Managed hosting is a boring superpower

Self-hosting has never been easier. Docker, DigitalOcean, Caddy, GitHub Actions — for about $12 a month and an afternoon of setup, you can run a production site with automatic HTTPS, continuous deployment, and a managed database.

So why do we still charge a monthly retainer for managed hosting? Because the hard part of hosting has never been the setup. It’s the 2am Saturday.

What breaks, actually breaks

In five years of running production sites, here’s what’s actually gone wrong:

• A cron job filling up the disk with log rotations that stopped working after an apt upgrade
• A TLS certificate failing to auto-renew because the ACME challenge couldn’t reach the server through a new firewall rule
• A database migration locking a critical table during a deploy
• An upstream dependency shipping a broken release that took the container out on restart
• A DDoS from a single abusive IP that the host’s default config didn’t block

Every one of these required a human looking at logs within minutes, not hours. A monitoring system that pages you is table stakes. Someone who answers the page is the whole product.

What managed hosting actually is

It’s four things:

• A human on call when something breaks
• A disaster recovery plan that has been tested this quarter
• Logs, metrics, and alerts configured for your specific workload
• Someone who knows the stack well enough to fix it, not just restart it

You can buy each one separately. Bundle them into one engagement and you get a site that stays up.

The goal of good hosting is to be boring. You should forget it exists.