Launch Week

The transition began with a final broadcast to our V2 users, signaling the end of an era. To ensure data integrity, we took a massive 16GB database dump and handed it off for the heavy lifting. The core challenge wasn’t just moving the data, but transforming it; we had to map our existing MySQL architecture into a new Postgres structure on Hetzner. This required a delicate dance of freezing new account creations on the V3 beta while we meticulously reshaped years of user information to fit our new vision.

By mid-afternoon, the real “chewing” began as our migration scripts merged the transformed V2 data with the existing V3 beta environment. Despite a minor bug requiring a second pass, we successfully synchronized the databases by 6:00 PM. We ultimately decided to hold the live switch until the following morning. Moving to production in the evening is a recipe for disaster; we wanted the full weight of our dev team available to handle the inevitable surge in server load and user feedback that comes with “flipping the switch.”

Launch day was a whirlwind of final checks: analytics, DNS settings, and OAuth redirects. At noon, we pulled the trigger. After a brief period of maintenance mode and some frantic .env variable updates, we were live.

The biggest lesson? Always purge the Cloudflare cache. Once we scaled the servers and the caching settled, the infrastructure stabilized. I spent the day in a marathon virtual meeting, debugging in tandem with the team. It was exhilarating but draining; I ended the day with a long walk in the park to recover from the screen fatigue.

The morning brought a “welcome gift”: the database had pinned at 100% CPU overnight due to the activity rankings. We doubled the instance size as a stopgap while we hunted down the inefficient queries.

I also spent time triaging feedback using Fider. It’s easy to let the negative comments sting, people naturally resist change, but the silent majority seemed to enjoy the update. We focused the rest of the day on performance tuning to ensure the database wouldn’t choke during the next peak.

The latter half of the week shifted toward SEO and stability. I audited our meta tags and de-indexed redundant pages (like individual competitions) to ensure Google focuses on our high-value practice modes. By Friday, things were stable enough to actually downsize our server instances, a great sign that our optimizations were working.

Saturday was for housekeeping and catching a few lingering bugs, like a stubborn issue with account deletions. By Sunday, I finally took a breath to monitor the servers and prep for the week ahead.

It is hard to believe it’s only been a week; the sheer volume of work & feedback makes it feel much longer. While there is still a long road ahead for bigfixing, performance tuning and long-term stability, I am happy with the relaunch. I’m looking forward to shipping new features and continually improving 10FastFingers for the whole world.