Monitoring
Check endpoints, journeys, servers and containers on a schedule, and hear about it the moment something breaks.
"Is it up?" is the easiest thing to monitor and the least useful. Most incidents that reach a customer are an endpoint that is up, returning 200, and wrong.
Creating a monitor#
- Open Monitoring and press New monitor.Point it at a saved request, a collection, or a URL.
- Set how often it runs.Every minute for something critical; every fifteen for the rest.
- Add assertions.This is the part that matters — see below.
- Choose who gets told.A monitor nobody is notified about only tells you what broke after you already knew.
What is worth watching#
Roughly in order of how often each earns its keep:
| Check | Catches |
|---|---|
| A deep health check — assert on the body, not just the status | A dependency down behind a still-serving app |
| Certificate expiry | The most avoidable total outage there is |
| Latency percentiles | A missing index, an N+1 that shipped, a cache not being hit |
| Response shape — a JSON Schema assertion | A field renamed or retyped; every client breaks, status stays 200 |
| A multi-step journey | Integration failures between individually healthy services |
| Servers and containers | The reason, when the endpoint check gives you only the symptom |
If you have nothing today, start with two: one deep health check per service asserting on the body, and certificate expiry on every public hostname. Between them they catch most outages that would otherwise reach a customer.
Other kinds#
- Synthetic — drive a real browser through a journey, for what an API check cannot see
- Server — CPU, memory and disk on a host
- Docker — container health and restart loops
- Status pages — publish the results on a public URL
Alerts#
- Set a failure threshold.Alerting on a single failed check is how alerts become noise people ignore.
- Choose channels.Email, webhook, or an incoming integration.
Frequently asked questions#
How do I monitor an API endpoint?
Create a monitor pointed at the endpoint or a saved request, set an interval, and add assertions on the response body rather than only the status code. Configure notifications so a failure reaches someone.
Can I get alerted before my SSL certificate expires?
Yes. A certificate expiry monitor fails ahead of the expiry date rather than on it, which turns a total outage into a ticket.