeline and further configured -

new delivlib.Pipeline(this, 'MyPipeline', {
  // ...
  autoBuild: true,
  autoBuildOptions: {
    publicLogs: true,
  },
});

Delivlib also separately exports the AutoBuild construct that can be used to configure AutoBuild on a project that doesn't have a pipeline associated, or for jobs that can be run outside of a pipeline.

new delivlib.AutoBuild(this, 'MyAutoBuild', {
  repo: // ...
});

Build

The second stage of a pipeline is to build your code. The following options allow you to do customize your build environment and scripts:

`buildSpec`: Build Script (optional)

The default behavior will use the buildspec.yaml file from the root of your source repository to determine the build steps.

See the the buildspec reference documentation in the CodeBuild User Guide.

Note that if you don't have an "artifacts" section in your buildspec, you won't be able to run any tests against the build outputs or publish them to package managers.

If you wish, you can use the buildSpec option, in which case CodeBuild will not use the checked-in buildspec.yaml:

import codebuild = require('@aws-cdk/aws-codebuild');

new delivlib.Pipeline(this, 'MyPipeline', {
  // ...
  buildSpec: codebuild.BuildSpec.fromObject({
    version: '0.2',
    phases: {
      build: {
        commands: [
          'echo "Hello, world!"'
        ]
      }
    },
    artifacts: {
      files: [ '**/*' ],
      'base-directory': 'dist'
    }
  }),
});

`buildImage`: Build container image (optional)

The Docker image to use for the build container.

Default: the default image (if none is specified) is a custom Docker image which is provided as part of the jsii distribution called jsii/superchain. It is an environment that supports building libraries that target all programming languages supported by jsii. Find more information on the contents of the jsii/superchain image on the jsii homepage.

You can use the AWS CodeBuild API to specify any Linux/Windows Docker image for your build. Here are some examples:

codebuild.LinuxBuildImage.fromDockerRegistry('golang:1.11') - use an image from Docker Hub
codebuild.LinuxBuildImage.UBUNTU_14_04_OPEN_JDK_9 - OpenJDK 9 available from AWS CodeBuild
codebuild.WindowsBuildImage.WIN_SERVER_CORE_2016_BASE - Windows Server Core 2016 available from AWS CodeBuild
codebuild.LinuxBuildImage.fromEcrRepository(myRepo) - use an image from an ECR repository

`env`: Build environment variables (optional)

Allows adding environment variables to the build environment:

new delivlib.Pipeline(this, 'MyPipeline', {
  // ...
  environment: {
    FOO: 'bar'
  }
});

Other Build Options

computeType: size of the AWS CodeBuild compute capacity (default: SMALL)
privileged: run in privileged mode (default: false)

Tests

The third stage of a delivlib pipeline is to execute tests. Tests are executed in parallel only after a successful build and can access build artifacts as defined in your buildspec.yaml.

The pipeline.addTest method can be used to add tests to your pipeline. Test scripts are packaged as part of your delivlib CDK app.

delivlib.addTest('MyTest', {
  platform: delivlib.ShellPlatform.LinuxUbuntu(), // or `ShellPlatform.Windows()`
  scriptDirectory: 'path/to/local/directory/with/tests',
  entrypoint: 'run.sh',
});

scriptDirectory refers to a directory on the local file system which must contain the entrypoint file. Preferably make this path relative to the current file using path.join(__dirname, ...).

The test container will be populated the build output artifacts as well as all the files from the test directory.

Then, the entry-point will be executed. If it fails, the test failed.

Publish

The last step of the pipeline is to publish your artifacts to one or more package managers. Delivlib is shipped with a bunch of built-in publishing tasks, but you could add your own if you like.

To add a publishing target to your pipeline, you can either use the pipeline.addPublish(publisher) method or one of the built-in pipeline.publishToXxx methods. The first option is useful if you wish to define your own publisher, which is class the implements the delivlib.IPublisher interface.

Built-in publishers are designed to be idempotent: if the artifacts version is already published to the package manager, the publisher will succeed. This means that in order to publish a new version, all you need to do is bump the version of your package artifact (e.g. change package.json) and the publisher will kick in.

You can use the dryRun: true option when creating a publisher to tell the publisher to do as much as it can without actually making the package publicly available. This is useful for testing.

