envato/stack_master

StackMaster is a CLI tool to manage CloudFormation stacks, with the following features:

• Synchronous visibility into stack updates. See exactly what is changing and what will happen before agreeing to apply a change.
• Dynamic parameter resolvers.
• Template compiler support for YAML and SparkleFormation.

Stack updates can cause a lot of damage if applied blindly. StackMaster helps with this by providing the operator with as much information about the proposed change as possible before asking for confirmation to continue. That information includes:

• Template body and parameter diffs.
• Change sets are displayed for review.
• Once the diffs & change set have been reviewed, the change can be applied and stack events monitored.
• Stack events will be displayed until an end state is reached.

Stack parameters can be dynamically resolved at runtime using one of the built in parameter resolvers. Parameters can be sourced from other stacks outputs, or by querying various AWS APIs to get resource ARNs, etc.

Installation

System-wide

gem install stack_master

# if you want linting capabilities:
pip install cfn-lint

Bundler

• pip install cfn-lint if you need lint functionality
• Add gem 'stack_master' to your Gemfile.
• Run bundle install
• Run bundle exec stack_master init to generate a directory structure and stack_master.yml file

Configuration

Stacks are defined inside a stack_master.yml YAML file. When running stack_master, it is assumed that this file will exist in the current working directory, or that the file is passed in with --config /path/to/stack_master.yml. Here's an example configuration file:

region_aliases:
  production: us-east-1
  staging: ap-southeast-2
stack_defaults:
  tags:
    application: my-awesome-app
  role_arn: service_role_arn
region_defaults:
  us-east-1:
    tags:
      environment: production
    notification_arns:
      - test_arn
  ap-southeast-2:
    tags:
      environment: staging
stacks:
  production:
    myapp-vpc:
      template: myapp_vpc.rb
      tags:
        purpose: front-end
    myapp-db:
      template: myapp_db.rb
      stack_policy_file: db_stack_policy.json
      tags:
        purpose: back-end
    myapp-web:
      template: myapp_web.rb
      tags:
        purpose: front-end
  staging:
    myapp-vpc:
      template: myapp_vpc.rb
      allowed_accounts: '123456789'
      tags:
        purpose: front-end
    myapp-db:
      template: myapp_db.rb
      allowed_accounts:
        - '1234567890'
        - '9876543210'
      tags:
        purpose: back-end
    myapp-web:
      template: myapp_web.rb
      tags:
        purpose: front-end
  eu-central-1:
    myapp-vpc:
      template: myapp_vpc.rb
      tags:
        purpose: vpc

S3

StackMaster can optionally use S3 to store the templates before creating a stack. This requires you to configure an S3 bucket in stack_master.yml:

stack_defaults:
  s3:
    bucket: my_bucket_name
    prefix: cfn_templates/my-awesome-app
    region: us-west-2

Additional files can be configured to be uploaded to S3 alongside the templates:

stacks:
  production:
    myapp-vpc:
      template: myapp_vpc.rb
      files:
        - userdata.sh

Directories

• templates - CloudFormation, SparkleFormation or CfnDsl templates.
• parameters - Parameters as YAML files.
• secrets - encrypted secret files.
• policies - Stack policy JSON files.

Templates

StackMaster supports CloudFormation templates in plain JSON or YAML. Any .yml or .yaml file will be processed as YAML, while any .json file will be processed as JSON. Additionally, YAML files can be pre-processed using ERB and compile-time parameters.

Ruby DSLs

By default, any template ending with .rb will be processed as a SparkleFormation template. However, if you want to use CfnDsl templates you can add the following lines to your stack_master.yml.

template_compilers:
  rb: cfndsl

Parameters

By default, parameters are loaded from multiple YAML files, merged from the following lookup paths from bottom to top:

• parameters/[stack_name].yaml
• parameters/[stack_name].yml
• parameters/[region]/[stack_name].yaml
• parameters/[region]/[stack_name].yml
• parameters/[region_alias]/[stack_name].yaml
• parameters/[region_alias]/[stack_name].yml

A simple parameter file could look like this:

key_name: myapp-us-east-1

Alternatively, a parameter_files array can be defined to explicitly list parameter files that will be loaded. If parameter_files are defined, the automatic search locations will not be used.

parameters_dir: parameters # the default
stacks:
  us-east-1:
    my-app:
      parameter_files:
        - my-app.yml # parameters/my-app.yml

Parameters can also be defined inline with stack definitions:

stacks:
  us-east-1:
    my-app:
      parameters:
        VpcId:
          stack_output: my-vpc/VpcId

Compile Time Parameters

Compile time parameters can be defined in a stack's parameters file, using the key compile_time_parameters. Keys in parameter files are automatically converted to camel case.

