Uptime & monitoring, the honest policy

What gets monitored, continuously

Uptime pings from an external service (not the same server the site lives on, a monitor that dies with the site is decoration), certificate expiry (the most common silent killer), DNS resolution, and, where the platform exposes it, error rates and deploy health. Backups are monitored for existence and freshness, because a backup job that quietly failed in March is discovered in June by people who needed it in April. Stores additionally get checkout-path checks, because a store that renders fine but cannot take payment is down in every way that matters, the checkout triage exists because of how often that one slips past uptime pings.

The honest math of "99.9% uptime"

Ninety-nine point nine percent means about 43.8 minutes of downtime per month, and that is the good tier. 99.5% is 3.65 hours. Any host or maintainer quoting "four nines" (4.4 minutes a year) without naming infrastructure is marketing, not arithmetic. The realistic promise: infrastructure chosen so downtime is rare and short, monitoring so it is known fast, and a human who acts on it the moment business hours allow. What is never promised: instant human response at 3am, because a one-person practice that claims a 3am pager is either lying or exhausted, and both end badly for your site.

What happens when something goes down

During business hours (IST): the alert lands, diagnosis starts, you get a message with what is known and what is being done, not silence while I "work on it." Most incidents (host hiccups, certificate lapses, a bad plugin update) resolve within the same window, and the broken-update discipline means updates that cause them were staged-tested in the first place. Outside business hours: the monitoring keeps watching, alerts queue, and response starts at the top of the next business day. For stores and revenue-critical sites that find this unacceptable, the honest advice is a host with a 24/7 escalation tier, named on the hosting page, with me coordinating the moment hours resume. Stating that limit plainly is the difference between a policy and a brochure.

The monthly report, the part that actually builds trust

Every month: uptime percentage with incident log, what was updated and why, backup status including the restore-test result, certificate and domain expiry runway, and anything that needs a decision. Short, factual, human. The maintenance retainer runs this whole discipline as its core loop; without it, monitoring is a smoke detector with no one home.