Uptime monitoring

Uptime monitoring is the automatic, repeated checking of a website, API or scheduled job, with an alert to the right people as soon as a check fails.

Why it matters for agencies

Agencies look after many client sites, and the usual way of finding out about an outage is the client ringing. Monitoring turns that into an alert for you first, with a record of how long the outage lasted.

How to do it or work it out

Monitor what the client would notice: the home page, key forms, the API, the domain's expiry and the scheduled jobs that should report in. Choose how often to check and who to alert.

Understand uptime percentages. Over a 30-day month, 99.9 percent uptime allows about 43 minutes of downtime, 99.5 percent allows about 3.6 hours.

Alert the right people, not everyone. Send outages to the on-call developer and a team channel, and keep slow-response warnings for a quieter route. Alert fatigue means the real alert gets ignored.

A worked example

A 30-day month has 43,200 minutes. At 99.9 percent uptime, 0.1 percent may be down, which is 43.2 minutes. At 99.5 percent, 0.5 percent may be down, which is 216 minutes, or 3.6 hours. A monitor checks a client's home page every 5 minutes and needs two failures before alerting. The site fails at 02:10 and again at 02:15, so the alert goes out at 02:15, and the incident records how long the outage lasted.

Common mistakes

  • Checking only the home page. It can be cached and healthy while checkout or login is broken, so add the checks a client would notice.
  • Forgetting expiry dates. A lapsed domain takes a site down just as surely as a server fault, and the date is known well in advance.
  • Alerting on a single failure. A brief network blip then wakes someone at night, so require two failures in a row.

All glossary terms. General information, not financial, legal or accounting advice.

Watch your clients' sites in Jasper

Monitor sites and APIs with HTTP, ping, DNS and heartbeat checks, and have the right people alerted first.