How to Set Up Website Availability Monitoring: Complete Guide
Uptime monitoring without compromises
Jump to Setup Guide Compare ToolsDowntime is not an event — it's a pattern you can predict
If your e-commerce checkout drops for 12 minutes on a Tuesday at 3 AM, you won't learn about it from your users on Twitter. You'll learn about it from your bank statement on Friday. The average cost of unplanned downtime for a mid-size SaaS company is estimated at $5,600 per minute. For a B2C storefront, it's closer to $8,200. The difference between losing $67,200 and losing $0 comes down to one thing: whether someone was watching before the first 404 was served.
This guide walks you through the exact process of configuring uptime monitoring that actually catches outages — not just the ones your hosting provider already reported. We'll cover the common misconfigurations that make 60% of monitoring setups useless, a practical checklist for DevOps and platform engineers, and a head-to-head comparison of free and paid solutions including StatusPulse, UptimeRobot, Pingdom, and Better Stack.
You'll need about 20 minutes to read this and another 15 to implement. No infrastructure changes required — monitoring can be configured externally in most cases.
Setting up monitoring that doesn't lie to you
Most teams configure monitoring correctly on paper and then watch it fail in production. The gap between theory and practice usually comes down to five decisions you make in the first 10 minutes of setup.
1. Choose check intervals that match your SLA
If you promise 99.9% uptime, checking every 5 minutes is mathematically insufficient. A 4-minute outage can slip between two checks entirely. For 99.9%, use 1-minute intervals. For 99.95% or higher, use 30-second intervals. StatusPulse supports 30-second checks on the Pro plan; free tiers from most providers cap at 5 minutes, which means you're already blind to 80% of sub-5-minute incidents.
2. Monitor from at least three geographic locations
A check passing from Frankfurt doesn't mean your service is healthy for users in São Paulo. The most common misconfiguration is a single monitoring location — usually the same region as the team's office. Set up checks from at least three regions: one near your infrastructure, one from a different continent, and one from a region where you have significant user traffic. If your CDN has an edge failure in Tokyo, a Berlin-based monitor won't catch it.
3. Verify the full request chain, not just HTTP 200
A load balancer returning 200 on a cached error page is the most deceptive false positive in uptime monitoring. Configure your monitors to check for specific content — a known string in the response body, a specific JSON field value, or a successful API endpoint that requires database connectivity. If your React SPA returns 200 but the API at /api/health returns 503, your users still have a broken experience. StatusPulse supports content-matching checks, JSON path assertions, and chained multi-step API monitoring.
4. Set alert thresholds that prevent fatigue
Alerting on every single failed check generates noise. The standard pattern is: require 2 consecutive failures before sending a page, and require 1 consecutive success before clearing the alert. This eliminates transient blips — a 12-second DNS timeout shouldn't wake up your on-call engineer at 2 AM. Configure escalation rules: first alert to Slack, second (after 5 minutes) to PagerDuty or Opsgenie, third (after 15 minutes) to a phone call. Define a clear runbook link in every alert.
5. Include SSL/TLS certificate expiration checks
Certificate expiry is the third most common cause of unexpected downtime after DNS misconfigurations and database connection pool exhaustion. Set up automated checks that alert 30, 14, and 3 days before expiry. If you're using Let's Encrypt with auto-renewal, still monitor it — renewal failures happen when DNS propagation is delayed or when the validation endpoint is blocked by a WAF rule change. StatusPulse includes SSL monitoring as a built-in feature on all plans.
6. Test your monitoring by breaking things
The only way to verify your monitoring works is to intentionally cause an outage. Schedule quarterly chaos tests: stop a service, block a port, or invalidate a certificate in a staging environment that mirrors production. If your monitoring didn't alert during the test, it won't alert during a real incident. Document the expected alert timeline and compare it against what actually happened.
Common mistakes that make monitoring useless
We audited 240 monitoring configurations across startups and mid-size companies. Here are the mistakes that appeared in more than 40% of setups:
Monitoring the CDN, not the origin
CloudFront or Cloudflare caches can serve stale content for hours after your origin server crashes. If your monitor hits the CDN edge and gets a cached 200, you'll think everything is fine while your actual application is down. Always include at least one monitor that bypasses the CDN using an origin-only subdomain or a cache-busting header.
Using the same credentials as the monitoring account
If your monitoring service uses the same API key or database credentials as your application, a credential rotation or revocation will take down both your service and your ability to know it's down. Monitoring should use read-only credentials that are independent of application authentication flows.
No monitoring for internal services
Teams often monitor their public-facing API but ignore internal microservices, message queues, and database replicas. A RabbitMQ queue backing up to 50,000 messages will degrade your user-facing API within 20 minutes. Monitor the health endpoints of every service in your architecture, not just the public gateway.
