Every HTTP check records how long your endpoint took, and anything slower than 10s counts as down. Push your own counters and gauges for the numbers a check cannot see: queue depth, database latency, error rate.
Works with your stack
It's easy to track signups. It's harder to know if your background worker is stuck, or if that new query you wrote is timing out for 5% of users.
When things feel "sluggish", you shouldn't have to guess why. You need raw, technical counters and gauges that tell you exactly which part of your system is sweating.
Counters and gauges from Node, Python, Ruby, Java, .NET and Go. Measure how long things take and how many things happened, then watch the line bend a day before it takes the service down with it.
Response time recorded on every HTTP check
Your own counters and gauges from six SDKs
Charts that show the slowdown before the outage