2026-10-07 · Operations & infrastructure
What “operating it” actually involves
Installing an application takes an afternoon. Keeping it running for years is a different job — here is what it is made of.
Almost every guide on the internet ends at the same point: the application starts, the browser shows a login screen, done. After that begins the part nobody writes about, because it is unspectacular and never stops.
Availability is a task, not a state
A service is not available because it was started once, but because someone notices when it stops. That means a check from the outside, a route the alert travels on, and a human who reacts. Without the third part the first two are decoration.
The uncomfortable truth is that most outages are not crashes. It is a full filesystem, an expired certificate, a renewal job that has been failing quietly for six weeks. None of it is visible while using the system — until everyone calls at once.
Updates are decisions, not routine
Security updates can simply be applied. Version upgrades cannot: they come with decisions about the database, the interfaces, and configuration nobody has touched in two years. Whoever avoids that decision still makes one — they stay on a version that will eventually stop receiving security updates.
That is why every update needs a way back. Not as a theory, but as a tested procedure.
A backup that was never restored is not a backup
Everybody knows the sentence and almost nobody follows it. Backups run, the file grows, everyone is content. Whether a working system can actually be rebuilt from it is something you only know once you have done it — and the worst moment to find out is the real emergency.
I test restores regularly. Not because a standard says so, but because I have found backups that had been faithfully saving an empty database for months.
The bill at the end
Add up monitoring, updates, certificates, backups, the occasional incident and the documentation, and each application carries a modest but constant load. For a single application nobody hires for that; for ten it pays off, but then you need someone reachable at night too.
This is exactly the question behind my Managed Apps offering: is it worth taking this part on, or is it worth handing it over.