Beacon

The brief

Existing uptime tools were either enterprise-priced or toy-grade. I wanted the middle: honest checks from multiple regions, alerts that respect on-call sanity, and public status pages fast enough to survive being the page everyone hits during an outage.

The build

• Check scheduler built on Laravel queues with per-monitor jitter, so thousands of checks don't thundering-herd on the minute.
• Alert escalation chains — notify the primary, wait, escalate — with quiet hours and dedup so one incident equals one thread, not forty pings.
• Status pages rendered to static HTML on state change and served straight from the edge cache. During an incident, the origin serves almost nothing.

Transmission log

• Runs 1,100+ monitors on a single modest VPS.
• Status pages: p99 load time 87ms during a real incident spike.
• The escalation logic has since been extracted and reused in two client projects.