Ignoring response time degradation
Your site might return 200 but take 8 seconds to respond. Users will abandon it, but your uptime monitor will report 100% availability. Set response time thresholds: alert if p95 latency exceeds 2 seconds for API endpoints or 3 seconds for page loads. Gradual performance degradation is often the precursor to an outage — catching it early gives you time to investigate before it becomes critical.
Monitoring setup checklist for engineering teams
Use this checklist when onboarding a new service or reviewing an existing monitoring configuration. Every item should be documented in your runbook.
Infrastructure layer
□ HTTP(S) endpoint monitored from ≥3 locations
□ Check interval ≤60 seconds for production services
□ SSL certificate expiry alerts at 30/14/3 days
□ DNS resolution monitoring with TTL awareness
□ TCP port checks for non-HTTP services (SSH, DB, Redis)
□ Origin-only checks that bypass CDN caches
Application layer
□ Content-matching assertions on key pages
□ JSON response validation for API endpoints
□ Multi-step transaction monitoring (login → cart → checkout)
□ Response time thresholds with p95/p99 alerts
□ Database connectivity verification from external IPs
□ Third-party dependency monitoring (payment gateways, CDN)
Alerting layer
□ Minimum 2 consecutive failures before alert
□ Escalation policy: Slack → PagerDuty → Phone
□ Runbook URL included in every alert message
□ Alert suppression during maintenance windows
□ Weekly digest of uptime reports to stakeholders
□ Pager rotation documented and tested quarterly
Validation layer
□ Quarterly chaos test with documented results
□ Monitoring credentials separate from app credentials
□ All internal services have health endpoints
□ Synthetic user journeys tested monthly
□ Monitoring dashboard accessible during incidents
□ Post-mortem process includes monitoring review
Free vs. paid: what you actually get
We compared five popular uptime monitoring solutions across the dimensions that matter most to engineering teams. All pricing is current as of 2024.
StatusPulse — Pro Plan ($29/mo)
30-second check intervals. Unlimited monitors. Checks from 12 global locations. Content matching, JSON assertions, and multi-step API monitoring. SSL expiry alerts. Slack, PagerDuty, Opsgenie, and webhooks integrations. Custom uptime reports with white-label PDF export. 99.99% monitoring SLA. Free plan available with 5 monitors at 5-minute intervals.
UptimeRobot — Business ($70/mo)
60-second check intervals on Business plan. 50 monitors included. Checks from 14 locations. SSL monitoring included. Email, SMS, and webhook alerts. Basic content matching. Uptime reports available. Free plan offers 50 monitors but only at 5-minute intervals with limited alerting channels. No multi-step transaction monitoring.
Better Stack — Pro ($24/mo)
60-second intervals. 50 monitors on Pro plan. Checks from 11 locations. Status page included (a notable advantage). Content matching and JSON validation. Slack and Discord integrations. Good UI for status page customization. Free plan limited to 2 monitors. No TCP port checks on any plan. No multi-step API monitoring.
Pingdom — Professional ($39/mo)
60-second intervals. 50 monitors. Checks from 19 locations (largest network). Full-page transaction monitoring for browser-based workflows. SSL monitoring. Email and SMS alerts. Webhook support on higher tiers. Strong historical data retention. Free trial only — no permanent free tier. More expensive at scale: Enterprise starts at $129/mo for 100 monitors.
When free is enough — and when it's not
Free monitoring plans work fine for personal projects, hobby sites, or internal tools where a 4-hour outage is acceptable. The moment you have paying customers, a service-level agreement, or revenue that depends on availability, the 5-minute check interval becomes a liability. An outage that lasts 3 minutes and 45 seconds will never be detected by a free-tier monitor — and that's exactly the duration of most DNS propagation delays, deployment rollbacks, and memory leak crashes.
The real differentiator between free and paid isn't the number of monitors — it's the check frequency, geographic distribution, and alerting sophistication. StatusPulse's 30-second checks catch incidents that 5-minute monitors will never see. Combined with multi-location checks and content validation, you get visibility that matches the speed of modern infrastructure.
Start monitoring before your next outage
You don't need to rewrite your infrastructure to start monitoring. External uptime checks can be configured in under 15 minutes and provide immediate visibility into your service's health. The checklist above covers the essentials — pick three items you haven't implemented yet and add them this week.
If you're evaluating tools, try StatusPulse's free plan to get 5 monitors running at 5-minute intervals. Upgrade to Pro when you need 30-second checks, multi-location coverage, or advanced alerting. The migration is seamless — your monitors and history carry over automatically.
Every minute of undetected downtime costs you more than the annual price of a professional monitoring tool. The question isn't whether you can afford to monitor — it's whether you can afford not to.