Observability

Application Health Check Guide

Design health endpoints that separate liveness, readiness, dependencies and degraded service states. Last updated September 1, 2026.

Design health endpoints that separate liveness, readiness, dependencies and degraded service states. This Formalint reference is designed for developers, DBAs, platform engineers and support teams who need an ordered troubleshooting path.

The page keeps the work practical: collect evidence, avoid leaking secrets, make the smallest safe change and leave a note another engineer can repeat.

When to use it

Practical workflow

StepWhat to verify
Define the signalDecide whether the symptom should be visible in logs, metrics, traces, health checks or alerts.
Keep fields searchableUse stable event names, correlation IDs, service names and safe identifiers.
Tune for actionAlerts should point to a user-visible problem, runbook or clear ownership path.

Command or evidence sample

GET /healthz
GET /readyz
GET /livez

Review checklist

  1. Write down the affected host, endpoint, job, service or browser context.
  2. Redact tokens, session cookies, passwords and customer identifiers.
  3. Capture timestamps, headers, logs or command output before changing settings.
  4. Prefer staged rollout, report-only mode, preview checks or single-URL cache purges where possible.
  5. Link the final note to a related Formalint page so the next step is obvious.

Common mistake

The common mistake is logging more text without making the event searchable, safe and tied to ownership.

Formalint's rule for these areas is simple: make the hidden system boundary visible before you tune it.

Related Formalint references

Continue with Api Correlation Id Logging Guide, Linux Journalctl Guide, Nginx Access Log Analysis Guide.

Frequently asked questions

Is this page enough for production approval?

No. Use it as a diagnostic and review aid, then follow your team's release, security and compliance process.

Can I paste real secrets or logs here?

No. Keep secrets and customer data out of browser tools. Use redacted examples and preserve sensitive evidence only in approved internal systems.