We're very good with servers

Here's ours. Every machine written down in code, every deploy through tests, monitoring on everything, backups we've restored from.

Every one of them is a line in one file. The network diagram is generated from that file, so it can't drift from reality without a check failing. If a machine isn't in the file, it doesn't exist as far as we're concerned. If it is in the file and isn't reporting, someone gets told.

It's dull, and it's why we sleep at night.

Monitoring that admits when it doesn't know

A check that can't see what it's checking reports unknown. We treat unknown as a problem: it never gets rounded up to green, and a dashboard full of green we can't vouch for is worse than one showing a few gaps honestly.

Backups we've actually restored from

A backup you've never restored is a hope, not a backup. Ours run on a schedule, a copy goes offsite, and restore tests run on a schedule too.

We also keep a list of what should be backed up and check reality against it. A new machine that nobody added to the backups shows up as missing, not as silence.

Silence is never good news

Every scheduled job checks in. If one stops checking in, that's an alert, because a job that's died quietly looks exactly like a job with nothing to report.

Fixes with a human in charge

Some fixes are routine: restart this service, rerun that sync. We've written those down as remedies, and the estate proposes them when something fails.

Nothing runs until the alert's been delivered and someone has said yes: a person, or an agent working within limits a person set, for named fixes and a dated window. No AI model sits in the path that runs the fix. When it's done, it checks the fix worked.

A report you can read

For the servers we look after for clients, we produce a monthly report: what was checked, what passed, what didn't, and what didn't run at all. That last line is the one most reports leave out.

The apps we build and keep running can run on servers we manage for you, with monitoring, backups and a monthly report.

Where should it live? On servers we run, or in your own account with the provider you already use, DigitalOcean or similar. Either way it's written down in code, so if you ever leave, it leaves with you.

Got a server that only exists as somebody's memory of what they typed once? The first job is writing it down. Then it can be monitored, backed up and rebuilt like the rest.