The following sections describe how to use each one of the built-in publishers.

npm.js (JavaScript)

The method pipeline.publishToNpm will add a publisher to your pipeline which can publish JavaScript modules to npmjs.

The publisher will search for js/*.tgz in your build artifacts and will npm publish each of them.

To create npm tarballs, you can use npm pack as part of your build and emit them to the js/ directory in your build artifacts. The version of the module is deduced from the name of the tarball.

To use this publisher, you will first need to store an npm.js publishing token in AWS Secrets Manager and supply the secret ARN when you add the publisher.

pipeline.publishToNpm({
  npmTokenSecret: { secretArn: 'my-npm-token-secret-arn' }
});

NuGet (.NET)

This publisher can publish .NET NuGet packages to nuget.org.

The publisher will search dotnet/**/*.nuget in your build artifacts and will publish each package to NuGet. To create .nupkg files, see Creating NuGet Packages. Make sure you output the artifacts under the dotnet/ directory.

To use this publisher, you will first need to store a NuGet API Key with "Push" permissions in AWS Secrets Manager and supply the secret ARN when you add the publisher.

Use pipeline.publishToNuGet will add a publisher to your pipeline:

pipeline.publishToNuGet({
  nugetApiKeySecret: { secretArn: 'my-nuget-token-secret-arn' }
});

Assembly Signature

Important: Limitations in the mono tools restrict the hash algorithms that can be used in the signature to SHA-1. This limitation will be removed in the future.

You can enable digital signatures for the .dll files enclosed in your NuGet packages. In order to do so, you need to procure a Code-Signing Certificate (also known as a Software Publisher Certificate, or SPC). If you don't have one yet, you can refer to Obtaining a new Code Signing Certificate for a way to create a new certificate entirely in the Cloud.

In order to enable code signature, change the way the NuGet publisher is added by adding an ICodeSigningCertificate for the codeSign key (it could be a CodeSigningCertificate construct, or you may bring your own implementation if you wish to use a pre-existing certificate):

pipeline.publishToNuGet({
  nugetApiKeySecret: { secretArn: 'my-nuget-token-secret-arn' },
  codeSign: codeSigningCertificate
});

Obtaining a new Code Signing Certificate

If you want to create a new certificate, the CodeSigningCertificate construct will provision a new RSA Private Key and emit a Certificate Signing Request in an Output so you can pass it to your Certificate Authority (CA) of choice:

Add a CodeSigningCertificate to your stack:

new delivlib.CodeSigningCertificate(stack, 'CodeSigningCertificate', {
  distinguishedName: {
    commonName: '<a name your customers would recognize>',
    emailAddress: '<your@email.address>',
    country: '<two-letter ISO country code>',
    stateOrProvince: '<state or province>',
    locality: '<city>',
    organizationName: '<name of your company or organization>',
    organizationalUnitName: '<name of your department within the origanization>',
  }
});

Deploy the stack:

$ cdk deploy $stack_name
...
Outputs:
$stack_name.CodeSigningCertificateXXXXXX = -----BEGIN CERTIFICATE REQUEST-----
...
-----END CERTIFICATE REQUEST-----

Forward the Certificate Signing Request (the value of the stack output that starts with -----BEGIN CERTIFICATE REQUEST----- and ends with -----END CERTIFICATE REQUEST-----) to a Certificate Authority, so they can provde you with a signed certificate.

Update your stack with the signed certificate obtained from the CA. The below example assumes you palced the PEM-encoded certificate in a file named certificate.pem that is in the same folder as file that uses the code:

// Import utilities at top of file:
import fs = require('fs');
import path = require('path');
// ...
new delivlib.CodeSigningCertificate(stack, 'CodeSigningCertificate', {
  distinguishedName: {
    commonName: '<a name your customers would recognize>',
    emailAddress: '<your@email.address>',
    country: '<two-letter ISO country code>',
    stateOrProvince: '<state or province>',
    locality: '<city>',
    organizationName: '<name of your company or organization>',
    organizationalUnitName: '<name of your department within the origanization>',
  },
  // Addin the signed certificate
  pemCertificate: fs.readFileSync(path.join(__dirname, 'certificate.pem'))
});

Redeploy your stack, so the self-signed certificate is replaced with the one received from your CA:
```
$ cdk deploy $stackName
```

Maven Central (Java)

This publisher can publish Java packages to Maven Central.

