Practical guides on uptime monitoring, alert escalation, and on-call workflows, written by the team building failover.io. If you run something that can't quietly go down, start here.
Practical walkthroughs on uptime monitoring, alerting, and on-call, published regularly.
A monitor watching from a single location can't tell your outage from its own network problems, so it pages you for both. This guide covers the false positive problem and why a failure should be confirmed from more than one vantage point before anyone's phone rings.
Read the guide →A site can serve "something went wrong" with a perfectly healthy status code, and a status-code monitor calls it up. This guide covers the ways broken pages ship behind 200s and how keyword checks catch the error your users actually see.
Read the guide →99.9% sounds close to perfect and allows 8 hours 46 minutes of downtime a year. This guide puts real hours on every nine, explains error budgets in plain terms, and covers which nine you should actually promise.
Read the guide →An expired TLS certificate takes the whole site down at a second you could see coming for months, and auto-renewal fails quietly when it fails. This guide covers watching the days-remaining number instead of the valid/invalid flag, with escalation at 30, 7, and 1 day.
Read the guide →A cron job that stops running throws no error and makes no request, so a normal monitor never sees it fail. This guide covers heartbeat monitoring: the job pings on each run, and the alert fires on the ping that never arrives.
Read the guide →PagerDuty prices per seat for a question a small team just needs answered: whose turn is it tonight. This guide covers running a timezone-aware on-call rotation, with overrides for real life, wired straight to your escalation chain.
Read the guide →A status page answers "is it down?" once, in public, instead of in twenty support emails during the outage. This guide covers building one from your existing monitors, theming it to your brand, and embedding it on your own site.
Read the guide →A stalled node answers every RPC call in 30ms with HTTP 200 while its block height sits frozen. The first hour this detection existed, it caught an amendment-blocked node in a public production pool. How it works and how to set it up.
Read the guide →Stripe does email you about failing webhooks, but late, and only for failures it can see from its side. A broken handler can lose days of events before that email arrives. This guide covers how to detect a stalled webhook pipeline within minutes using heartbeat monitoring.
Read the guide →Escalation isn't just for big ops teams. This guide covers how a one-person or small team can build an alert chain that reliably reaches a human, without buying enterprise on-call software.
Read the guide →Email and Slack alerts are easy to sleep through. This guide covers why a ringing phone is the most reliable way to catch a 3 a.m. outage, the options for setting it up, and the trade-offs of each.
Read the guide →A node can return HTTP 200 and still be broken: stalled, out of sync, or erroring inside the JSON body. This guide covers what to actually check and how to monitor an RPC endpoint properly.
Read the guide →5 monitors, 60-second checks, all 8 free-tier channels. No credit card. Commercial use allowed.
Start monitoring free →