GreatSoftwares - From Monolith to Billions

GreatPages hired me to write code. A year later I was CTO. Between those two sentences sat one failing monolith and one argument I chose not to have.

The product let anyone publish a site in minutes. The monolith under it crashed often, hit ceilings no one could explain, and hid which part had broken. Everyone agreed it was failing. They disagreed about when to stop and rebuild, and a rewrite that size is one long bet that pays out only at the end, on a platform where customers publish right now.

A new developer had no standing to make everyone wait for that bet, so I built the better version beside the one running. I rebuilt the queries behind the worst failures right in production, no migration window, each earning its traffic under real load first. When the customer charts went down and support drowned, the replacement was already live. I moved traffic onto it and the platform came back. That was the argument, made in production instead of a meeting.

As CTO I broke the monolith into services with the platform still online, monitoring built into the migration instead of bolted on after. Capacity went to the components under pressure instead of the whole platform, which brought the bill down. One failure stopped dragging its neighbors with it, and when something broke, monitoring named the service.

Then there were the certificates

Every custom domain needed a Let's Encrypt certificate renewed every ninety days, and expiry does not slow a page, it takes it off the internet. A known vendor promised to migrate all of ours in about a day. I built a second path on Cloudflare first, on our own schedule, expecting never to use it. One day became six. Thousands of customer pages sat near expiry. I cut over to Cloudflare and ran the migration myself. No page went down, because the fallback was finished before the failure started.

Over two years the platform went from a monolith that could not carry its own growth to one serving billions of visits. The same move each time: keep what works running, build its replacement beside it, never let someone else's bad week become your outage.