Skip to main content


Stacks are a way to express the complete infrastructure needed for an environment composed of Components using a standard YAML configuration


Stacks are a central SweetOps abstraction layer that is used to instantiate Components. They’re a set of YAML files that follow a standard schema to enable a fully declarative description of your various environments. This empowers you with the ability to separate your infrastructure’s environment configuration settings from the business logic behind it (provided via components).

SweetOps utilizes a custom YAML configuration format for stacks because it’s an easy-to-work-with format that is nicely portable across multiple tools. The stack YAML format is natively supported today via Atmos , the terraform-yaml-stack-config module, and Spacelift via the terraform-spacelift-cloud-infrastructure-automation module.


Stacks define a generic schema for expressing infrastructure

How-to Guides


We have a number of important conventions around stacks that are worth noting.


Make sure you’re already familiar with the core Concepts.

Stack Files

Stack files can be very numerous in large cloud environments (think many dozens to hundreds of stack files). To enable the proper organization of stack files, SweetOps recommends the following:

  • All stacks should be stored in a stacks/ folder at the root of your infrastructure repository.

  • Name individual environment stacks following the pattern of $environment-$stage.yaml

  • For example, $environment might be ue2 (for us-east-2) and $stage might be prod which would result in stacks/ue2-prod.yaml

  • For any global resources (as opposed to regional resources), such as Account Settings, IAM roles and policies, DNS zones, or similar, the environment for the stack should be gbl to connote that it’s not tied to any region.

  • For example, to deploy the iam-delegated-roles component (where all resources are global and not associated with an AWS region) to your production account, you should utilize a stacks/gbl-prod.yaml stack file.


When you have a configuration that you want to share across various stacks, use catalogs. Catalogs are the SweetOps term for shared, reusable configuration.

By convention, all shared configuration for stacks is put in the stacks/catalog/ folder, which can then be used in the root stacks/ stack files via import. These files use the same stack schema. Learn more about How to Use Imports and Catalogs in Stacks.

There are a few suggested shared catalog configurations that we recommend adopting:

  • Global Catalogs: For any configuration to share across all stacks.

  • For example, you create a stacks/catalog/globals.yaml file and utilize import wherever you need that catalog.

  • Environment Catalogs: For any configuration you want to share across environment boundaries.

  • For example, to share configuration across ue2-stage.yaml and uw2-stage.yaml stacks, you create a stacks/catalog/stage/globals.yaml file and utilize import in both the ue2-stage.yaml and uw2-stage.yaml stacks to pull in that catalog.

  • Stage Catalogs: For any configuration that you want to share across stage boundaries.

  • For example, to share configuration across ue2-dev.yaml, ue2-stage.yaml, and ue2-prod.yaml stacks, you create a stacks/catalog/ue2/globals.yaml file and import that catalog in the respective dev, stage, and prod stacks.

  • Base Components: For any configuration that you want to share across all instances of a component.

  • For example, you’re using the eks component and you want to ensure all of your EKS clusters are using the same Kubernetes version, you create a stacks/catalog/component/eks.yaml file which specifies the eks component’s vars.kubernetes_version. You can then import that base component configuration in any stack file that uses the eks component.

  • More information in the below section.

Component Inheritance

Using a component catalog, you can define the default values for all instances of a component across your stacks. But it is also important to note that you can provide default values for multiple instances of a component in a single stack using the component inheritance pattern via the component key / value:


Pro tip: You can also use our inheritance model for the polymorphism of components.

# stacks/catalog/component/s3-bucket.yaml
enabled: false
user_enabled: false
acl: private
grants: null
versioning_enabled: true

# stacks/uw2-dev.yaml
- catalog/component/s3-bucket
- catalog/dev/globals
- catalog/uw2/globals

component: s3-bucket
enabled: true
acl: public
name: public-images

component: s3-bucket
enabled: true
name: export-data

# ...

In the above example, we’re able to utilize the default settings provided via the s3-bucket base component catalog, while also creating multiple instances of the same component and providing our own overrides. This enables maximum reuse of global component configuration.

Terraform Workspace Names

In atmos and the accompanying terraform automation modules like terraform-spacelift-cloud-infrastructure-automation the terraform workspaces will be automatically created when managing components. These workspaces derive their names from the stack name and the component name in question following this pattern: $env-$stage-$component. The result is workspace names like ue2-dev-eks or uw2-prod-mq-broker.

Pro Tips

Here are some tips to help you write great stacks:

  1. Use {} for empty maps, but not just a key with an empty value.

  2. Use consistent data types for deep merging to work appropriately with imports (e.g. don’t mix maps with lists or scalars).

