Configuration

What does terraboot do

It generates terraform-readable json. Make sure you have latest terraform installed, they do tend to fix bugs from one version to the other

Modules

terraboot is divided in 4 modules at the moment:

• vpc (generated by the vpc-vpn-infra function): for vpc, subnets, NAT, vpn, ELK, monitoring and alerting boxes (the setup of the boxes is not automated at this point, see install-dns, install-icinga, install-influx, install-logstash)
• vpc dns (generated by the vpc-public-dns function): for a number of public DNS names on the VPC (requires having a public domain on Route53).
• cluster (geenrated by cluster-infra function): for individual clusters - the idea being that you can have several clusters per vpn.
• cluster dns (generated by cluster-publlic-dns function): for cluster-specific DNS (also requires having a public domain on Route53)

The configuration can reside in both separate files and the code used to call terraboot:

• edn files per module for fixed details per part of the infrastructure
• variable parameters (like instance types or open ports) in the calls to terraboot. This is a suggestion - your mileage may vary and you may wish to integrate more parameters in the configuration file as suited.

edn files

This should contain details which are mostly fixed over the whole infrastructure. One of these file is passed in as an argument to lein run, for instance lein run resources/terraboot-staging.edn.

{:region "your-aws-region"
 :bucket-name "your-s3-bucket"
 :aws-profile "your-aws-profile"
 :account-number "your-aws-account-number" ;; to generate ARN
 :azs [:your-regions]
 :target "your-target"}

The regions should be in single-letter keyword format (say [:a :b :c]). The target should be the name of the part of the infrastructure you want to build using this configuration file (see below, for example code where the configuration is being used).

Variable configuration

The more variable configuration goes in the clojure code. By more variable I mean IP address ranges, names, open ports, instance types, disk sizes. As this is likely to be a moving target, it's best to check the signature of the various functions to work out which parameters should be used. The scenario we're going for now is a 'main' function calling the module functions with both a parameter EDN file and parameters

(ns my-setup.infra
  (:require [terraboot.core :refer :all]
            [terraboot.vpc :as vpc]
            [terraboot.public-dns :as dns]
            [terraboot.cluster :as cluster]
            [clojure.edn :as edn]
            [clojure.java.io :as io]))

(defn get-config
  "Gets info from a config file."
  [url]
  (edn/read-string (slurp url)))

(defn generate-json [edn-path]
  (let [{:keys [account-number
                region
                azs
                bucket-name
                aws-profile
                target]}) (get-config (edn-path)
        mesos-ami "" ;; desired CoreOS AMI
        default-ami "" ;; desired Ubuntu AMI
        vpc-cidr-block "" ;; VPC address range
        dns-zone "mastodonc.net" ;; if public dns is to be used
        dns-zone-id ""] ;; AWS zone id
        (condp = target
          "vpc" (do (to-file (vpc/vpc-vpn-infra {... parameters}) "vpc/vpc.tf")
                    (to-file (dns/vpc-public-dns {... parameters}) "vpc/vpc-dns.tf"))
          "dataplatform" (do (to-file (cluster/cluster-info {... parameters}) "dataplatform/dataplatform.tf")
                             (to-file (dns/cluster-public-dns {... parameters}) "dataplatform/dataplatform-dns.tf")))))

(defn -main [edn-path]
  (generate-json edn-path))

The *.tf files referred to in the code are the terraform json file terraform will consume. It's recommended to put them in their own directory, since terraform reads all tf file in a directory.

Running it all

How to run a component: generate relevant configuration file

lein run resources/terraboot-vpc.edn # takes edn path

In relevant directory (where tf files live)

terraform plan .

If planning result looks like what you'd expect (green, number of resources to plan)

terraform apply .