Orchestrates large Cloudformation deployments

pip install cloudformation-seed==0.12.0


Cloudformation Seed

Build Status


This is a script that will help you deploy your Cloudformation project without hassle:

  • Handle Cloudformation deployments of any scale
  • Allow to do multiple deployments of the same code with a different installation name
  • Automate Lambda code handling
  • Get rid of hard dependencies of Cloudformation Exports, instead pass around Output values between stacks
  • Package the whole deployment in a Docker image and version it

It will:

  • Automatically create an S3 bucket according to the project name
  • Upload the Cloudformation templates into the bucket
  • Package and checksum your Lambda code and upload it into the bucket
  • Upload arbitrary artifacts into the bucket so that they are available to your deployment
  • Create and manage Cloudformation stacks
  • Create, roll out and manage Stacksets


You need a Mac or a Linux machine/VM to run the Seed. Windows is not supported, but it may work on Windows too if you really want it to.

You need to have Docker on the workstation.

Every Cloudformation template you use has to have 4 mandatory parameters that will be supplied by the Seed:

  1. TemplatesS3Bucket - the Seed will automatically create an S3 bucket and every template will have its name passed down in this parameter, so it can be made available to Lambda functions, autoscaling groups, e.t.c.
  2. InstallationName - installation name is what makes you able to deploy your project multiple times without name clashes. Every template will have it in this parameter and you have to use it in the names of your resources to make them unique across multiple installations
  3. RuntimeEnvironment - name of the runtime environment (read Deployment configuration)
  4. Route53ZoneDomain - DNS domain associated with your deployment. The Seed doesn't require it to exist, you can use it as part of your resource naming convention

Here's a snippet you can copy and paste:

    Type: String
    Description: S3 Bucket with the components templates
    Type: String
    Description: Unique DNS stack installation name
    Type: String
    Description: The runtime environment config tag
    Default: dev
    Type: String
    Description: Route53 zone domain that represents the environment

Quick start

First things first:

  1. Create a new directory for your project
  2. Copy everything from the examples directory to the root of the project
  3. Edit parameters/dev.yaml to your needs
  4. Add more templates with .cf.yaml extensions under the cloudformation directory and include them in parameters/dev.yaml


Authenticate to AWS using your method of choice, make sure that you have set the AWS Region you need for deployment. Run cloudformation-seed -c my-project -i x0 -e dev -d my.domain.cld deploy


Take the dockerfiles and makefiles from the examples directory and massage them around to suit your needs.

Deep dive

Seed bucket

The Seed will automatically create an S3 bucket for operating the deployment. The name of the bucket is derived from the installation name and project name from Makefile.particulars. The name of the bucket will be passed down to every Cloudformation template in your deployment as TemplatesS3Bucket

Deployment configuration

The RUNTIME_ENVIRONMENT=dev clause in the deployment directive points to the configuration file dev.yaml located under the parameters directory.

You can have multiple runtime environments for the same project with different configuration, for example if you have dev, test and prod environments that reuse the same Cloudformation but need different configuration, for example VPC and subnet IDs.

A runtime environment is a YAML file that:

  • defines the sequence in which the Cloudformation stacks will be deployed
  • sets parameters for the Cloudformation stacks

The runtime environment contains two sections:


In this section you can specify Cloudformation parameters that will be picked up by every stack in the deployment as a default value (i.e. if a stack has the same parameter on it it will take precedence)


  VpcId: vpc-00000000

You can use !StackOutput (read below) in common-parameters and it will work as expected.

Instead of common-parameters you can also use YAML anchors like this:

SAMLUsername: &SAML_USERNAME okta_sso

  - name: centralservices-iam-set
    type: stackset
    template: sets/
      SSMLogsLambdaS3Key: !LambdaZip
      SAMLUsername: *SAML_USERNAME

You can also tag your stacks/stacksets by defining your tags as a dictionary and referencing them using the YAML anchors within your stacks like this:

tags_a: &TAGSA
  testkey1: testvalue1
  testkey2: testvalue2

tags_b: &TAGSB
  testkey3: testvalue3

  - name: example-stackset-template
    type: stackset
    template: sets/
      - account: '000000000000'
    tags: *TAGA

  - name: my-project-kms-decrypt-lambda
    template: support/
      LambdaSourceS3Key: !LambdaZip
    tags: *TAGSB


Main configuration where you describe the Cloudformation stacks you want to deploy.


