Uploaded July 2025 | Updated September 2026, 2 weeks ago
makeitwork.tv 👈 full version shipping any week now
We implement the following one-liner bash script the Elixir Way™:
dig cdn-2025-02-25.internal AAAA +short | while read ipv6; do curl -so /dev/null -w "%{http_code} %{method} %{url_effective}\n" -X PURGE "http://[$ipv6]:9000/"; done
Find the code and the conversation in this pull request 🐙 github.com/thechangelog/changelog.com/pull/549
🤔 THE PROBLEM: CDN is serving stale content, with some entries expired for over 4 minutes. We need a way to explicitly purge cache across all CDN instances rather than waiting for user requests to trigger refreshes.
🙇♂️ THE SOLUTION: DNS lookups to discover all instances and coordinate purges across them.
🧑🔧 THE DNS THING (02:26): Using Fly.io's internal DNS system, we implement IPv6 lookups to discover all CDN instances. Claude Code helps. It works the first time 🎉
🐞 BUGS & PREDATORS (3:00): Undefined function errors, mock testing issues, and the classic double-slash URL bug. No Claude Code, just experience.
🧐 DID IT WORK? (6:56): Let's force push straight into production and see what happens 😱
This session isn't just a tutorial - it's an authentic look at collaborative coding, where DNS magic meets Elixir elegance to conquer distributed caching.
If you're into DevOps, CDNs, or just love seeing bugs squashed in real time, this gives you the tools to understand (and maybe implement) your own purge system.
00:00 The Problem
00:46 The Solution
02:26 The DNS thing
03:00 Bugs & Predators
04:43 Let's fire this up
06:56 Let's force push it
07:25 Is it working?
08:53 No no no
makeitwork.tv 👈 full version shipping any week now
We implement the following one-liner bash script the Elixir Way™:
dig cdn-2025-02-25.internal AAAA +short | while read ipv6; do curl -so /dev/null -w "%{http_code} %{method} %{url_effective}\n" -X PURGE "http://[$ipv6]:9000/"; done
Find the code and the conversation in this pull request 🐙 github.com/thechangelog/changelog.com/pull/549
🤔 THE PROBLEM: CDN is serving stale content, with some entries expired for over 4 minutes. We need a way to explicitly purge cache across all CDN instances rather than waiting for user requests to trigger refreshes.
🙇♂️ THE SOLUTION: DNS lookups to discover all instances and coordinate purges across them.
🧑🔧 THE DNS THING (02:26): Using Fly.io's internal DNS system, we implement IPv6 lookups to discover all CDN instances. Claude Code helps. It works the first time 🎉
🐞 BUGS & PREDATORS (3:00): Undefined function errors, mock testing issues, and the classic double-slash URL bug. No Claude Code, just experience.
🧐 DID IT WORK? (6:56): Let's force push straight into production and see what happens 😱
This session isn't just a tutorial - it's an authentic look at collaborative coding, where DNS magic meets Elixir elegance to conquer distributed caching.
If you're into DevOps, CDNs, or just love seeing bugs squashed in real time, this gives you the tools to understand (and maybe implement) your own purge system.
00:00 The Problem
00:46 The Solution
02:26 The DNS thing
03:00 Bugs & Predators
04:43 Let's fire this up
06:56 Let's force push it
07:25 Is it working?
08:53 No no no
