stackup

Stackup provides a CLI and a simplified Ruby API for dealing with AWS CloudFormation stacks.

• Why?
• Installation
• Command-line usage
  • Stack create/update
  • Specifying parameters
  • YAML support
  • AWS credentials
  • Using URLs as inputs
  • Stack deletion
  • Stack inspection
  • Change-set support
• Programmatic usage
• Rake integration
• Docker image
• Releasing

Stackup provides some advantages over using awscli or aws-sdk directly:

• It treats stack changes as synchronous, streaming stack events until the stack reaches a stable state.

• A Stack#up facade for create/update frees you from having to know whether your stack already exists or not.

• Changes are (mostly) idempotent: "no-op" operations - e.g. deleting a stack that doesn't exist, or updating without a template change - are handled gracefully (i.e. without error).

$ gem install stackup

The entry-point is the "stackup" command.

Most commands operate in the context of a named stack:

$ stackup STACK-NAME ...

Called with --list, it will list stacks:

$ stackup --list
foo-bar-test
zzz-production

Use sub-command "up" to create or update a stack, as appropriate:

$ stackup myapp-test up -t template.json

This will:

• update (or create) the named CloudFormation stack, using the specified template
• monitor events until the stack update is complete

Requests will retry 3 times by default. After this limit is exceeded, ERROR: Rate exceeded failures will be logged. You can increase the limit using the --retry-limit option, or by setting the $AWS_API_RETRY_LIMIT environment variable.

For more details on usage, see

$ stackup STACK up --help

Stack parameters can be read from a file, e.g.

$ stackup myapp-test up -t template.json -p parameters.json

These files can be either JSON or YAML format, see YAML support for more information.

Parameters can be specified as simple key-value pairs:

{
  "IndexDoc": "index.html"
}

but also supports the extended JSON format used by the AWS CLI:

[
  {
    "ParameterKey": "IndexDoc",
    "ParameterValue": "index.html",
    "UsePreviousValue": false
  }
]

You may specify -p multiple times; stackup will read and merge all the files:

$ stackup myapp-test up -t template.json \
  -p defaults.json \
  -p overrides.json

Or, you can specify one or more override parameters on the command-line, using -o with -p:

$ stackup myapp-test up -t template.json \
  -p defaults.json \
  -o IndexDoc=index-override.html
  -o ContentDoc=content-override.html

Stack tags can be read from a file, e.g.

$ stackup myapp-test up -t template.json --tags tags.json

These files can be either JSON or YAML format, see YAML support for more information.

Tags are specified as simple key-value pairs:

{
  "environment": "dev"
}

CloudFormation requires that some stacks explicitly acknowledge certain capabilities before creation. This helps to prevent the creation of stacks with unintended privileges.

If your stack includes IAM resources, you must specify either the CAPABILITY_IAM capability, or the CAPABILITY_NAMED_IAM capability if they have custom names.

If your stack template contains macros or nested stacks, you must specify the CAPABILITY_AUTO_EXPAND capability.

Capabilities can be provided via the --capability CLI option.

$ stackup myapp-test up -t template.json \
  --capability CAPABILITY_NAMED_IAM \
  --capability CAPABILITY_AUTO_EXPAND

stackup includes defaults to including CAPABILITY_NAMED_IAM capability if, and only if, no capabilities are specified. This is to provide backwards compatibility with previously deployed stacks and may be removed in a future release.

stackup supports input files (template, parameters, tags) in YAML format, as well as JSON.

It also supports the abbreviated YAML syntax for Cloudformation functions, though unlike the AWS CLI, Stackup (by default) normalises YAML input to JSON before invoking CloudFormation APIs.

If you don't want normalisation of the YAML input to JSON, then use the --preserve-template-formatting flag to the up or change-set create commands.

Note: normalisation of S3 / HTTP URL stored templates is never done, as Cloudformation collects these directly.

The stackup command-line looks for AWS credentials in the standard environment variables.

