health/CHANGELOG.md

Changelog

All notable changes to this module will be documented in this file.

The format is based on Keep a Changelog, and this module adheres to Semantic Versioning.

0.9.0 - 2026-03-18

Added

• Level type — int representing component criticality; zero value is LevelCritical
• LevelCritical constant (value 0) — a failing critical component sets overall status to DOWN and returns HTTP 503
• LevelDegraded constant (value 1) — a failing degraded component sets overall status to DEGRADED and returns HTTP 200
• Checkable interface — HealthCheck(ctx context.Context) error, Name() string, Priority() Level; implemented by infrastructure components
• Logger interface — duck-typed minimal logger (Debug, Info, Warn, Error, WithContext) satisfied by logz.Logger without importing logz
• ComponentStatus struct — JSON-serialisable per-component result with fields status string, latency string (omitempty), and error string (omitempty)
• Response struct — JSON-serialisable overall response with fields status string and components map[string]ComponentStatus
• NewHandler(logger Logger, checks ...Checkable) http.Handler — constructs an http.Handler that runs all registered checks concurrently, collects results via a buffered channel, and writes a JSON Response
• Parallel check execution: each Checkable.HealthCheck call runs in its own goroutine; a buffered channel sized to the check count prevents goroutine leaks
• Per-request 5-second deadline derived from context.WithTimeout(r.Context(), 5*time.Second) and propagated to all check goroutines
• Overall status aggregation: UP if all checks pass; DEGRADED (HTTP 200) if at least one degraded component fails and no critical component fails; DOWN (HTTP 503) if any critical component fails
• Check latency measurement: each goroutine records time.Since(start) and includes it as a string in ComponentStatus
• Content-Type: application/json response header set on every response

Design Notes

• All checks run in parallel goroutines and report through a buffered channel; the buffer is sized exactly to the number of registered checks at construction time, so the handler is guaranteed to drain all results without blocking even if it returns early.
• The two-level criticality model (LevelCritical / LevelDegraded) gives orchestrators and load balancers a clean binary HTTP signal (200 vs 503) while still surfacing partial degradation in the JSON body for monitoring systems.
• The Logger and Checkable interfaces are defined entirely within this package using duck typing — no micro-lib module is imported, keeping health a pure stdlib package (Tier 0/1) that infra packages can satisfy without a circular dependency.