A collection of Terraform-based blueprints
Bedrock - Building blocks for composable Cloud architectures
Table of Contents
- Introduction - What is Bedrock?
- Getting started - How to use Bedrock
- Development - Guide for contributing to the Bedrock project
Bedrock is a collection of managed role-based policies and Terraform-based blueprints to assist with provisioning infrastructure and services. When we design modern computing architectures it is important to consider security, reliability and efficiency as equally important concerns. With public Cloud architectures in particular, security must be addressed throughout the entire architecture, and not just at the perimiter.
Role-based access control (RBAC) allows us to restrict actors to the minimum required permissions, which is commonly referred to as the principle of least privilege, and is the basis for the architectural blueprints provided by Bedrock.
The Bedrock blueprints are based on popular tools such as Terraform and Cloudformation, and provide both an informative and practical approach to infrastructure provisioning. You are encouraged to explore and critique these blueprints as they should continue to evolve over time.
The purpose of Bedrock is not only to provide best-practice blueprints for modern architectures, but to explore and educate about the challenges, decisions and opinions behind the designs themselves. As such, Bedrock aims to avoid a "black box" approach and is designed to encourage hacking and examining the underlying code.
Blueprints provide an opinionated approach to provisioning an infrastructure component based on common and/or best- practice approaches.
Blueprints may be either platform-based (i.e. AWS, Azure, Openstack, etc.) or service-based (i.e NGINX, Bastion, Squid Proxy, etc.). Typically they are identified using one of the following naming conventions:
<platform>-<service>[-<component>] examples: aws-ec2-autoscaling, aws-ecs-cluster, azure-cdn, openstack-network, etc. or; <service>[-<component>][-<platform>] examples: nginx-reverseproxy-aws, bastion-do, rancher-stack, squidproxy-aws, etc.
A manifest provides a way to define one or more collections of blueprints (a constellation) and associated configurations that make up a complete architecture.
Often a single blueprint is not sufficient to define a Cloud architecture as they are more likely to be distributed across multiple services. For example, you might have an ECS cluster for web/API applications, RDS or DynamDB for persistence, and S3 for archiving.
Within each of these tiers are additional ancillary services such as route53 for well-known endpoints and/or service discovery, cloudwatch events/triggers, etc. The collection of blueprints that encapsulates and independent function is called a constellation. Multiple constellations may be defined in a single manifest such that an entire architecture may be provisioned.
A manifest also provides a higher-order language that can be used to unambiguously describe novel Cloud architectures that are composed of well-defined blueprints.
Each Bedrock module is designed to exist independently from the rest, to allow maximum portability and sharing of code.
You can go into any module directory and run
terraform init && terraform apply to execute a blueprint in your
For example, if you require a Bastion host in your AWS account, do the following:
# provision IAM roles for creating Bastion host cd terraform/blueprints/bastion/roles terraform init && terraform apply # provision a new Bastion EC2 host cd ../aws terraform init && terraform apply -var bastion_user=bedrock
You can also build any Bedrock module as a Docker image, that can be run anywhere Docker is supported. Building images is supported via a Makefile in the root directory, which you can execute as follows:
# Build Docker images for Bastion roles and instance make bastion-roles bastion-aws
The Makefile will ensure dependencies are build in the right order, and includes support for tagging and push to a remote Docker registry:
# Tag and push to AWS ECR REGISTRY=<aws_account_id>.dkr.ecr.ap-southeast-2.amazonaws.com make tag push bastion-aws
After building an image you can now use the provided scripts to execute the blueprint in the current working directory:
# provision a new Bastion EC2 host export BEDROCK_REGISTRY=<aws_account_id>.dkr.ecr.ap-southeast-2.amazonaws.com bastion/scripts/bastion-aws.sh init && bastion/scripts/bastion-aws.sh apply
As the Docker images for Bedrock blueprints can be run anywhere that supports Docker, it is now possible to integrate with automated deployment and provisioning tools also. This might include build tools such as Jenkins or Bamboo, and also integrated with Cloud platforms such as AWS via the AWS Developer Tools (CodeBuild, CodePipeline, etc.).
As an example, we might configure a CodeBuild project to provision a blueprint using configuration from an S3 Bucket. Using S3 Bucket notifications we can trigger a build by simply updating a blueprint configuration. This allows for a very minimal effort approach to provisioning sophisticated and secure architectures whilst retaining the ability to maintain and evolve the designs.
Comparison with other tools
Terragrunt provides a wrapper to Terraform that enforces consistency and reduces code duplication across multiple modules. Whilst Bedrock offers similar module grouping via the manifest file, it does not impose constraints on the code in each module, nor does it hide the underlying Terraform code.
Astro offers dependency mapping between Terraform modules and concurrent execution. Bedrock provides rudimentary dependency mapping (via sequential execution of the manifest), it doesn't yet support concurrent execution.
