terraform environment init.sh

init.sh

#!/bin/bash

# Usage: ./init.sh once to initialize remote storage for this environment.
# Subsequent tf actions in this environment don't require re-initialization, 
# unless you have completely cleared your .terraform cache.
#
# terraform plan  -var-file=./production.tfvars 
# terraform apply -var-file=./production.tfvars 

tf_env="production"

terraform remote config -backend=s3 \
                        -backend-config="bucket=hound-terraform-state" \
                        -backend-config="key=$tf_env.tfstate" \
                        -backend-config="region=us-east-1"

echo "set remote s3 state to $tf_env.tfstate"

path: $gitroot/infra/terraform/env-prod/init.sh

So, the question I have about this, is does it always sync the state to the same .terraform file? So if I'm running a dev cycle, and do env-dev/init.sh, finalize it, and then do env-prod/init.sh, will the local file be .terraform/terraform.tfstate no matter what?

Has that been a problem (following along with your wonderful and timely blog post, btw).

Thanks!

@sean-abbott I guess it will use $(PWD) to use .terraform . So if you switch dirs (cd env-dev or cd env-production) and then run init.sh you'll get:

[env-dev]
   - [.terraform]
      - .terraform.state
[env-prod]
   - [.terraform]
      - .terraform.state

We're doing much of what you've outlined here (hooray, we're doing it right too!), but my question is this: if you have a terraform remote state resource, why do you need to initialize it with the shell script? We had our own shell script already and it was my thinking that putting the remote state into Terraform as a resource would eliminate the need for it. BY which I mean, we had the shell script and NO terraform remote state resource declaration and it worked just fine.

Thoughts?

Check this out. https://github.com/pporada-gl/terraform-stuff