As an example:

# parameters/some_stack.yml
vpc_cidr: 10.0.0.0/16
compile_time_parameters:
  subnet_cidrs:
    - 10.0.0.0/28
    - 10.0.2.0/28

SparkleFormation

Compile time parameters can be used for SparkleFormation templates. It conforms and allows you to use the Compile Time Parameters feature.

CloudFormation YAML ERB

Compile time parameters can be used to pre-process YAML CloudFormation templates. An example template:

# templates/some_stack_template.yml.erb
Parameters:
  VpcCidr:
    Type: String
Resources:
  Vpc:
    Type: AWS::EC2::VPC
    Properties:
      CidrBlock: !Ref VpcCidr
  # Given the two subnet_cidrs parameters, this creates two resources:
  # SubnetPrivate0 with a CidrBlock of 10.0.0.0/28, and
  # SubnetPrivate1 with a CidrBlock of 10.0.2.0/28
  <% params["SubnetCidrs"].each_with_index do |cidr, index| %>
  SubnetPrivate<%= index %>:
    Type: AWS::EC2::Subnet
    Properties:
      VpcId: !Ref Vpc
      AvailabilityZone: ap-southeast-2
      CidrBlock: <%= cidr %>
  <% end %>

Parameter Resolvers

Parameter values can be sourced dynamically using parameter resolvers.

One benefit of using parameter resolvers instead of hard coding values like VPC IDs and resource ARNs is that the same configuration works cross region/account, even though the resolved values will be different.

Cross-account parameter resolving

One way to resolve parameter values from different accounts to the one StackMaster runs in, is to assume a role in another account with the relevant IAM permissions to execute successfully.

This is supported in StackMaster by specifying the role and account properties for the parameter resolver in the stack's parameters file.

All parameter resolvers are supported.

vpc_peering_id:
  role: cross-account-parameter-resolver
  account: 1234567890
  stack_output: vpc-peering-stack-in-other-account/peering_name

an_array_param:
  role: cross-account-parameter-resolver
  account: 1234567890
  stack_outputs:
    - stack-in-account1/output
    - stack-in-account1/another_output

another_array_param:
  - role: cross-account-parameter-resolver
    account: 1234567890
    stack_output: stack-in-account1/output
  - role: cross-account-parameter-resolver
    account: 0987654321
    stack_output: stack-in-account2/output

my_secret:
  role: cross-account-parameter-resolver
  account: 1234567890
  parameter_store: ssm_parameter_name

An example of use case where cross-account parameter resolving is particularly useful is when setting up VPC peering where you need the VPC ID of the peer. Without the ability to assume a role in another account, the only option was to hard code the peer's VPC ID.

Stack Output

The stack output parameter resolver looks up outputs from other stacks in the same or different region. The expected format is [(region|region-alias):]stack-name/(OutputName|output_name).

vpc_id:
  # Output from a stack in the same region
  stack_output: my-vpc-stack/VpcId

bucket_name:
  # Output from a stack in a different region
  stack_output: us-east-1:init-bucket/bucket_name

zone_name:
  # Output from a stack in a different region using its alias
  stack_output: global:hosted-zone/ZoneName

This is the most used parameter resolver because it enables stacks to be split up into their separated concerns (VPC, web, database etc) with outputs feeding into parameters of dependent stacks.

Secret

Note: The GPG parameter resolver has been extracted into a dedicated gem. Please install and follow the instructions for the stack_master-gpg_parameter_resolver gem.

