Skip to content

schollii/terraform-local-gen-files

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Overview

This module makes it easy to extract terraform state at the time of terraform apply and automatically update config files used by other systems or tools separate from terraform such as helm or kubectl, but which depend on some infrastructure state or settings such as database addresses, security group ids, etc.

This module has been used in several closed source projects to generate helm values files and could surely be used to generate kustomize files, ansible files, etc. This module is especially useful to those not using the kubernetes or helm providers.

image

image

The main concepts in this module "stacks", "namespaces" and "config" files:

  • a stack is an instance of a set of resources that describe a complete environment. Eg a VPC, subnets etc, plus an EKS cluster, plus IAM, security groups, databases, lambdas, all for one environment, is a stack. There could be a dev stack, a staging stack, a prod stack. In the dev stack there could be multiple instances of an application, each for a different developer. The precise meaning will depend on your situation, but the term is used to refer to the totality of a set of resources that logically belong together.
  • a namespace is a subdivision of this stack into blocks. Think kubernetes, but this is just the term used for subdivision.
  • a config is a text file or set of text files that contain settings that are required to do something external to terraform with the resources of the stack/namespace. This could be the values files for a particular deployment of an "app" in a namespace of the stack (vpc + cluster).

Usage

Using this is very simple, and best understood by looking in the examples folder. Simplest example: you have YOUR_TF_MODULE/main.tf, and you need to get the AWS RDS DB address into the helm chart values file of YOUR_APP_DIR which represents a container deployed in your kubernetes cluster:

  • In YOUR_APP_DIR/config/_templates_/base-auto-root-values.yaml put the following:

    user_url: http://${cluster_name}/user/id
    security:
      group1: ${sg_ingress}
    region: ${aws_region}
    
  • in YOUR_TF_MODULE/main.tf, put this:

    module "gen_helm_values" {
      source = "schollii/gen-files/local"
    
      stack_id     = "YOUR_CLUSTER_NAME"
      config_roots = ["YOUR_APP_DIR/config"]
      tpl_name     = "values.yaml"
    
      # the keys and values in template_vars are completely arbitrary:
      template_vars = {
        cluster_name  = "YOUR_CLUSTER_NAME"
        sg_ingress    = "SOME_SG_ID"
        aws_region    = "us-east-1"
      }
    }
    
  • In YOUR_TF_MODULE, run terraform init then terraform apply: this will load the YOUR_APP_DIR/config/_templates_/base-auto-root-values.yaml file as a terraform template, replace all occurrences of the keys of tmpl_vars by the corresponding values, and write the result to YOUR_APP_DIR/config/stacks/YOUR_CLUSTER_NAME/base-auto-root-values.yaml.

This scales nicely:

  • you can have multiple stacks/clusters: each app will have the correct values for all the different stacks they need values from, in clearly recognizable folders.
  • you can "specialize" the templates for individual namespaces (check out namespace variable), and for individual clusters
  • you can have a global config file at the root of the stacks folder (it is therefore static)
  • you can have a large number of services, you don't need to remember which service depends on which terraform state data; once in the template, and in your terraform config files, the association is easy to identify at any time
  • independent modules of your terraform config can each be responsible for their own outputs, no need to trickle outputs all the way to the top. Eg YOUR_APP_DIR/config/_templates_/module1.yaml and $YOUR_APP_DIR/config/_templates_/module2.yaml
  • when terraform apply changes something in the infrastructure that impacts services, it is clear which services are impacted. You can then run helm install/upgrade or kubectl apply on only those services.

The main challenge is to decide on the order in which to combine all the generated config files. However for helm and kustomize this is typically fairly obvious: from most to least specific, ie global/base, auto-root, auto-cluster, auto-namespace; and at each level, the generated configs take precedence over any static ones. Since the files are known for each service, a helmfile can be useful to capture the list of config files and their order.

Contributions

Contributions and requests are welcome. It could be made less specific to kubernetes if there is interest.