What I build · 70
Monitoring that tells the truth
A dashboard you have learned to ignore is worse than no dashboard. I build monitoring around a small number of alerts that are always worth reading.
What I build
- Watching the right things. Disk space before it runs out. The backup that did not run last night. The certificate that expires next month. The job that quietly stopped running three weeks ago — silence is the commonest failure in small systems, and it is the one most monitoring never catches.
- A heartbeat for scheduled work. Each job reports when it runs; the alert fires when a report is missing, not when something crashes.
- A quiet channel — email or a phone notification, your choice — with a rule that the same fault is never reported twice in a day and nothing pages anyone at 3 a.m. for something that can wait until morning.
- Honest states. A check that cannot reach the thing it watches says “unknown,” never “fine.” A stale reading is labelled stale.
- Optional: I watch it too. If you keep me on, the same alerts reach me, and I tell you what I did about them.
What I don’t do
I don’t install a monitoring service that reports to a vendor’s cloud and goes blind when your link does.