• I need to keep my old ghost blog running.
• My version of ghost is too old to be upgraded and I don't want to spend time re-develop a static site.
• I did not find any nice gatsby/hugo templates.
• I will use the smallest EC2 instance (2$/months), ECS/ASG (free), and a cloudfront (cheap) to keep the site up.
• However, it is not 100% HA, need to link manually the DNS to cloudfront if the instance get destroyed. If so a healthcheck will alert me to relaunch the terraform plan with the new instance DNS.

• Hosting: AWS cloud
• Front: Ghost in docker (version 0.11.3, newer makes my blog fail)
• Container orchestrator: ECS (1 EC2 instance t2.nano)
• Data: EFS + daily backup
• CDN: Cloudfront + Cert
• Code: Github
• Infra: Terraform cloud
• Price: < 3$ / month

User --> Route53 --> Cloudfront --> EC2 DNS (docker ghost on ECS) <-- EFS (blog data)

• Purchase domain d3vblog.com in route53
• Create custom certificate in zone us-east-1 (for cloudfront to see)

• Deploy with:

cd terraform/dev
nano main.tf    <-- edit variables
terraform init  
terraform apply

• If using terraform cloud: change workspace -> setting to "local" (to use your local aws credential)
• First deploy should not work because the blog data are not in EFS.

• Delete the ECS service ghost
• Connect to the EC2 instances, mount the EFS volume (check its DNS name in the console):

sudo mkdir /mnt/efs
sudo mount -t efs fs-9921c253.efs.eu-west-1.amazonaws.com /mnt/efs/

• And copy the data from a backup to /mnt/efs/. ex: scp -r -i ~/.ssh/aws-greg-eu-west-1.pem ~/Documents/blog/blog-data ec2-user@54.194.140.51:~/

• There are no LB because of cost (20$ per month). So if instance goes down, I have to update cloudfront origin with the EC2 instance public DNS.
• Get the DNS of the ECS instance and replace the variable instance_dns in terraform/dev/main.tf
• Run again a terraform apply

• Try the website on instance and CloudFront url
• Try with your DNS: https://d3vblog.com

• Use backup to backup every day your EFS data.

• Terraform will not let you create a subcription to SNS, so you are alerted when the blog goes down.
• You have to go via the console to SNS, region us-east-1(!), topics, create an EMAIL subscription, and validated the url in your mailbox.

• Go the EC2/reserved instance and buy a linux t2.nano in the region your deployed the ECS cluster. Insted of paying 5$/month for a t2.nano your should be able to find some 2.2$ / month
• Domain purchase should be around 10$ / month

Maintenance cost of this infra / month:

• 1 EC2 instance: 2.2$ for converible reserved (5$ for on demand)
• 1 heathcheck: 0.5$ ($0.50* per health check / month)
• Efs disk (< 50MB): < 0.02$ ($0.33 /GB/month)
• Efs backup: ~0$ ($0.06 per GB-Month)
• ECS: free
• Total ~ 3$ / month

• To run manually the container: docker run -d -p 2368:2368 -e NODE_ENV=production --name ghost -v $PWD/ghost-backup:/var/lib/ghost ghost:0.11.3
• Check that blog data are present, and the rights are set
• Manual backup to laptop

tar -czvf ghost.1.06.2020.tar.gz /mnt/efs/   # from server
scp -i ~/.ssh/greg-eu-west-1.pem ec2-user@ip:~/ghost.1.06.2020.tar.gz ~/GBC/BACKUP/.  # From laptop

• Ami: amzn2-ami-ecs-hvm-2.0.20200430-x86_64-ebs -> New ami-07a1802c113adc855

• Healtcheck alarm issue