API uptime monitoring

What this team needs to watch

An API has no human watching it render. Its users are other programs that retry, time out, and eventually fail their own customers. The useful signal is whether documented endpoints return the right status, the right shape, and an acceptable latency under real conditions.

• Assert on response shape, not just a 200 status.
• Certificate expiry on the API host takes everything down at once.
• Latency creep often precedes a hard failure.

How Sandglass covers it

Run HTTP checks against representative endpoints, add content checks that assert a known field in the JSON response, and put SSL certificate checks on the API hostname so an expiring certificate never becomes a surprise outage.

• Group checks by service, environment, or client so production alerts stay separate from staging noise.
• Combine HTTP, ping, TCP, content, and SSL certificate checks to match each failure mode you care about.
• Add heartbeat checks for scheduled jobs that no customer watches directly but everyone depends on.

Where teams get it wrong

A reachable port is not a working API. A check that only confirms a TCP connection will stay green while the service returns 500s, so assert on the response body, not just the handshake.