  - name: in-cld-managed-zone                            # name of the CF stack, INSTALLATION_NAME will be prepended
    template: centralservices/           # CF template relative to cloudformation directory
    parameters:                                          # Parameters to the CF stack
      ManagedZoneDomainName: in.cld
      ManagingAccountArns:                               # List parameters turn into comma-separated values
        - arn:aws:iam::000000000000:root
        - arn:aws:iam::111111111111:root
        - arn:aws:iam::222222222222:root

  - name: in-cld-provisioning                            # name of CF stack, INSTALLATION_NAME will be prepended
    template: centralservices/   # CF template relative to cloudformation directory
      LambdaSourceS3Key: !LambdaZip     # points to the lambda function under src/provisionR53 (read below)
      SharedServiceR53ZoneRoleArn: !StackOutput in-cld-managed-zone.ManagedZoneCrossAccountRole    # will take the output called ManagedZoneCrossAccountRole from the above stack called in-cld-managed-zone
      Route53DomainName: !StackOutput in-cld-managed-zone.ManagedZoneDomainName
      ExportOutputs: 'false'                             # put numbers and booleans in quotes

  - name: centralservices-iam-set
    type: stackset                                       # set type to stackset
    template: sets/
    parameters:                                          # parameters to the StackSet
      SSMLogsLambdaS3Key: !LambdaZip
      SAMLUsername: *SAML_USERNAME
    pilot:                                               # when StackSet is updated only update instances in these accounts
        - '000000000000'
    rollout:                                             # manage StackSet instances
      - account: '000000000000'
        override:                                        # parameter override
          Route53ZoneDomain: prod.innablr.lan
      - account: '111111111111'
        regions:                                         # in this account it goes into two regions
          - ap-southeast-2
          - eu-west-1
          Route53ZoneDomain: preprod.innablr.lan
      - account: '222222222222'
          Route53ZoneDomain: dev.innablr.lan
      - account: '999999999999'
        regions: []                                      # this is how you delete an instance
          Route53ZoneDomain: dontwant.innablr.lan

Automated Lambda functions

If your deployment contains Lambda function they can be handled by the Seed automatically. In the examples directory you can find an example of a Lambda function called kmsParameters

  1. Create a directory under src for your Lambda, say kmsParameters
  2. Do the development
  3. Create a Makefile in the directory you have created and make sure that the default target of the Makefile produces a zip-file, say
  4. In your runtime environment configuration use !LambdaZip to pass the zip-file name to the CloudFormation template (see the example above)

If your Lambda function is used in a StackSet and needs to be available from other AWS accounts make sure that you give access to the Seed bucket from those accounts. Refer to the stack that is included in the examples.

Arbitrary artifacts

If you want to include any configuration objects for your software or other relatively lightweight artifacts you can create a directory called config/<runtime_environment> under the root of your project and anything you put in this directory will be uploaded in the Seed S3 bucket under a key called config.

Let's say you have config/dev/myapp_cert.pem and you deploy a runtime configuration called dev. The file will be uploaded in the bucket as config/myapp_cert.pem.

Configuration tags

In the runtime environment configuration you can use the following tags in stack parameters specification:

  1. !LambdaZip - will pass the correct S3 key to the uploaded, so you can use it in your Lambda resources together with TemplatesS3Bucket

  2. !CloudformationTemplate support/ - works very similar to !Lambdazip but for Cloudformation templates. Will pass down the correct S3 key to the specified CloudFormation stack

  3. !StackOutput stack-name.OutputName - will read the corresponding output from the specified stack and pass it down here. The stack needs to have been created above in the sequence.

  4. !ArtifactVersion, !ArtifactRepo and !ArtifactImage - these three tags are used together with a release manifest in release management

Release management can read a release manifest file if you specify it in the -m commandline argument. Release manifest contains images, their versions and other information about the software that is being deployed by the Seed. You can then inform your Cloudformation stacks about the versions and images you are deploying using the !ArtifactVersion, !ArtifactRepo and !ArtifactImage tags in the runtime environment configuration.

More documentation about release management is coming soon.