This publisher expects to find a local maven repository under the java/ directory in your build output artifacts. You can create one using the altDeploymentRepository option for mvn deploy (this assumes dist if the root of your artifacts tree):

$ mvn deploy -D altDeploymentRepository=local::default::file://${PWD}/dist/java

Use pipeline.publishToMaven to add this publisher to your pipeline:

pipeline.publishToMaven({
  mavenLoginSecret: { secretArn: 'my-maven-credentials-secret-arn' },
  signingKey: mavenSigningKey,
  stagingProfileId: '11a33451234521'
});

In order to configure the Maven publisher, you will need at least three pieces of information:

Maven Central credentials (mavenLoginSecret) stored in AWS Secrets Manager
GPG signing key (signingKey) to sign your Maven packages
Staging profile ID (stagingProfileId) assigned to your account in Maven Central.

The following sections will describe how to obtain this information.

GPG Signing Key

Since Maven Central requires that you sign your packages you will need to create a GPG key pair and publish it's public key to a well-known server:

This library includes a GPG key construct:

const mavenSigningKey = new delivlib.OpenPGPKeyPair(this, 'MavenCodeSign', {
  email: 'your-email@domain.com',
  identity: 'your-identity',
  secretName: 'maven-code-sign',
  pubKeyParameterName: 'mavenPublicKey',
  keySizeBits: 4096,
  expiry: '1y',
  version: 1.0
});

After you've deployed your stack once, you can go to the SSM Parameter Store console and copy the public key from the new parameter created by your stack under the specified secret name. Then, you should paste this key to any of the supported key servers (recommended: https://keyserver.ubuntu.com).

Sonatype Credentials

In order to publish to Maven Central, you'll need to follow the instructions in Maven Central's OSSRH Guide and create a Sonatype account and project via JIRA:

Create JIRA account
Create new project ticket
Once you have the user name and password of your Sonatype account, create an AWS Secrets Manager secret with a username and password key/value fields that correspond to your account's credentials.

Staging Profile ID

After you've obtained a Sonatype account and Maven Central project:

Log into https://oss.sonatype.org
Select "Staging Profiles" from the side bar (under "Build Promotion")
Click on the "Releases" staging profile that you registered
The URL of the page should change and include your profile ID. For example: https://oss.sonatype.org/#stagingProfiles;11a33451234521

This is the value you should assign to the stagingProfileId option.

PyPI (Python)

This publisher can publish modules to PyPI.

This publisher will publish all files under the python/ directory in your build output artifacts to PyPI using the following command:

twine upload --skip-existing python/**

To use this publisher, you will need to an account with PyPI. Then store your credentials in an AWS Secrets Manager secret, under the username and password fields.

Now, use pipeline.publishToPyPi to add this publisher to your pipeline:

pipeline.publishToPyPi({
  loginSecret: { secretArn: 'my-pypi-credentials-secret-arn' }
});

GitHub Releases

This publisher can package all your build artifacts, sign them and publish them to the "Releases" section of a GitHub project.

This publisher relies on two files to produce the release:

build.json a manifest that contains metadata about the release.
CHANGELOG.md (optional) the changelog of your project, from which the release notes are extracted. If not provided, no release notes are added to the release.

The file build.json is read from the root of your artifact tree. It should include the following fields:

{
  "name": "<project name>",
  "version": "<project version>",
  "commit": "<sha of commit>"
}

This publisher does the following:

Create a zip archive that contains the entire build artifacts tree under the name ${name}-${version}.zip.
Sign the archive using a GPG key and store it under ${name}-${version}.zip.sig
Check if there is already a git tag with v${version} in the GitHub repository. If there is, bail out successfully.
If there's a CHANGELOG.md file, and extract the release notes for ${version} (uses changelog-parser)
Create a GitHub release named v${version}, tag the specified ${commit} with the release notes from the changelog.
Attach the zip archive and signature to the release.

To add a GitHub release publisher to your pipeline, use the pipeline.publishToGitHub method:

pipeline.publishToGitHub({
  githubRepo: targetRepository,
  signingKey: releaseSigningKey
});

The publisher requires the following information:

The target GitHub project (githubRepo): see instructions on how to connect to a GitHub repository. It doesn't have to be the same repository as the source repository, but it can be.
A GPG signing key (signingKey): a delivlib.SigningKey object used to sign the zip bundle. Make sure to publish the public key to a well-known server so your users can validate the authenticity of your release (see GPG Signing Key for details on how to create a signing key pair and extract it's public key). You can either use

GitHub Pages

This publisher allows you to publish versioned static web-site content to GitHub Pages.