Parameter Store

An alternative to the secrets store, uses the AWS SSM Parameter store to protect secrets. Expects a parameter of either String or SecureString type to be present in the same region as the stack. You can store the parameter using a command like this

aws ssm put-parameter --region <region> --name <parameter name> --value <secret> --type (String|SecureString)

When doing so make sure you don't accidentally store the secret in your .bash_history and you will likely want to set the parameter to NoEcho in your template.

db_password:
  parameter_store: ssm_parameter_name

1Password Lookup

An alternative to the secrets store is accessing 1password secrets using the 1password cli (op). You declare a 1password lookup with the following parameters in your parameters file:

# parameters/database.yml
database_password:
  one_password:
    title: production database
    vault: Shared
    type: password

1password stores the name of the secret in the title. You can pass the vault you expect the secret to be in. Currently we support two types of secrets, passwords and secureNotes. All values must be declared, there are no defaults.

For more information on 1password cli please see here

EJSON Store

ejson is a tool to manage asymmetrically encrypted values in JSON format. This allows you to keep secrets securely in git/Github and gives anyone the ability the capability to add new secrets without requiring access to the private key. ejson_wrapper encrypts the underlying EJSON private key with KMS and stores it in the ejson file as _private_key_enc. Each time an ejson secret is required, the underlying EJSON private key is first decrypted before passing it onto ejson to decrypt the file.

First, generate an ejson file with ejson_wrapper, specifying the KMS key ID to be used:

gem install ejson_wrapper
ejson_wrapper generate --region us-east-1 --kms-key-id [key_id] --file secrets/production.ejson

Then, add the ejson_file argument to your stack in stack_master.yml:

stacks:
  us-east-1:
    my_app:
      template: my_app.json
      ejson_file: production.ejson

finally refer to the secret key in the parameter file, i.e. parameters/my_app.yml:

my_param:
  ejson: "my_secret"

Additional configuration options:

• ejson_file_region The AWS region to attempt to decrypt private key with
• ejson_file_kms Default: true. Set to false to use ejson without KMS.

Security Group

Looks up a security group by name and returns the ARN.

ssh_sg:
  security_group: SSHSecurityGroup

Security Groups

An array of security group names can also be provided.

ssh_sg:
  security_groups:
    - SSHSecurityGroup
    - WebAccessSecurityGroup

SNS Topic

Looks up an SNS topic by name and returns the ARN.

notification_topic:
  sns_topic_name: PagerDuty

Latest AMI by Tag

Looks up the latest AMI ID by a given set of tags.

web_ami:
  latest_ami_by_tags: role=web,application=myapp

Note that the corresponding array resolver is named latest_amis_by_tags.

Latest AMI by attribute

Looks up the latest AMI ID by a given set of attributes. By default it will only return AMIs from the account the stack is created in, but you can specify the account ID or certain keywords mentioned in the aws documentation

This selects the latest wily hvm AMI from Ubuntu (using the account id):

bastion_ami:
  latest_ami:
    owners: 099720109477
    filters:
      name: ubuntu/images/hvm/ubuntu-wily-15.10-amd64-server-*

A set of possible attributes is available in the AWS documentation.

Any value can be an array of possible matches.

Latest Container from Repository

Looks up the a Container Image from an ECR repository. By default this will return the latest container in a repository. If tag is specified we return the sha digest of the image with this tag. This avoids the issue where CloudFormation won't update a Task Definition if we use a tag such as latest, because it only updates resources if a parameter has changed. This allows us to tag an image and deploy the latest version of that image via CloudFormation and avoids versioning our tags and having to store the metadata about the latest tag version somewhere.

Returns the docker repository URI, i.e. aws_account_id.dkr.ecr.region.amazonaws.com/container@sha256:digest

container_image_id:
  latest_container:
    repository_name: nginx # Required. The name of the repository
    registry_id: "012345678910" # The AWS Account ID the repository is located in. Defaults to the current account's default registry. Must be in quotes.
    region: us-east-1 # Defaults to the region the stack is located in
    tag: production # By default we'll find the latest image pushed to the repository. If tag is specified we return the sha digest of the image with this tag

Environment Variable

Lookup an environment variable:

db_username:
  env: DB_USERNAME

ACM Certificates

Find an ACM certificate by domain name:

cert:
  acm_certificate: www.example.com

Custom parameter resolvers

New parameter resolvers can be created in a separate gem.

