Wrote about how to safely run a globally unique process in an Elixir cluster, and a scary story from the past!
Learn about :global for registering unique processes in a cluster, and where using it can go horribly wrong. Or at least, a little bit wrong. And how to avoid weird edge cases. This was a lot of fun and nostalgic to write, hope you enjoy it
Nice article. It makes sense in retrospect, but I didn’t know you could handle global processes in the supervision tree by handling the already-started error and link+return the other pid. Cool trick
I think it would be worth mentioning Highlander in your article as well. It provides the same functionality as the SingletonManager proposed.