Prompt
My uptime checks keep firing false positives
Latest observation
Nov 5, 2026
OpenAI APIWeb search: off
False positives in uptime checks are usually caused by one of these:
Common causes
- Too aggressive thresholds: One failed ping or one slow response marks the service down.
- Short timeouts: The monitor gives up before the app/server can respond.
- Transient network issues: Brief packet loss, DNS hiccups, or regional routing problems.
- Load spikes / cold starts: Serverless apps, autoscaling, or newly restarted services can be briefly unavailable.
- Health endpoint problems:
/healthmay depend on databases or downstream services and fail even when the core app is fine. - Monitor location issues: Only one region checking can make localized network problems look like outages.
- TLS / certificate warnings: Expired or misconfigured certs may fail checks unexpectedly.
- Proxy/CDN/WAF interference: Security layers may block the monitoring probe or rate-limit it.
- Flaky dependencies: If your health check calls external APIs, their issues become your “uptime” issues.
How to reduce them
- Use a lightweight health endpoint that only verifies the app process is alive, not every dependency.
- Add retries / require multiple failures before alerting.
- Increase timeout slightly if your app is sometimes slow.
- Check from multiple regions to avoid localized network false alarms.
- Separate “live” and “ready” checks:
- liveness = app is running
- readiness = app can serve traffic
- Avoid alerting on single blips; use a rolling window.
- Exclude known maintenance windows or deployments from alerting.
- Inspect logs around alert times to see if failures are real or monitor-induced.
A good baseline
For many services:
- timeout: 5–10s
- interval: 1 min
- alert after: 2–3 consecutive failures
- multiple check locations: 2+
If you want, I can help you troubleshoot this more precisely. Send me:
- what uptime tool you use,
- what endpoint it checks, and
- an example false-positive alert time/log.