Skip to content

tmp_readme.md

Latest commit

 

History

History
85 lines (60 loc) · 3.07 KB

tmp_readme.md

File metadata and controls

85 lines (60 loc) · 3.07 KB

High level picture

Sources:

  • A. Cloudcraft Blueprint (json)
  • B. Lucidchart API (if exists)
  • C. AWS API (Describe AWS-API or python script similar to the one by Lucidchart can generate json)
  • D. Custom json/hcl files
  • E. modules.tf UI produces json, which is sent to MTG endpoint
  • F. Cloudformation json/yaml
  • G. draw.io (can be self-hosted, and used by financial institutions)

All gets into modules.tf's generator (MTG) which creates "infrastructure as code (terragrunt + docs)"

Mapping

Cloudcraft and modules.tf mapping of variables is defined in modules-metadata/{module}.json and updated by shell script update_modules_metadata.sh

isMultiAZ (cloudcraft) ==> multi_az (modules.tf) ==> multi_az (terraform-aws-modules)

To-do

  • Convert blueprint.json (cloudcraft) into generic json config for cookiecutter (output.json)
  • Cookiecutter should render one directory with all content as specified in json config
  • Create zip archive, upload it to S3
  • Return content of zip file back to the user to initiate download
  • Make content of archive real and usable
  • Print only required key/values into terraform.tfvars
  • Improve config generator to follow links/edges/etc
  • Get connections between (using data-sources):
    • ELB & ASG target_group_arns (Figure out how to pass dynamic value of target_group_arns (eg, ${data.terraform_remote_state.elb.arn}) to autoscaling group terraform module. Maybe using "locals", "overrides", data-sources, tfstate, etc)
    • S3 & Cloudfront
    • ELB or ALB
    • VPC
    • Security group
  • Generate real meta data from modules (or hardcode some)
  • Add basic S3 and Cloudfront modules
  • Circleci - https://serverless.com/blog/automating-ci-for-python-serverless-app-with-circleci/

@todo: # terragrunt hook which scans terraform.tfvars and replaces "data.terraform_remote_state.vpc.vpc_id" with the the real value fetched from datasource before calling the main module; using null_resource; http; external as a "glue"

Before IPEXPO (19.8.2018):

  • Generate Terraform code instead of Terragrunt

Extract data from "nodes" and convert it into single "terraform-aws-module" + available arguments (instanceType, instanceSize, elbType)

Notes

"A" get concept of "area/zone", which means it can be grouped into single "infrastructure module" before passing to modules.tf web-site and populate the forms or to modules.tf converter (MTC) before sending to MTG.

Output structure

project/
  eu-west-1/
    layer1/
      terraform.tfvars
    layer2/
      terraform.tfvars
    common.tfvars
    terraform.tfvars
  .gitignore
  LICENSE
  README.md
  ...

Run pre-commit after code was rendered:

cd final

// iterate through all files: pre-commit run -c $(pwd)/.pre-commit-config.yaml --files $(pwd)/eu-west-1/rds/terraform.tfvars pre-commit run -c $(pwd)/.pre-commit-config.yaml --files $(pwd)/....

Comments from Dmytro:

  1. if it python 3 - mb better to use pathlib https://www.scivision.co/python-idiomatic-pathlib-use/
  2. https://docs.python.org/3/library/typing.html