To create a resolver named my_resolver:

• Create a new gem using your favorite tool
• The gem structure must contain the following path:

lib/stack_master/parameter_resolvers/my_resolver.rb

• That file needs to contain a class named StackMaster::ParameterResolvers::MyResolver that implements a resolve method and an initializer taking 2 parameters :

module StackMaster
  module ParameterResolvers
    class MyResolver < Resolver
      array_resolver # Also create a MyResolvers resolver to handle arrays

      def initialize(config, stack_definition)
        @config = config
        @stack_definition = stack_definition
      end

      def resolve(value)
        value
      end
    end
  end
end

• Note that the filename and classname are both derived from the resolver name passed in the parameter file. In our case, the parameters YAML would look like:

vpc_id:
  my_resolver: dummy_value

Resolver Arrays

Most resolvers support taking an array of values that will each be resolved. Unless stated otherwise in the documentation, the array version of the resolver will be named with the pluralized name of the original resolver.

When creating a new resolver, one can automatically create the array resolver by adding a array_resolver statement in the class definition, with an optional class name if different from the default one.

module StackMaster
  module ParameterResolvers
    class MyResolver < Resolver
      array_resolver class_name: 'MyCustomArrayResolver'
      ...
    end
  end
end

In that example, using the array resolver would look like:

my_parameter:
  my_custom_array_resolver:
    - value1
    - value2

Array parameter values can include nested parameter resolvers.

For example, given the following parameter definition:

my_parameter:
  - stack_output: my-stack/output # value resolves to 'value1'
  - value2

The parameter value will resolve to:

my_parameter: 'value1,value2'

ERB Template Files in SparkleFormation templates

An extension to SparkleFormation is the user_data_file! method, which evaluates templates in templates/user_data/[file_name]. Most of the usual SparkleFormation methods are available in user data templates. Example:

# templates/user_data/app.erb
REGION=<%= region! %>
ROLE=<%= role %>

And used like this in SparkleFormation templates:

# templates/app.rb
  user_data user_data_file!('app.erb', role: :worker)

You can also use the joined_file! method which evaluates templates in templates/config/[file_name]. It is similar to user_data_file! but doesn't do base64 encoding. Example:

# templates/config/someconfig.conf.erb
my_variable=<%= ref!(:foo) %>
my_other_variable=<%= account_id! %>

# templates/ecs_task.rb
container_definitions array!(
  -> {
    command array!(
      "-e",
      joined_file!('someconfig.conf.erb')
    )
    ...
  }
)

Compiler Options & Alternate Template Directories

StackMaster allows you to separate your stack definitions and parameters from your templates by way of a template_dir key in your stack_master.yml. You can also pass compiler-specific options to the template compiler to further customize SparkleFormation or CfnDsl's behavior. Combining the 2 lets you move your SFN templates away from your stack definitions. For example, if your project is laid out as:

project-root
|-- envs
  |-- env-1
    |-- stack_master.yml
  |-- env-2
    |-- stack_master.yml
|-- sparkle
  |-- templates
    |-- my-stack.rb

Your env-1/stack_master.yml files can reference common templates by setting:

template_dir: ../../sparkle/templates
stack_defaults:
  compiler_options:
    sparkle_path: ../../sparkle

stacks:
  us-east-1:
    my-stack:
      template: my-stack.rb

Loading SparklePacks

SparklePacks can be pre-loaded using compiler options. This requires the name of a rubygem to require followed by the name of the SparklePack, which is usually the same name as the Gem.

stacks:
  us-east-1:
    my-stack:
      template: my-stack-with-dynamic.rb
      compiler_options:
        sparkle_packs:
          - vpc-sparkle-pack

The template can then simply load a dynamic from the sparkle pack like so:

SparkleFormation.new(:my_stack_with_dynamic) do
   dynamic!(:sparkle_pack_dynamic)
end

Note though that if a dynamic with the same name exists in your templates/dynamics/ directory it will get loaded since it has higher precedence.