The publisher commits the entire contents of the docs/ directory into the root of the specified GitHub repository, and also under the ${version}/ directory of the repo (which allows users to access old versions of the docs if they wish).

NOTE: static website content can grow big. Therefore, this publisher will always force-push to the branch without history (history is preserved via the versions/ directory). Make sure you don't protect this branch against force-pushing or otherwise the publisher will fail.

This publisher depends on the following artifacts:

build.json: build manifest (see schema above)
docs/**: the static website contents

This is how this publisher works:

Read the version field from build.json
Clone the gh-pages branch of the target repository to a local working directory
Rsync the contents of docs/** both to versions/${version} and to / of the working copy.
Commit and push to the gh-pages branch on GitHub

NOTE: if docs/ contains a fully rendered static website, you should also include a .nojekyll file to bypass Jekyll rendering.

To add this publisher to your pipeline, use the pipeline.publishToGitHubPages method:

pipeline.publishToGitHubPages({
  githubRepo,
  sshKeySecret: { secretArn: 'github-ssh-key-secret-arn' },
  commitEmail: 'foo@bar.com',
  commitUsername: 'foobar',
  branch: 'gh-pages' // default
});

In order to publish to GitHub Pages, you will need the following pieces of information:

The target GitHub repository (githubRepo). See instructions on how to connect to a GitHub repository. It doesn't have to be the same repository as the source repository, but it can be.
SSH private key (sshKeySecret) for pushing to that repository stored in AWS Secrets Manager which is configured in your GitHub repository as a deploy key with write permissions.
Committer email (commitEmail) and username (commitUsername).

To create an ssh deploy key for your repository:

Follow this guide to produce a private/public key pair on your machine.
Add the deploy key to your repository with write permissions.
Create an AWS Secrets Manager secret and paste the private key as plaintext (not key/value).
Use the name of the AWS Secrets Manager secret in the sshKeySecret option.

Metrics

The Pipeline construct automatically creates the following metrics in CloudWatch for the configured pipelines. These are published under the namespace 'CDK/Delivlib'.

Execution Failures: The number of failures of the pipeline execution. When a pipeline execution fails, a '1' is recorded and forevery success, a '0' is recorded.

Metric Name: Failures Dimensions:
- Pipeline: The pipeline name in CodePipeline.
Action Failures: The number of failures per action per pipeline. An execution failure can be due to multiple actions failing. For every action failure, a '1' is recorded and for every success, a '0' is recorded.

Metric Name: Failures Dimensions:
- Pipeline: The pipeline name in CodePipeline.
- Action: THe name of the action that succeeded or failed.

Automatic Bumps and Pull Request Builds

GitHub Access

If your source repository is GitHub, in order to enable these features you will need to manually connect AWS CodeBuild to your GitHub account. Otherwise, you will receive the following error message:

No Access token found, please visit AWS CodeBuild console to connect to GitHub
(Service: AWSCodeBuild; Status Code: 400; Error Code: InvalidInputException;
Request ID: ab458603-6fd4-11e8-9310-ff116e0423f9)

To connect, go to the AWS CodeBuild console, click "Create Project", select a GitHub source and hit "Connect". There is no need to save the new project. This needs to be done once per account/region.

Automatic Bumps

A bump is the process of incrementing the version number of the project. When the version number is incremented and a commit is pushed to the master branch, the publishing actions will release the new version to all repositories.

This feature enables achieving full continuous delivery for libraries.

To enable automatic bumps, you will first need to determine how to perform a bump in your repository. What command should be executed in order to increment the version number, update change log, etc.

The bump command is expected to perform the bump and issue a commit and a tag to the local repository with the version number.

For JavaScript projects, the standard-version tool will do exactly that, so it is the recommended mechanism for such projects.

Once a bump is committed, the commit will be pushed either to a dedicated branch called bumps/VERSION or to a branch of your choosing such as master.

To set up bumps, simply call autoBump on your pipeline. The following example sets up a bump on the default schedule (12pm UTC daily) which will automatically push the to "master" (which will trigger a release).

const bump = pipeline.autoBump({
  bumpCommand: 'npm i && npm run bump',
  branch: 'master'
});

You can customize the environment used for running the bump script.

