How we made a Ruby method 200x faster · Campsite

One morning, Campsite was slow. Normally snappy interactions had loading spinners and delayed data. Our observability tooling let us know it wasn’t just us, and it gave us some hints about what was happening. An Axiom monitor sent us a “high API request queue time” message via our Campsite integration. Web server processes were fully utilized, and requests were forced to wait. Fly.io metrics showed HTTP response times were significantly elevated, but didn’t show an increase in traffic. PlanetScale insights didn’t show any increase in database query latency. The spikes in request queue times and response times lined up with a deploy, so with no better leads, we reverted the change, which had refactored how we transform rich text. Request queue times and response times returned to normal levels, and Campsite felt fast again.