Skip to content

automation tooling playground

Notifications You must be signed in to change notification settings

ckdake/automation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

53 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Automation

tflint trivy infracost test1-sample-app-web-ci

Terraform for AWS plaground

The goal of this AWS setup is to have an automated AWS Organization set up that:

  1. Is fully compliant with SOC2
  2. Is fully compliant with AWS, CIS, and NIST security standards.
  3. Has minimal AWS cost overhead
  4. Facilitates easily testing out things in AWS test accounts

Things to have readily available to be able to test other things:

  1. a docker container in ECR running as an ECS service
  2. lambdas triggered by EventBridge

Running locally

This requires env vars with user credentials that can assume to adminstrator. If aws sts get-caller-identity works, you are good, otherwise:

export AWS_ACCESS_KEY_ID=
export AWS_SECRET_ACCESS_KEY=

Currently only works in tenants/management with terraform apply.

Run prowler to populate Security Hub with any breaking things it fines by: cd tools/prowler/ && ./install-prowler.sh && ./run-prowler.sh

Run trivy to scan for vulns: cd tenants/management/ && trivy config . --ignorefile ./.trivyignore.yaml

Bootstrapping a new account

  1. Add the account in tenants/management/organization.tf and terraform apply
  2. Log into the account's root user with "forgot password" and get some keys, set up profile
  3. Create a tenants/${TENANTNAME} folder, copy a main.tf from another tenant, and update the account id in the provider block and the backend key block. (TODO(ckdake): move this to a terragrunt template)
  4. cd into tenants/${TENANTNAME} and run ../../scripts/bootstrap-tenant.sh

Using the app-workspace tenant

This is a unique tenant, as a proof-of-concept for using Terraform workspaces and persisting some state. This could be used to manage a "review-app" pattern.

Each workspace stores some output in it's state, in a way that is usable in future terraform apply runs. The use case is for situations where different jobs in different places need to make changes to the same terraform-managed application. One example is separate pipelines for "frontend" and "backend" that both use a terraform apply to update an ECS task definition, for either the frontent or the backend. We don't want the frontend pipeline to know about the changes to the task defintion for the backend task. Any time terraform apply is run, it will only change things that are passed in via TF_VAR env vars, and all other things will persist. Defaults are set in the locals.tf.

cd tenants/app-workspace
terraform workspace select -or-create my-app-one
TF_VAR_config_value_one=test1 terraform apply
TF_VAR_config_value_two=test2 terraform apply

TODO

  • implement "automation:prod" pattern of docker container to run jobs
  • implement Application/Environment tagging, including in modules
  • implement SSO into AWS from somewhere
  • implement SSM ssh pattern
  • Sample-app
    • automatic deploy on container push to ECR
    • talk to an AWS Serverless V2
    • talk to an ElasticCache
    • Use a Secret = [ ] Cloudwatch dashboard
  • saml2aws for logging in, what to use for IdP?
  • import everything in root account, test1 account, test2 account
    • dhcp options, internet gateway, org delegation config
  • get prowler checks to 100% green
  • setup github actions for terraform plan, terraform apply, prowler
  • lots more

About

automation tooling playground

Resources

Stars

Watchers

Forks

Languages