First look at terraform

Terraform is client based like Ansible (which is good). Very cloud focussed and relies on Cloud API's to manage the infrastructure. This is good as it means you dont have to worry about authentication and ports etc but it does need to be a supported cloud vendor. You couldn't use it manage your own on-premise servers (presumably).

Example config file:

provider "aws" {
  access_key = "AKIAIFZVQTBBG3YF46LQ"
  secret_key = "SfEjiU0I2I7wxXQqbRWLPYCbEfdSxhI/4r44CgaV"
  region     = "us-east-1"
}

resource "aws_instance" "example" {
  ami           = "ami-2757f631"
  instance_type = "t2.micro"
}

> terraform init
> terraform apply
> terraform show 

Terraform is purely declarative. Ansible IS declarative but you need to be disciplined as it can be used imperatively.

One blog calls Terraform an 'orchestration' tool with ansible being a more traditional configuration management tool.

Some changes can be done in-place (~) but others, an obvious one being an image change, require a replacement (+/-).

> terraform destroy

Implicit dependencies are used to determine the order in which resources are created, eg, a ref from an elastic IP to a machine. That is, Terraform will infer that the machine needs to be created first and then the EIP. This is in fact the preferred way, for other case an explicit 'depends_on' property can be used.

PROVISIONERS are custom scripts that are run once when the resource is created, they are therefore only suited bootstrapping a server and NOT suited to config management. If CM is needed, then it should be bootstrapped here. If provisioning fails, the resource is marked as TAINTED, not rollback but will be destroyed and recreated on the next execution plan.

Variables can be used for inputs and outputs.

Comments

Post a Comment

Popular posts from this blog

https://simonsmith.io/dipping-a-toe-into-functional-js-with-lodash-fp/ Great blog on Functional programming. For me does an excellent job of highlighting the difference between pure functions and bonified functional programming. lodash fp is a wrapper around lodash to provide immutable, auto curried, iteratee first data last functions. Currying seems to make sense here, just functions that return a function. The data param is often omitted at the end to make the function more re-usable. Currying, just keeps returning a function until it has all the arguments. This is why it makes sense to have the data arg last so it can be provided last. I still can't see the functional programming being realistically easier to use but pure functions definitely make a lot of sense.
Read more

A brief look into Cloudformation

Image
In this blog, i just wanted to record details and my thoughts on following a first tutorial on Cloudformation. This is the free AWS offering for managing and provisioning infrastructure resources within a cloud environment - effectively their IaC service. A cloud formation template describes a group of resources that form a stack. Resources can references resources within the template or outside of the stack entirely. Template can be written in JSON or YAML. A stack is free in AWS but the resources within are priced as normal. Templates have several useful features: * Parameters * Mappings for conditional values (eg AMI ID's differ between regions) * Functions can be used to join values etc * Output values Ultimately i would like to compare and contrast with Terraformer. Updating a stack This tutorial describes how to update a stack implementing a sample PHP app on apache. MetaData/Init section for an EC2 instance allows you to specify things like which packages ar
Read more