If a bump fails, the bump.alarm CloudWatch alarm will be triggered.

NOTE: there is currently no way for the bump command to indicate to the system that a bump is not needed (i.e. no changes have been made to the library).

Failure Notifications

Pipelines can be configured with notifications that will be sent on any failure in pipeline's stages. Notifications can be sent to either a Slack channel or a Chime room. The following code configures one of each -

// Slack
const teamChannel = new chatbot.SlackChannelConfiguration(this, {
  // ...
});
pipeline.notifyOnFailure(PipelineNotification.slack({
  channels: [teamChannel]
}));

// Chime
const teamRoomWebhook = 'https://hooks.chime.aws/incomingwebhooks/1c3588c7-623d-4799-af9b-8b1818fca779?token=cUMzOVA4OXl8MXxCaHJlZ0RUVm03TmZVMkpoTzlwa3NVbXJCam8tNWF3UGdzemVqZndsZERV';
pipeline.notifyOnFailure(PipelineNotification.chime({
  webhookUrl: [ teamRoomWebhook ]
}));

ECR Mirror

Builds commonly use Docker images from DockerHub as their base image. In fact, delivlib defaults its build image to jsii/superchain. However, DockerHub has throttles in place for the volume of unauthenticated and authenticated pulls. This can cause CodeBuild jobs that run frequently to fail from DockerHub's throttling.

The EcrMirror construct can be used to synchronize, on a specific schedule, Docker images between DockerHub and a local ECR registry in the AWS account.

new EcrMirror(this, 'RegistrySync', {
  sources: [
    MirrorSource.fromDockerHub('jsii/superchain'),
    MirrorSource.fromDockerHub('python:3.6'),
  ],
  dockerhubCredentials: // ...
  schedule: events.Schedule.cron( ... ),
})

You can also use the MirrorSource.fromDirectory() API if you would like to build a new Docker image based on a Dockerfile. The Dockerfile should be placed at the top level of the specified directory.

In addition to this, an EcrMirrorAspect is available that can walk the construct tree and replace all occurrences of Docker images in CodeBuild projects with ECR equivalents if they are found in the provided EcrMirror construct. This can be applied to an entire stack as so -

const stack = new MyStack(...);
// ...
Aspects.of(stack).add(new EcrMirrorAspect(ecrMirrorStack.mirror));

Contributing

See the contribution guide for details on how to submit issues, pull requests, setup a development environment and publish new releases of this library.

License

This library is licensed under the Apache 2.0 License.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

README.md

README.md

aws-delivlib

Pipeline Structure

Installation

Source

`repo`: Source Repository (required)

CodeCommit

GitHub

`branch`: Source Control Branch (optional)

Pull Request Builds

Build

`buildSpec`: Build Script (optional)

`buildImage`: Build container image (optional)

`env`: Build environment variables (optional)

Other Build Options

Tests

Publish

npm.js (JavaScript)

NuGet (.NET)

Assembly Signature

Obtaining a new Code Signing Certificate

Maven Central (Java)

GPG Signing Key

Sonatype Credentials

Staging Profile ID

PyPI (Python)

GitHub Releases

GitHub Pages

Metrics

Automatic Bumps and Pull Request Builds

GitHub Access

Automatic Bumps

Failure Notifications

ECR Mirror

Contributing

License

Files

README.md

Latest commit

History

README.md

File metadata and controls

aws-delivlib

Pipeline Structure

Installation

Source

repo: Source Repository (required)

CodeCommit

GitHub

branch: Source Control Branch (optional)

Pull Request Builds

Build

buildSpec: Build Script (optional)

buildImage: Build container image (optional)

env: Build environment variables (optional)

Other Build Options

Tests

Publish

npm.js (JavaScript)

NuGet (.NET)

Assembly Signature

Obtaining a new Code Signing Certificate

Maven Central (Java)

GPG Signing Key

Sonatype Credentials

Staging Profile ID

PyPI (Python)

GitHub Releases

GitHub Pages

Metrics

Automatic Bumps and Pull Request Builds

GitHub Access

Automatic Bumps

Failure Notifications

ECR Mirror

Contributing

License

`repo`: Source Repository (required)

`branch`: Source Control Branch (optional)

`buildSpec`: Build Script (optional)

`buildImage`: Build container image (optional)

`env`: Build environment variables (optional)