health/health.go

package health

import (
	"context"
	"encoding/json"
	"net/http"
	"time"
)

// Level represents the criticality of a component to the overall application health.
type Level int

const (
	// LevelCritical indicates a component essential for the application.
	// If any critical component is DOWN, the overall status is DOWN (503).
	LevelCritical Level = iota
	// LevelDegraded indicates a component that is important but not essential.
	// If a degraded component is DOWN, the overall status is DEGRADED (200).
	LevelDegraded
)

// Checkable is the interface that infrastructure components implement.
type Checkable interface {
	HealthCheck(ctx context.Context) error
	Name() string
	Priority() Level
}

// Logger is the minimal interface health needs — satisfied by logz.Logger via duck typing.
type Logger interface {
	Debug(msg string, args ...any)
	Info(msg string, args ...any)
	Warn(msg string, args ...any)
	Error(msg string, err error, args ...any)
	WithContext(ctx context.Context) Logger
}

// ComponentStatus represents the health state of an individual component.
type ComponentStatus struct {
	Status  string `json:"status"`
	Latency string `json:"latency,omitempty"`
	Error   string `json:"error,omitempty"`
}

// Response is the JSON body returned by the health handler.
type Response struct {
	Status     string                     `json:"status"`
	Components map[string]ComponentStatus `json:"components"`
}

type handler struct {
	logger Logger
	checks []Checkable
}

// NewHandler returns an http.Handler for the health endpoint.
// Runs all checks concurrently with a 5-second timeout.
// Returns 200 (UP/DEGRADED) or 503 (DOWN).
func NewHandler(logger Logger, checks ...Checkable) http.Handler {
	return &handler{logger: logger, checks: checks}
}

func (h *handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
	logger := h.logger.WithContext(r.Context())
	logger.Debug("health: running checks")

	ctx, cancel := context.WithTimeout(r.Context(), 5*time.Second)
	defer cancel()

	type result struct {
		name     string
		status   ComponentStatus
		priority Level
	}

	resChan := make(chan result, len(h.checks))

	for _, check := range h.checks {
		go func(chk Checkable) {
			start := time.Now()
			err := chk.HealthCheck(ctx)
			latency := time.Since(start).String()

			status := "UP"
			errMsg := ""
			if err != nil {
				errMsg = err.Error()
				if chk.Priority() == LevelDegraded {
					status = "DEGRADED"
				} else {
					status = "DOWN"
				}
			}

			resChan <- result{
				name:     chk.Name(),
				priority: chk.Priority(),
				status: ComponentStatus{
					Status:  status,
					Latency: latency,
					Error:   errMsg,
				},
			}
		}(check)
	}

	overallStatus := "UP"
	httpStatus := http.StatusOK
	components := make(map[string]ComponentStatus)

	for range h.checks {
		res := <-resChan
		components[res.name] = res.status

		switch res.status.Status {
		case "DOWN":
			overallStatus = "DOWN"
			httpStatus = http.StatusServiceUnavailable
		case "DEGRADED":
			if overallStatus == "UP" {
				overallStatus = "DEGRADED"
			}
		}
	}

	resp := Response{Status: overallStatus, Components: components}
	w.Header().Set("Content-Type", "application/json")
	w.WriteHeader(httpStatus)
	if err := json.NewEncoder(w).Encode(resp); err != nil {
		logger.Error("health: failed to encode response", err)
	}
}