  3. Use YAML anchors to DRY up a config within a single file.


IMPORTANT Anchors work only within the scope of a single file boundary and not across multiple imports.

Stack Schema

The official JSON Schema document for Stacks can be found here. The below is a walk-through of a complete example utilizing all capabilities.

# stacks/ue2-dev.yaml

# `import` enables shared configuration / settings across different stacks
# The referenced files are deep merged into this stack to support granular configuration capabilities
# Merge the below `stacks/catalog/*.yaml` files into this stack to provide any shared `vars` or `components.*` configuration
- catalog/globals
- catalog/ue2/globals
- catalog/dev/globals

# `vars` provides shared configuration for all components -- both terraform + helmfile
# Used to determine the name of the workspace (e.g. the 'dev' in 'ue2-dev')
stage: dev
# Used to determine the name of the workspace (e.g. the 'ue2' in 'ue2-dev')
environment: ue2

# Define cross-cutting terraform configuration
# `terraform.vars` provides shared configuration for terraform components
vars: {}

# `backend_type` + `backend` provide configuration for the terraform backend you
# would like to use for all components. This is typically defined in `globals.yaml`
# atmos + our modules support all options that can be configured for a particular backend.
# `backend_type` defines which `backend` configuration is enabled
backend_type: s3 # s3, remote, vault
encrypt: true
bucket: "eg-uw2-root-tfstate"
key: "terraform.tfstate"
dynamodb_table: "eg-uw2-root-tfstate-lock"
role_arn: "arn:aws:iam::999999999999:role/eg-gbl-root-terraform"
acl: "bucket-owner-full-control"
region: "us-east-2"
remote: {}
vault: {}

# Define cross-cutting helmfile configuration
# `helmfile.vars` provides shared configuration for terraform components
account_number: "999999999999"

# Components are all the top-level units that make up this stack

# All terraform components should be listed under this section.

# List one or more Terraform components here
# Provide automation settings for this component
# Provide spacelift specific automation settings for this component
# (Only relevant if utilizing terraform-spacelift-cloud-infrastructure-automation)

# Controls whether or not this workspace should be created
# NOTE: If set to 'false', you cannot reference this workspace via `triggers` in another workspace!
workspace_enabled: true

# Override the version of Terraform for this workspace (defaults to the latest in Spacelift)
terraform_version: 0.13.4

# Which git branch trigger's this workspace
branch: develop

# Controls the `autodeploy` setting within this workspace (defaults to `false`)
auto_apply: true

# Add extra 'Run Triggers' to this workspace, beyond the parent workspace, which is created by default
# These triggers mean this component workspace will be automatically planned if any of these workspaces are applied.
- ue2-dev-second-component
- gbl-root-example1

# Set the Terraform input variable values for this component.
my_input_var: "Hello world! This is a value that needs to be passed to my `first-component` Terraform component."
bool_var: true
number_var: 47

# Complex types like maps and lists are supported.
- example1
- example2

key1: value1
key2: value2

# Every terraform component should be uniquely named and correspond to a folder in the `components/terraform/` directory
my_input_var: "Hello world! This is another example!"

# You can also define component inheritance in stacks to enable unique workspace names or multiple usages of the same component in one stack.
# In this example, `another-second-component` inherits from the base `second-component` component and overrides the `my_input_var` variable.
component: second-component
my_input_var: "Hello world! This is an override."

# All helmfile components should be listed under this section.

# Helmfile components should be uniquely named and correspond to a folder in the `components/helmfile/` directory
# Helmfile components also support virtual components

# Set the helmfile input variable values for this component.
installed: true
enableCertManager: true

# `workflows` enable the ability to define an ordered list of operations that `atmos` will execute. These operations can be any type of component such as terraform or helmfile.
# See "Getting started with Atmos" documentation for full details: /tutorials/atmos-getting-started/

# `workflows` is a map where the key is the name of the workflow that you're defining

# `description` should provide useful information about what this workflow does
description: Deploy helmfile charts in the specific order

# `steps` defines the ordering of the jobs that you want to accomplish

# `job` entries defined `atmos` commands that you want to execute as part of the workflow
- job: helmfile sync cert-manager
- job: helmfile sync external-dns
- job: helmfile sync alb-controller
- job: helmfile sync metrics-server
- job: helmfile sync ocean-controller
- job: helmfile sync efs-provisioner
- job: helmfile sync idp-roles
- job: helmfile sync strongdm
- job: helmfile sync reloader
- job: helmfile sync echo-server