← Course Index
Ref
Primary Source
Better Stack — Uptime Monitoring Guide
Practical guide to setting up uptime monitoring with realistic alerting thresholds. Read →
Why Uptime Monitoring?
Sentry catches errors in running code. Uptime monitoring catches when your app stops running entirely . A monitor pings your /health endpoint every 30–60 seconds. If it doesn't respond, you get alerted. You know before your users do.
The /health Endpoint
app.get('/health', async (req, res) => {
const health = { status: 'healthy', uptime: process.uptime() };
try {
await db.query('SELECT 1');
health.database = 'ok';
} catch {
health.database = 'down';
health.status = 'degraded';
}
try {
await redis.ping();
health.redis = 'ok';
} catch {
health.redis = 'down';
health.status = 'degraded';
}
res.status(health.status === 'healthy' ? 200 : 503).json(health);
});
Free Uptime Tools
Tool Free tier Best for
UptimeRobot 50 monitors, 5-min checks Getting started Better Stack 10 monitors, 3-min checks Startups — great integrations Freshping 50 monitors, 1-min checks Fastest free checks
Alerting Best Practices
Alert on 2+ consecutive failures — not just 1 (reduces false alarms)Alert channels: Slack #alerts + PagerDuty/OpsGenie for on-callSSL cert expiry monitoring: alert 30 days beforeStatus page: publish at status.yourapp.com for transparent communication
Incident Response
Acknowledge — confirm someone is lookingAssess — docker compose ps, docker compose logs -f apiMitigate — restart service or rollback deployCommunicate — update status pagePostmortem — write what happened, why, prevention
💡
Write a runbook before you need it: "if alert X fires, do steps 1-5." At 3am you want a checklist, not a problem-solving session.
Check Your Understanding
1. /health returns 200 but doesn't check the database. What problem does this cause?
The endpoint will be slow to respond If the database goes down, the monitor reports "healthy" — you don't get alerted /health must return a string, not JSON Uptime monitors don't parse JSON responses
2. You get 3 false alarm alerts weekly because your server hiccups for 30 seconds. Fix?
Switch from Slack to email alerts Increase check frequency to 30 seconds Require 2+ consecutive failures before alerting Check the dashboard manually instead of using alerts
3. Your SSL cert expired. A user found out and emailed you. What monitoring would have prevented this?
Sentry error tracking Server uptime monitoring SSL certificate expiry monitoring — alert 30 days before expiry Google Search Console