CodePipeline: How to reference nested CloudFormati

Question

Our CloudFormation templates are stored in GitHub. Inside CodePipeline we're using GitHub as our Source, but we can't reference nested CloudFormation Stacks when they're not stored on S3.

How can we reference CloudFormation nested Stacks when using GitHub as our source in CodePipeline?

If this is not possible, how can we upload the CloudFormation Templates from GitHub to S3 between the Source Stage (from GitHub) and the Deploy Stage in CodePipeline?

Answer 1

Besides the solution with a lambda stage, a simple approach is to use CodeBuild and AWS SAM.

In the main CloudFormation template (let's call it main.yaml), use 'Transform: AWS::Serverless-2016-10-31'

AWSTemplateFormatVersion: '2010-09-09'
Transform: AWS::Serverless-2016-10-31
Resources:
  NestedTemplate:
    AWS::CloudFormation::Stack
    Properties:
      TemplateUri: ./nested-template.yaml

Note that you just need to put the relative path to the child template instead of an absolution s3 uri.

Add a CodeBuild stage with the following buildspecification.yaml

version: 0.1
phases:
  build:
    commands:
      aws cloudformation package --template-file main.yaml --output-template-file transformed_main.yaml --s3-bucket my_bucket

artifacts:
  type: zip
  files:
    - transformed_main.yaml

The build command 'aws cloudformation package' will upload the nested-template.yaml to the s3 bucket 'my_bucket' and inject the absolute s3 uri to the transformed template.

In the CloudFormation deployment stage, use 'Create change set' and 'Execute change set' to create the stack. Note that 'Create or update stack' does not work for 'Transform: AWS::Serverless-2016-10-31'.

Here are the docs you may find useful:

1. http://docs.aws.amazon.com/cli/latest/reference/cloudformation/package.html
2. http://docs.aws.amazon.com/lambda/latest/dg/automating-deployment.html

The second doc shows how to deploy a lambda function, but it is essentially the same to reference a nested stack.

Answer 2

There are two approaches I can think of to reference nested CloudFormation Stacks from a GitHub source for a CodePipeline deployment:

1. pre-commit Git hook

Add a pre-commit client-side Git hook that runs aws cloudformation package on your template, committing a generated template with the S3 reference to your GitHub repository alongside the changes to the source template.

The benefit to this approach is that you can leverage the existing template-rewriting logic in aws cloudformation package, and you don't have to modify your existing CodePipeline configuration.

2. Lambda pipeline Stage

Add a Lambda-based pipeline Stage that extracts the specified nested-stack template file from the GitHub Source Artifact and uploads it to a specified location in S3 referenced in the parent stack template.

The benefit to this approach is that the Pipeline will remain entirely self-contained, without any extra pre-processing step required by the committer.

I've published a complete reference example implementation to wjordan/aws-codepipeline-nested-stack:

AWSTemplateFormatVersion: 2010-09-09
Description: Infrastructure Continuous Delivery with CodePipeline and CloudFormation, with a project containing a nested stack.
Parameters:
  ArtifactBucket:
    Type: String
    Description: Name of existing S3 bucket for storing pipeline artifacts
  StackFilename:
    Type: String
    Default: cfn-template.yml
    Description: CloudFormation stack template filename in the Git repo
  GitHubOwner:
    Type: String
    Description: GitHub repository owner
  GitHubRepo:
    Type: String
    Default: aws-codepipeline-nested-stack
    Description: GitHub repository name
  GitHubBranch:
    Type: String
    Default: master
    Description: GitHub repository branch
  GitHubToken:
    Type: String
    Description: GitHub repository OAuth token
  NestedStackFilename:
    Type: String
    Description: GitHub filename (and S3 Object Key) for nested stack template.
    Default: nested.yml
Resources:
  Pipeline:
    Type: AWS::CodePipeline::Pipeline
    Properties:
      RoleArn: !GetAtt [PipelineRole, Arn]
      ArtifactStore: 
        Type: S3
        Location: !Ref ArtifactBucket
      Stages:
      - Name: Source
        Actions:
        - Name: Source
          ActionTypeId:
            Category: Source
            Owner: ThirdParty
            Version: 1
            Provider: GitHub
          Configuration:
            Owner: !Ref GitHubOwner
            Repo: !Ref GitHubRepo
            Branch: !Ref GitHubBranch
            OAuthToken: !Ref GitHubToken
          OutputArtifacts: [Name: Template]
          RunOrder: 1
      - Name: Deploy
        Actions:
        - Name: S3Upload
          ActionTypeId:
            Category: Invoke
            Owner: AWS
            Provider: Lambda
            Version: 1
          InputArtifacts: [Name: Template]
          Configuration:
            FunctionName: !Ref S3UploadObject
            UserParameters: !Ref NestedStackFilename
          RunOrder: 1
        - Name: Deploy
          RunOrder: 2
          ActionTypeId:
            Category: Deploy
            Owner: AWS
            Version: 1
            Provider: CloudFormation
          InputArtifacts: [Name: Template]
          Configuration:
            ActionMode: REPLACE_ON_FAILURE
            RoleArn: !GetAtt [CFNRole, Arn]
            StackName: !Ref AWS::StackName
            TemplatePath: !Sub "Template::${StackFilename}"
            Capabilities: CAPABILITY_IAM
            ParameterOverrides: !Sub |
              {
                "ArtifactBucket": "${ArtifactBucket}",
                "StackFilename": "${StackFilename}",
                "GitHubOwner": "${GitHubOwner}",
                "GitHubRepo": "${GitHubRepo}",
                "GitHubBranch": "${GitHubBranch}",
                "GitHubToken": "${GitHubToken}",
                "NestedStackFilename": "${NestedStackFilename}"
              }
  CFNRole:
    Type: AWS::IAM::Role
    Properties:
      AssumeRolePolicyDocument:
        Statement:
        - Action: ['sts:AssumeRole']
          Effect: Allow
          Principal: {Service: [cloudformation.amazonaws.com]}
        Version: '2012-10-17'
      Path: /
      ManagedPolicyArns:
      # TODO grant least privilege to only allow managing your CloudFormation stack resources
      - "arn:aws:iam::aws:policy/AdministratorAccess"
  PipelineRole:
    Type: AWS::IAM::Role
    Properties:
      AssumeRolePolicyDocument:
        Statement:
        - Action: ['sts:AssumeRole']
          Effect: Allow
          Principal: {Service: [codepipeline.amazonaws.com]}
        Version: '2012-10-17'
      Path: /
      Policies:
        - PolicyName: CodePipelineAccess
          PolicyDocument:
            Version: '2012-10-17'
            Statement:
              - Action:
                - 's3:*'
                - 'cloudformation:*'
                - 'iam:PassRole'
                - 'lambda:*'
                Effect: Allow
                Resource: '*'
  Dummy:
    Type: AWS::CloudFormation::WaitConditionHandle
  NestedStack:
    Type: AWS::CloudFormation::Stack
    Properties:
      TemplateURL: !Sub "https://s3.amazonaws.com/${ArtifactBucket}/${NestedStackFilename}"
  S3UploadObject:
    Type: AWS::Lambda::Function
    Properties:
      Description: Extracts and uploads the specified InputArtifact file to S3.
      Handler: index.handler
      Role: !GetAtt LambdaExecutionRole.Arn
      Code:
        ZipFile: !Sub |
          var exec = require('child_process').exec;
          var AWS = require('aws-sdk');
          var codePipeline = new AWS.CodePipeline();
          exports.handler = function(event, context, callback) {
            var job = event["CodePipeline.job"];
            var s3Download = new AWS.S3({
                credentials: job.data.artifactCredentials,
                signatureVersion: 'v4'
            });
            var s3Upload = new AWS.S3({
                signatureVersion: 'v4'
            });
            var jobId = job.id;
            function respond(e) {
              var params = {jobId: jobId};
              if (e) {
                params['failureDetails'] = {
                  message: JSON.stringify(e),
                  type: 'JobFailed',
                  externalExecutionId: context.invokeid
                };
                codePipeline.putJobFailureResult(params, (err, data) => callback(e));
              } else {
                codePipeline.putJobSuccessResult(params, (err, data) => callback(e));
              }
            }
            var filename = job.data.actionConfiguration.configuration.UserParameters;
            var location = job.data.inputArtifacts[0].location.s3Location;
            var bucket = location.bucketName;
            var key = location.objectKey;
            var tmpFile = '/tmp/file.zip';
            s3Download.getObject({Bucket: bucket, Key: key})
              .createReadStream()
              .pipe(require('fs').createWriteStream(tmpFile))
              .on('finish', ()=>{
                exec(`unzip -p ${!tmpFile} ${!filename}`, (err, stdout)=>{
                if (err) { respond(err); }
                s3Upload.putObject({Bucket: bucket, Key: filename, Body: stdout}, (err, data) => respond(err));
              });
            });
          };
      Timeout: 30
      Runtime: nodejs4.3
  LambdaExecutionRole:
    Type: AWS::IAM::Role
    Properties:
      AssumeRolePolicyDocument:
        Version: '2012-10-17'
        Statement:
        - Effect: Allow
          Principal: {Service: [lambda.amazonaws.com]}
          Action: ['sts:AssumeRole']
      Path: /
      ManagedPolicyArns:
      - "arn:aws:iam::aws:policy/service-role/AWSLambdaBasicExecutionRole"
      - "arn:aws:iam::aws:policy/AWSCodePipelineCustomActionAccess"
      Policies:
      - PolicyName: S3Policy
        PolicyDocument:
          Version: '2012-10-17'
          Statement:
            - Effect: Allow
              Action:
                - 's3:PutObject'
                - 's3:PutObjectAcl'
              Resource: !Sub "arn:aws:s3:::${ArtifactBucket}/${NestedStackFilename}"