The prerequisites for running the examples below are as follows:
Ensure you have set the following environment variables:
AWS_ACCESS_KEY_ID = <your AWS access key> AWS_SECRET_ACCESS_KEY = <your AWS secret key> AWS_DEFAULT_REGION = <your AWS region>
Build Docker images
$ cd blueprints && make all
The following manifests are required to configure appropriate IAM roles/permissions
$ bin/bedrock.py -m manifests/iam.yml init && bin/bedrock.py -m manifests/iam.yml apply
Bedrock Manifest Tool
A collection of helper tools for Bedrock are available in the Python Package Index (pypi.org), and can be installed as follows:
$ pip install brock
The first manifest you run should initialise your environment, for example:
## Contents of bedrock-init.yml name: AWS Initialisation description: Create Bedrock IAM roles and an S3 bucket for storing Terraform state constellations: policies: iam-policies: roles: iam-roles: default: terraform-state-aws:
You can then run this using the Bedrock Manifest Tool:
$ bmt init -m bedrock-init.yml $ TF_APPLY_ARGS=-auto-approve bmt apply -m bedrock-init.yml
The first command will initialise the Terraform providers/plugins, whilst the second will provision the IAM roles and S3 bucket required for Bedrock provisioning.
NOTE: As of this writing the above manifest will create 15 IAM policies, 2 IAM roles and 1 S3 bucket.
If you ever need to switch accounts where Terraform state is stored you can do this by reconfiguring the Terraform backend as follows:
TF_INIT_ARGS=-reconfigure bmt init -m bedrock-init.yml
Provision a Bastion host
The following manifest file describes how to deploy an EC2 instance as a Bastion host:
name: AWS Bastion Host description: Provision an EC2 instance for Bastion constellations: bastion: bastion-roles: bastion-aws: bastion_user: fortuna
You can execute this file as follows:
$ bmt apply -m bastion.yml -c ssh_key=@~/.ssh/id_rsa.pub
This command will apply the manifest in your AWS account, overriding the
input variable using the public SSH key from your local SSH configuration.
Deploy Lambda Layers
The following manifest will create common Lambda Layers for use in Lambda functions:
name: Lambda Layers description: Create Lambda Layers to support Bedrock Lambda blueprints constellations: aws-java-sdk-lambda: lambda-layer: layer_name: aws-java-sdk-lambda description: Support for the AWS Lambda SDK for Java content_path: /tools/aws-java-sdk-lambda/build/layer runtimes: - java8 groovy-runtime: lambda-layer: layer_name: groovy-runtime description: Support for the Groovy JVM language content_path: /tools/groovy-runtime/build/layer runtimes: - java8 python-requests: lambda-layer: layer_name: python-requests description: Python requests package plus dependencies content_path: /tools/python-requests/packages runtimes: - python3.6
Note that this manifest requires an additional volume to provide input for the
$ bmt apply -m lambda-layers.yml -v "$PWD/tools:/tools"
To make use of Bedrock you must have access to a public Cloud and a local environment with Docker installed.
If you intend to build the Docker images from source you will also require make to be installed.
Configuration will depend on the Cloud environment(s) available, but typically will involve setting an API key in your environment that allows Terraform access to resources in your account.
When using API keys we want to restrict access to the bare minimum required to perform the required tasks (Principle of least privilege). As such you should ensure the associated user has just the permissions outlined in the table below:
|Digital Ocean||API||Full access|
You can manage your Terraform state using either local storage or with an AWS S3 bucket. It is advisable to use S3 to protect your state, in which case you will require a user with the following IAM permissions:
|S3 Bucket create||Creates an S3 bucket containing Terraform state||arn:aws:iam::<AWS account ID>:policy/bedrock-terraform-state|
|IAM Policy create||Creates an IAM policy to allow access to read/write to the S3 Bucket||arn:aws:iam::<AWS account ID>:policy/bedrock-terraform-state|
For provisioning blueprints in AWS you will require a user with the following IAM permissions:
|Terraform state (*)||Read/write permissions to S3 bucket containing Terraform state||arn:aws:iam::<AWS account ID>:policy/bedrock-terraform-state|
|Assume role||Can assume role required for blueprint provisioning||arn:aws:iam::<AWS account ID>:role/-bedrock-|
- Note that the
Terraform statepermission is only required when state is stored in AWS.
Building Docker images
If you want to build the blueprint Docker images locally a
Makefile is provided that supports the following
# Build all blueprints with the default settings make all # Build a specific set of blueprints make <blueprint> [<blueprint>...] # Build blueprints with a specific Terraform version BUILD_ARGS="--build-arg TERRAFORM_VERSION=0.11.4" make all # Generate documentation (README) for all blueprints make docs # Tag and push blueprints to a custom registry REGISTRY="https://my.docker.registry/bedrock" make tag push
Open source software is made stronger by the community that supports it. Through participation you not only contribute to the quality of the software, but also gain a deeper insight into the inner workings.
Contributions may be in the form of feature enhancements, bug fixes, test cases, documentation and forum participation. If you have a question, just ask. If you have an answer, write it down.
And if you are somehow constrained from participation, through corporate policy or otherwise, consider financial support. After all, if you are profiting from open source it's only fair to give something back to the community that make it all possible.
Use the following tools to provision a pre-configured developer environment.
$ vagrant up $ ssh -p 2222 -L 8080:localhost:8080 vagrant@localhost # password: vagrant
$ docker build -t bedrock-env . $ ./developer-env.sh
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size bedrockcli-1.0.5.tar.gz (16.2 kB)||File type Source||Python version None||Upload date||Hashes View|