Just writing a quick thing to say this blog is generated with a tool called Hugo. It takes a bunch of Markdown files and generates some static HTML files, one per post, plus some extra files for scaffolding (e.g. landing page, list of tags, etc).

There are also a bunch of metadata that can be specified (via a yaml blob at the top of each .md file) which describe which tags apply to a post, the creation time, short description, etc. For example, this page’s metadata is:

---
title: "Hugo"
description: |
  Small description of Hugo static site generator.  
author: Aimee
date: 2024-07-19T15:09:50-04:00
tags:
- devlog
- intro
mood: []
toc: false
draft: true
---

Just writing a quick thing to say this blog is generated with
...

If a post has the draft: true property, then it is not, by default, generated into a static file and thus is not publically available on the site. Generally, I run hugo server -D --renderToMemory locally when writing a new post to (a) render drafts (-D flag), and (b) to output to memory only (--renderToMemory flag) so as to not pollute the output directory.

When I’m ready to publish, I remove the draft: true property and run hugo --minify to generate minified static HTML pages in the output directory. I then commit the newly generated files to Git.

Deploying

Deploying to my actual webserver is done via some Git magic. After Hugo generates files, those are committed to a Git repo. On the actual blog webserver (a k8s nginx service) there is a helper process that polls/fetches updates from the Git repo.

So, nginx serves the static files from the most recent commit to Git. The repo is checked on a regular basis for updates. Since the site is just static files, nothing needs to compile on the server-side and no processes/binaries need to restart. It just serves whatever is checked out onto the local disk.