Ghost Blog Project

What is it?

This repository provides all the necessary tooling to deploy an instance of the Ghost CMS on Amazon Web Services.

What does it do?

This repository will use Hashicorp's Terraform tool to deploy all the architecture required to run an enterprise ready CMS. The Terraform code will create the following:

VPC, Subnets, Internet Gateway, NAT Gateways
Route 53 records, TLS Certificate, Application Load Balancers, Security Groups
An Elastic Container Service, Task Definition to run the community maintained Docker image of of the Ghost CMS

For more information on the infrastructure configuration, see the diagram below.

How?

The entire stack, including public DNS records can be deployed directly from your computer.

Prerequisites

The Terraform cli installed locally. This has been tested with v0.13.5
An AWS account with an AWS IAM user with Admin level privileges with an access and secret key configured
A DNS domain registered with Route53

Config

Clone the repo locally git clone git@github.com:peterbamsey/ghost-blog.git
Open the terraform.sh.default file and enter your AWS Access and Secret keys as well as the target AWS region

export AWS_ACCESS_KEY_ID="your_access_key_here"
export AWS_SECRET_ACCESS_KEY="your_secret_access_key_here"
export AWS_DEFAULT_REGION="target_aws_region"

Save the file as terraform.sh
Open prod.tfvars.default and enter your Route53 managed domain name and target AWS region. The sub-domain and environment parameters can stay as default

region      = "eu-west-1"
domain-name = "yourdomain.com"
sub-domain  = "blog"
environment = "prod"

Save the file as prod.tfvars
Execute the terraform.sh script with ./terraform "prod" from the command line.
You will be prompted to confirm the changes that Terraform will make. Type yes at the prompt.
Wait a for the Terraform run to complete. After local DNS servers have updated your new blog site should be at https://blog.yourdomain.com
To destroy the infrastructure and blog site so you are no longer charged, uncomment the last line in the terraform.sh file and re-run the script

# Uncomment the line below line to destroy the stack
#terraform destroy -var-file="${ENVIRONMENT}.tfvars"

Scaling the instances

This project is configured to allow horizontal scaling of both the application and database tiers. To scale the application tier change the default value of the number-of-applicaiton-instances variable.

variable "number-of-application-instances" {
  default = 1
  description = "The number of ECS Fargate tasks to run concurrnetly. Used for horizontal scaling of the application tier"
  type = number
}

To scale the database tier change the default value of the ``


variable "number-of-db-instances" {
  default = 2
  description = "The number of RDS read replica instances to run concurrently.  Use for horizontal sclaing of the db tier"
  type = number
}

To improve this configuration we would create an autosclaing configuration for both the ECS Fargate task and the RDS Aurora service which automatically scales the instances based on a predefined metric.

Monitoring

This project creates a very basic example monitoring dashboard in AWS Cloudwatch. It uses metric widgets to display basic information about the application. The Dashboard name is prod-ghost.

We can improve this considerably by templating the metric widgets and dashbaords to make the module more reusable.

Alerting

This project creates a very basic alerting system by utilising Cloudwatch Alarms and an SNS topic. When the ECS instances go above a certain CPU utilisation, an alarm triggers that pushes a message to the SNS topic and sends an SMS message to the number defined in:

variable "sms-number" {
  description = "The number to send alarms to"
  type = string
  default = "+44012345678910"
}

This module is currently unfinished, though the foundations are present.

peterbamsey/ghost-blog