pipeline_manual_approvals_onestage.jpgRecently, AWS announced that it added manual approval actions to AWS CodePipeline. In doing so, you can now model your entire software delivery process – whether it’s entirely manual or a hybrid of automated and manual approval actions.
In this post, I describe how you can add manual approvals to an existing pipeline – manually or via CloudFormation – to minimize your CodePipeline costs.


The AWS CodePipeline pricing model is structured to incentivize two things:

  • Frequent Code Commits
  • Long-lived Pipelines

This is because AWS charges $1 per active pipeline per month. Therefore, if you were to treat these pipelines as ephemeral, you’d likely be paying more than you might be otherwise consuming. While in experimentation mode, you might be regularly launching and terminating pipelines as you determine the appropriate stages and actions for an application/service, once you’ve established this pipeline, the change lifecycle is likely to be much less.
Since CodePipeline uses compute resources, AWS had to make a decision on whether they incentivize frequent code commits or treat pipelines ephemerally – as they do with other resources like EC2. If they’d chosen to charge by the frequency activity then it could result in paying more when committing more code – which would be a very bad thing since you want developers to be committing code many times a day.


While we tend to prefer an immutable approach in most things when it comes to the infrastructure, the fact is that different parts of your system will change at varying frequencies. This is the case with your pipelines. Once your pipelines have been established, typically, you might make add, edit, or remove some stages and actions but probably not every day.
Our “workaround” is to use CloudFormation’s update capability to modify our pipeline’s stages and actions without incurring the additional $1 that we’d get charged if we were to launch a new active pipeline.
The best way to apply these changes is to make the minimum required changes in the template so that errors are prevalent if they do occur.

Manual Approvals

There are many reasons your software delivery workflow might require manual approvals including exploratory testing, visual inspection, change advisory boards, code reviews, etc.
Some other reasons for manual approvals include canary and blue/green deployments – where you might make final deployment decisions once some user or deployment testing is complete.
With manual approvals in CodePipeline, you can now make the approval process a part of a fully automated software delivery process.

Create and Connect to a CodeCommit Repository

Follow these instructions for creating and connecting to an AWS CodeCommit repository: Create and Connect to an AWS CodeCommit Repository. Take note of the repository name as you’ll be using it as a CloudFormation user parameter later. The default that I use the lab is called codecommit-demo but you can modify this CloudFormation parameter.

Launch a Pipeline

Click the button below to launch a CloudFormation stack that provisions AWS CodePipeline with some default Lambda Invoke actions.

Once the CloudFormation has launched successfully, click on the link next to the PipelineUrl Output from your CloudFormation stack. This launches your pipeline. You should see a pipeline similar to the one in the figure below.

Update a Pipeline

To update your pipeline, click on the Edit button at the top of the pipeline in CodePipeline. Then, click the (+) Stage link in between the Staging the Production stage. Enter the name ExploratoryTesting for the stage name, then click the (+) Action link. The add action window displays. Choose the new Approval Action category from the drop down and enter the other required and optional fields, as appropriate. Finally, click the Add action button.


Once you’ve done this, click on the Release change button. Once it goes through the pipeline stages and actions, it transitions to the Exploratory Testing stage where your pipeline should look similar to the figure below.
At this time, if your SNS Topic registered with the pipeline is linked to an email address, you’ll receive an email message that looks similar to the one below.
As you can see, you can click on the link to be brought to the same pipeline where you can approve or reject the “stage”.

Applying Changes in CloudFormation

You can apply the same updates to CodePipeline that you had previously manually performed in code using CloudFormation update-stack. We recommend you minimize the incremental number of changes you apply using CloudFormation so that they are specific to CodePipeline changes. This is because limiting your change sets often results in limiting the amount of time you spend troubleshooting any problems.
Once you’ve manually added the new manual approval stage and action, you can use your AWS CLI to get the JSON configuration that you can use in your CloudFormation update template. To do this, run the following command substituting {YOURPIPELINENAME} with the name of your pipeline.

aws codepipeline get-pipeline --name {YOURPIPELINENAME} >pipeline.json

You’ll also notice that this command pipes the output to a file that you can use as a means of copying and formatting as part of stage and action configuration in CodePipeline. For example, the difference between the initial pipeline and the updated pipeline is shown in the JSON configuration below.

                  "CustomData":"Approval or Reject this change after running Exploratory Tests"

You can take this code and add it to a new CloudFormation template so that it’s between the Staging and Production stages. Once you’ve done this, go back to your command line and run the update-stack command from your AWS CLI. An example is shown below. You’ll replace the {CFNSTACKNAME} with your stack name. If you want to make additional changes to the new stack, you can download the CloudFormation template and update it to an S3 location you control.

aws cloudformation update-stack --stack-name {CFNSTACKNAME} --template-url https://s3.amazonaws.com/stelligent-public/cloudformation-templates/github/labs/codepipeline/codepipeline-updates-after.json --region us-east-1 --capabilities="CAPABILITY_IAM" --parameters ParameterKey=RepositoryBranch,UsePreviousValue=true ParameterKey=RepositoryName,UsePreviousValue=true ParameterKey=S3BucketLambdaFunction,UsePreviousValue=true ParameterKey=SNSTopic,UsePreviousValue=true

By running this command against the initial stack, you’ll see the same updates that you’d manually defined previously. The difference is that it’s defined in code which means you can version, test and deploy changes.
An alternative approach is to manually apply the changes using Update Stack through from your CloudFormation stack. You’ll enter the new CloudFormation template as an input and CloudFormation will determine which changes it will apply to your infrastructure. You see a screenshot of the change that CloudFormation will apply below.


By incorporating manual approvals into your software delivery process, you can fully automate its workflow. You learned how you can apply changes to your pipeline using CloudFormation as a way of minimizing your costs while providing a repeatable, reliable update process through code.

Sample Code

The code for the examples demonstrated in this post are located at https://github.com/stelligent/cloudformation_templates/tree/master/labs/codepipeline. Let us know if you have any comments or questions @stelligent or @paulduvall.
Stelligent is hiring! Do you enjoy working on complex problems like figuring out ways to automate all the things as part of a deployment pipeline? Do you believe in the “everything-as-code” mantra? If your skills and interests lie at the intersection of DevOps automation and the AWS cloud, check out the careers page on our website.



My colleagues at Stelligent including Eric Kascic and Casey Lee provided some use cases for manual approvals.

Stelligent Amazon Pollycast
Voiced by Amazon Polly