Every change I make to a live server starts the same way: with a copy of whatever I’m about to touch, and a written note of how to put it back.
It sounds slow. In practice it’s the opposite. When a change goes wrong — and on a long enough timeline, one always does — the difference between a five-minute rollback and an afternoon of guesswork is whether that copy exists.
Three habits that do the work
- Measure before changing. Look at real logs and real numbers first, so the change fixes the problem you actually have.
- Keep a way back. A timestamped backup beside the file, a rollback step written down, and a timer that undoes the change if I can’t prove it worked.
- Verify afterwards. “It restarted” isn’t the same as “it works” — check the thing the change was for, and check nothing else broke.
None of this is clever. It’s just the habits that keep websites and mailboxes boring, which is exactly what hosting should be.