↓ Ir para o conteúdo principal

← todas as notas

📎 Webclip

Applying the Well-Architected Framework, Small Edition

The post starts with the goal of combining many feeds into one super feed and publishing it through a Hugo site. A Python script can write feed items as Hugo posts, but keeping it updated on a schedule leads to a cloud design choice.

The first approach uses AWS Lambda, CloudWatch Events, and Amazon EFS, but the Lambda /tmp limit and the need for a NAT gateway in a VPC make the cost unattractive. The final choice is an EC2 instance started on demand, configured with user-data and cloud-init, then used to run the update script, build the site, sync it to S3, and terminate itself.

Reading notes
#

  • The problem is to combine many feeds into one super feed and publish it on a Hugo site.
  • Hugo stores content as text files, so the solution writes feed items as unique posts in the content tree.
  • A Python script can collect feeds and write them into Hugo posts.
  • Lambda can run the script on a schedule, but the repo size exceeds the 512 MB /tmp limit.
  • Lambda needs a VPC to access Amazon EFS, and internet access from that VPC requires a NAT gateway.
  • The NAT gateway makes the Lambda plus EFS setup too expensive for a site whose hosting cost is otherwise very low.
  • EC2 in a public subnet avoids the NAT gateway and can run the workflow with user-data and cloud-init.
  • The EC2 instance installs tools, gets a GitHub token from Secrets Manager, clones the repo, runs the feed update, builds the site, syncs to S3, clears cache, and terminates itself.
  • The author weighs the solution against the five Well-Architected pillars, with cost optimization and security as the main concerns.
  • Reliability is addressed through a low blast radius, shutdown behavior, and billing alerts.
  • Security is handled with a locked-down security group, least-privilege IAM roles, Secrets Manager, and tag-based permission control.
  • The main lesson is that the Well-Architected Framework applies continuously, and the best design can change when cost or other constraints change.