CosmicClash/server/allocator/health.go

package allocator

import (
	"net/http"
	"sync"
	"time"
)

// Health records only complete successful cycles. Readiness ages out when
// provider or durable-store work repeatedly fails or stalls, while liveness
// remains independent so Kubernetes does not restart a healthy process for a
// dependency outage.
type Health struct {
	mu                  sync.RWMutex
	lastSuccessfulCycle time.Time
}

func (h *Health) ObserveSuccessfulCycle(at time.Time) {
	if h == nil || at.IsZero() {
		return
	}
	h.mu.Lock()
	h.lastSuccessfulCycle = at
	h.mu.Unlock()
}

func (h *Health) Ready(now time.Time, maxStale time.Duration) bool {
	if h == nil || now.IsZero() || maxStale <= 0 {
		return false
	}
	h.mu.RLock()
	lastSuccess := h.lastSuccessfulCycle
	h.mu.RUnlock()
	return !lastSuccess.IsZero() && !now.Before(lastSuccess) && now.Sub(lastSuccess) <= maxStale
}

// RoleHandler exposes metrics plus distinct process-liveness and dependency-
// progress readiness endpoints on the allocator's private listener.
func RoleHandler(metrics *Metrics, health *Health, maxStale time.Duration, now func() time.Time) http.Handler {
	mux := http.NewServeMux()
	mux.Handle("/metrics", MetricsHandler(metrics))
	mux.HandleFunc("/healthz", methodGet(func(w http.ResponseWriter, _ *http.Request) {
		w.Header().Set("Content-Type", "text/plain; charset=utf-8")
		_, _ = w.Write([]byte("ok\n"))
	}))
	mux.HandleFunc("/readyz", methodGet(func(w http.ResponseWriter, _ *http.Request) {
		if now == nil || !health.Ready(now(), maxStale) {
			http.Error(w, "allocator has no recent successful cycle", http.StatusServiceUnavailable)
			return
		}
		w.Header().Set("Content-Type", "text/plain; charset=utf-8")
		_, _ = w.Write([]byte("ready\n"))
	}))
	return mux
}

func methodGet(next http.HandlerFunc) http.HandlerFunc {
	return func(w http.ResponseWriter, r *http.Request) {
		if r.Method != http.MethodGet {
			w.WriteHeader(http.StatusMethodNotAllowed)
			return
		}
		next(w, r)
	}
}