You can also use the --with-role option to temporarily assume a different IAM role, for stack operations:

$ stackup myapp-test up -t template.json \
  --with-role arn:aws:iam::862905684840:role/deployment

You can use the --service-role-arn option to pass a specific IAM service role for CloudFormation to use for stack operations:

$ stackup myapp-test up -t template.json \
    --service-role-arn arn:aws:iam::862905684840:role/cloudformation-role

(for more information on CloudFormation service roles, see AWS' documentation).

You can use either local files, or HTTP URLs, to specify inputs; stack template, parameters, etc.

$ stackup mystack up \
  -t https://s3.amazonaws.com/mybucket/stack-template.json

Where a template URL references an object in S3, stackup leverages CloudFormation's native support for such URLs, enabling use of much larger templates.

Non-S3 URLs are also supported, though in that case stackup must fetch the content itself:

$ stackup mystack up \
  -t https://raw.githubusercontent.com/realestate-com-au/stackup/main/examples/template.yml

Sub-command "delete" deletes the stack.

Inspect details of a stack with:

$ stackup myapp-test status
$ stackup myapp-test resources
$ stackup myapp-test outputs

You can also create, list, inspect, apply and delete change sets using stackup.

$ stackup myapp-test change-sets
$ stackup myapp-test change-set create -t template.json
$ stackup myapp-test change-set inspect
$ stackup myapp-test change-set apply

The change-set name defaults to "pending", but can be overridden using --name.

The change-set create subcommand, like the up command, supports --service-role-arn to specify a service role.

It is impossible to create a change set with no changes. By default, stackup will only return successfully if a change set was actually created, and will otherwise fail. If the --no-fail-on-empty-change-set option is provided, stackup will return successfully if a change set was created or if no change set was created because no changes were needed.

Get a handle to a Stack object as follows:

stack = Stackup.stack("my-stack")

You can pass an Aws::CloudFormation::Client, or client config, to Stackup, e.g.

stack = Stackup(credentials).stack("my-stack")

See {Stackup::Stack} for more details.

Stackup integrates with Rake to generate handy tasks for managing a stack, e.g.

require "stackup/rake_tasks"

Stackup::RakeTasks.new("app") do |t|
  t.stack = "my-app"
  t.template = "app-template.json"
end

providing tasks:

rake app:diff       # Show pending changes to my-app stack
rake app:down       # Delete my-app stack
rake app:inspect    # Show my-app stack outputs and resources
rake app:up         # Update my-app stack

Parameters and tags may be specified via files, or as a Hash, e.g.

Stackup::RakeTasks.new("app") do |t|
  t.stack = "my-app"
  t.template = "app-template.json"
  t.parameters = "production-params.json"
  t.tags = { "environment" => "production" }
end

Stackup is also published as a Docker image. Basic usage is:

docker run --rm \
    -v "`pwd`:/cwd" \
    -e AWS_ACCESS_KEY_ID -e AWS_SECRET_ACCESS_KEY -e AWS_SESSION_TOKEN \
    -e AWS_DEFAULT_REGION \
    realestate/stackup:latest ...

If you're sensible, you'll replace "latest", with a specific version.

The default working-directory within the container is /cwd; hence the volume mount to make files available from the host system.

This policy grants the principal all actions required by stackup up for any cloudformation stack:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "cloudformation:CreateStack",
                "cloudformation:DeleteStack",
                "cloudformation:DescribeStackEvents",
                "cloudformation:DescribeStackResource",
                "cloudformation:DescribeStacks",
                "cloudformation:SetStackPolicy",
                "cloudformation:UpdateStack"
            ],
            "Resource": [
                "*"
            ]
        }
    ]
}

auto/test will run the tests in a Docker container. auto/lint will run the linter in a Docker container.

Releasing is done mostly by CI. The automated process will push tags to GitHub, push the docker images to DockerHub. It won't yet push the gem to rubygems.org, but we're working ok it..

Prerequisites:

• You must have a rubygems account with permission to push to the stackup gem. (auto/release-gem will ask for your username and password)
• You have bumped the version number in lib/stackup/version.rb.
• You have checked the CHANGES.md file to make sure it is reasonable and has the changes for this new version.

To release the rubygem.

1. auto/release-gem
2. On REA's internal CI tool of choice, find the build for stackup-ci and trigger a new build from HEAD. This will always grab the latest code from here and release the rest of the parts.