Templates can be also loaded from sparkle packs by defining sparkle_pack_template. The name corresponds to the registered symbol rather than specific name. That means for a sparkle pack containing:

SparkleFormation.new(:template_name) do
  ...
end

we can use stack defined as follows:

stacks:
  us-east-1:
    my-stack:
      template: template_name
      compiler: sparkle_formation
      compiler_options:
        sparkle_packs:
          - some-sparkle-pack
        sparkle_pack_template: true

Allowed accounts

The AWS account the command is executing in can be restricted to a specific list of allowed accounts. This is useful in reducing the possibility of applying non-production changes in a production account. Each stack definition can specify the allowed_accounts property with an array of AWS account IDs or aliases the stack is allowed to work with.

This is an opt-in feature which is enabled by specifying at least one account to allow.

Unlike other stack defaults, the allowed_accounts property values specified in the stack definition override values specified in the stack defaults (i.e., other stack property values are merged together with those specified in the stack defaults). This allows specifying allowed accounts in the stack defaults (inherited by all stacks) and override them for specific stacks. See below example config for an example.

stack_defaults:
  allowed_accounts: '555555555'
stacks:
  us-east-1:
    myapp-vpc: # only allow account 555555555 (inherited from the stack defaults)
      template: myapp_vpc.rb
      tags:
        purpose: front-end
    myapp-db:
      template: myapp_db.rb
      allowed_accounts: # only allow these accounts (overrides the stack defaults)
        - '1234567890'
        - my-account-alias
      tags:
        purpose: back-end
    myapp-web:
      template: myapp_web.rb
      allowed_accounts: [] # allow all accounts (overrides the stack defaults)
      tags:
        purpose: front-end
    myapp-redis:
      template: myapp_redis.rb
      allowed_accounts: '888888888' # only allow this account (overrides the stack defaults)
      tags:
        purpose: back-end

In the cases where you want to bypass the account check, there is the StackMaster flag --skip-account-check that can be used.

Commands

stack_master help # Display up to date docs on the commands available
stack_master init # Initialises a directory structure and stack_master.yml file
stack_master list # Lists stack definitions
stack_master apply [region-or-alias] [stack-name] # Create or update a stack
stack_master apply [region-or-alias] [stack-name] [region-or-alias] [stack-name] # Create or update multiple stacks
stack_master apply [region-or-alias] # Create or update stacks in the given region
stack_master apply # Create or update all stacks
stack_master --changed apply # Create or update all stacks that have changed
stack_master --yes apply [region-or-alias] [stack-name] # Create or update a stack non-interactively (forcing yes)
stack_master diff [region-or-alias] [stack-name] # Display a stack template and parameter diff
stack_master drift [region-or-alias] [stack-name] # Detects and displays stack drift using the CloudFormation Drift API
stack_master delete [region-or-alias] [stack-name] # Delete a stack
stack_master events [region-or-alias] [stack-name] # Display events for a stack
stack_master outputs [region-or-alias] [stack-name] # Display outputs for a stack
stack_master resources [region-or-alias] [stack-name] # Display outputs for a stack
stack_master status # Displays the status of each stack
stack_master tidy # Find missing or extra templates or parameter files
stack_master compile # Print the compiled version of a given stack
stack_master validate # Validate a template
stack_master lint # Check the stack definition locally using cfn-lint
stack_master nag # Check the stack template with cfn_nag

Applying updates - stack_master apply

The apply command does the following:

• Compiles the proposed stack template and resolves parameters.
• Fetches the current state of the stack from CloudFormation.
• Displays a diff of the current stack and the proposed stack.
• Creates a change set and displays the actions that CloudFormation will take to perform the update (if the stack already exists).
• Asks if the update should continue.
• If yes, the API calls are made to update or create the stack.
• Stack events are displayed until CloudFormation has finished applying the changes.

Demo:

Drift Detection - stack_master drift

stack_master drift us-east-1 mystack uses the CloudFormation APIs to trigger drift detection and display resources that have changed outside of the CloudFormation stack. This can happen if a resource has been updated via the console or CLI directly rather than via a stack update.

Diff - stack_master diff

stack_master diff us-east-1 mystack displays whether the computed parameters or template differ to what was last applied in CloudFormation. This can happen if the template or computed parameters have changed in code and the change hasn't been applied to this stack.

Maintainers

• Steve Hodgkiss
• Glen Stampoultzis

License

StackMaster uses the MIT license. See LICENSE.txt for details.

Contributing

1. Fork it ( http://github.com/envato/stack_master/fork )
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create new Pull Request