tf-registry

Self Hosted Terraform Registry backed by S3

Usage

tf-registry Provides a simple http server that implements the Terraform Module Registry Protocol.

Terraform Registry Server

Usage: tf-registry [flags] 

Flags:
  -bucket string
    	aws s3 bucket name containing terraform modules
  -port string
    	port for HTTP server (default "3000")
  -prefix string
    	optional path prefix for modules in s3
  -profile string
    	aws named profile to assume (default "default")

Uploading Modules

tf-registry uses S3 as its backend storage, and at the moment it is read only, meaning that modules must be manually uploaded to s3 before they can be retrieved via tf-registry.

The path format must match the expected format: s3://<bucket>/[optional_prefix]/<registry_namespace>/<module_name>/<provider>/<version>/<module_name>.tgz

Example upload script (run from local module path):

#!/bin/bash

BUCKET="tf-registry-storage"
MODULE_NAME=$(basename $PWD)
REGISTY_NAMESPACE="nalbury"
PROVIDER="aws"
VERSION="1.0.0"

TMP_DIR=$(mktemp -d)
tar -czf ${TMP_DIR}/${MODULE_NAME}.tgz .
aws s3 cp ${TMP_DIR}/${MODULE_NAME}.tgz \
  s3://${BUCKET_NAME}/${REGISTRY_NAMESPACE}/{MODULE_NAME}/${PROVIDER}/${VERSION}/${MODULE_NAME}.tgz
rm -rf ${TMP_DIR}

Using Modules from the Registry

Once the module has been uploaded, and the server is running, you can then reference a module using the standard registry source format:

module "eks_cluster" {
  source  = "tf-registry.mydomain.io/nalbury/my-aws-module/aws"
  version = "~> 1.0.0"
}

NOTE Terraform will only install modules if your registry is served over HTTPS. You can use ngrok for a local server if necessary.

TODO

Aside from any TODOs mentioned in the code, tf-registry should ideally have:

• Build CI + Release Artficats + Container Image
• Helm Chart for running tf-registry
• Terraform Module for running tf-registry (hosted publicly)
• Module upload support either via a custom client (wrap s3 api), or via the HTTP API directly
• Authentication
• Provider registry support
• Additional backend storage providers (gcp, azure, local FS)