e2e-provisioning

This is an example of how terraform can be used for end-to-end provisioning, not only of "logical" resources but also for full integration with other parts of the process, e.g.: Infrastructure, Service Discovery, Monitoring, Paging...

It is important to understand that this type of provisioning is not focus on delivering "applications" yet peripherical and sorrounding infrastructured used by applications, e.g.: databases, caches, queues, loadbalacers, clusters rather than "containers". Containers or any sort of app deployment mechanism (e.g.: binary deployment, code-replacement, lambda functions) should live outside terraform as terraform is not a deployment tool.

To illustrate the example an example company is using:

• AWS Resources, e.g.: "RDS", "ElastiCache" - All terraform resources for AWS are using a mocked resource (null_resource) to facilitate this playground environment
• Consul for Service Discovery - All AWS-created resources are registered on consul for service discovery from the application side
• ...

Distributed ownership, centralized controls

End-to-end Provisioning is extremely coupled with company culture, not all provisioning methods work with all company structures. This specific approach tends to work better with a YBIYRI strategy. It is important to highligh which roles/teams are expected to have be around:

• platform-providers - Engineers working on ensuring modules and mechanisms around
  • Platform engineers are expected own the processes, documentation and communicatin of infra-offerings rather than owning the infra itself. They create, maintain and document offerings for things such as: Databases, Cache, Serverless Components, LoadBalancers, CDNs...
  • Platform engineers review Pull-requests from service-owners and ensure quality controls continue to be in-place for all new infra created, either by simply code-review or building more robust controlling mechanisms to automate aways common problems
  • example roles: Infra Engineers, Database Engineers, Network Engineers, SREs, ...
  • example interactions:
    • "database engineer creates a new engine module for provisioning Amazon Aurora"
    • "database engineer makes it possible to upgrade Postgres versions to 14.2"
    • "infra engineer fixes a bug on all MySQL backups"
    • "SRE creates a new SLO monitor for all ElastiCache instances"
    • "SRE enables default MultiAZ read-replicas for all production Postgres instances"
    • "platform engineer reviews pull-requests from product to add a new Postgres cluster"
• service-owners - Engineers working on their services/applications which require peripherical infrastructure
  • Service owners are expected to also own their infrastructure, using everything platform creates, reading internal documentation and being aware of all oferings.
  • example roles: Back-end Engineers, Front-end Engineers, SREs, ...
  • example interactions:
    • "finance engineer creates a new Database"
    • "front-end engineer updates Loadbalancer Headers"
    • "customer acquisition engineer creates a pull-request to increase cache memory"
    • "support engineer creates a pull-request to change backup configuration of a database"

Makefile

apply             Run terraform apply
compose           Start all dependencies via docker
format            Format the code
help              List help commands
init              Lock all terraform projects dependencies
install_terraform Install terraform version with tfenv
plan              Run terraform plan
setup             Install all dependencies

File structure

├── Makefile  - Overall commands
├── README.md - Documentation
├── docker-compose.yml - Dependencies docker
├── accounts
│   ├── production -> Production resources
│   │   ├── main.tf
│   │   ├── providers.tf
│   │   └── versions.tf
│   └── staging -> Staging resources
│       ├── main.tf
│       ├── providers.tf
│       └── versions.tf
└── modules
    ├── aws -> AWS related modules
    │   ├── elasticache -> A module to create ElastiCache Clusters
    │   └── rds -> A module to create RDS Clusters
    ├── consul -> Consul related modules
    │   ├── node -> A module to create consul nodes
    │   └── service -> A module to create consul services
    └── platform -> "Platform" modules
        ├── cache -> A module to create a full end-to-end cache
        └── database -> A module to create a full end-to-end database

• accounts - There are individual "accounts", those can be any isolation level you want to set but usually represent "orgs" and/or "environments"
  • e.g.: "platform-production", "platform-staging", "payments-production",...
  • For simplicity here only production/staging is used
  • Each folder is its own terraform project for state isolation purposes
• modules -> contain both generic modules for external providers as much as "platform" modules, which are mostly an end-to-end combination of the otheres to setup everything related to that resource