The case for boring infrastructure.

We run Clikn on Postgres, Redis, and a handful of Node processes behind Cloudflare. No Kubernetes. No service mesh. No data lake. Here’s why the boring stack is the one that lets us sleep.

The temptation

Every early-stage company has a moment where someone says “what if we need to scale?” and reaches for Kubernetes. It’s the wrong question. The right question is: what’s the simplest thing that could work at 100× our current load? For us, the answer was Postgres and a single replica.

We’ve watched peers spend six months standing up service meshes they don’t need. That’s six months of features unshipped, bugs unchased, customers unheard.

“The stack that lets you sleep is the stack that’s still boring in eighteen months.”

What we actually use

Postgres 16 with pg_bouncer, a Redis cluster for hot counters, Node 20 for the API, and Cloudflare Workers for the redirect path itself. The redirect never hits an origin server — it’s resolved entirely at the edge against a KV store that’s updated by a webhook when a link is created or modified.

That’s it. That’s the whole thing.