What monitoring systems measure
- Uptime — does the server respond at all?
- Response time — how fast the first byte arrives.
- SSL validity — certificate expiry and chain health.
- DNS resolution — does the name still resolve correctly?
- Content checks — is the expected text or status code present?
- Synthetic transactions — does login or checkout still complete end-to-end?
Designing alerts that get acted on
Bad alerts are worse than no alerts. Follow three rules: (1) require multi-region confirmation before paging — one node hiccuping is not an outage, (2) tier severity so SSL-expires-in-7-days is an email and origin-server-down is a page, (3) route by business hours so 3 a.m. alerts only fire for true emergencies.
Health metrics worth tracking long-term
Beyond raw uptime, track: mean time to detect (MTTD), mean time to recover (MTTR), error budget consumption, and Apdex-style satisfaction scores. These metrics tell you whether reliability is improving over time, not just whether last week was bad.
The business case for monitoring
Downtime is the most expensive bug you'll ever ship. Even a basic monitoring setup pays for itself the first time it catches an SSL expiry, DNS misconfiguration, or origin outage before customers do. For SaaS, e-commerce, and any business with paid acquisition, the marketing cost of a single bad outage exceeds years of monitoring spend.