bzzz/pkg/integration/slurp_reliability.go

package integration

import (
	"context"
	"crypto/sha256"
	"encoding/json"
	"fmt"
	"log"
	"math"
	"math/rand"
	"os"
	"path/filepath"
	"sync"
	"time"
)

// CircuitState represents the state of a circuit breaker
type CircuitState int

const (
	CircuitClosed CircuitState = iota
	CircuitOpen
	CircuitHalfOpen
)

// String returns string representation of circuit state
func (s CircuitState) String() string {
	switch s {
	case CircuitClosed:
		return "CLOSED"
	case CircuitOpen:
		return "OPEN"
	case CircuitHalfOpen:
		return "HALF_OPEN"
	default:
		return "UNKNOWN"
	}
}

// CircuitBreaker implements circuit breaker pattern for SLURP client
type CircuitBreaker struct {
	mu                sync.RWMutex
	state             CircuitState
	failureCount      int
	consecutiveFailures int
	lastFailureTime   time.Time
	nextRetryTime     time.Time
	
	// Configuration
	maxFailures       int           // Max failures before opening circuit
	cooldownPeriod    time.Duration // How long to stay open
	halfOpenTimeout   time.Duration // How long to wait in half-open before closing
	
	// Metrics
	totalRequests     int64
	successfulRequests int64
	failedRequests    int64
}

// NewCircuitBreaker creates a new circuit breaker
func NewCircuitBreaker(maxFailures int, cooldownPeriod, halfOpenTimeout time.Duration) *CircuitBreaker {
	return &CircuitBreaker{
		state:           CircuitClosed,
		maxFailures:     maxFailures,
		cooldownPeriod:  cooldownPeriod,
		halfOpenTimeout: halfOpenTimeout,
	}
}

// CanProceed checks if request can proceed through circuit breaker
func (cb *CircuitBreaker) CanProceed() bool {
	cb.mu.Lock()
	defer cb.mu.Unlock()
	
	cb.totalRequests++
	
	switch cb.state {
	case CircuitClosed:
		return true
		
	case CircuitOpen:
		if time.Now().After(cb.nextRetryTime) {
			cb.state = CircuitHalfOpen
			log.Printf("🔄 Circuit breaker moving to HALF_OPEN state")
			return true
		}
		return false
		
	case CircuitHalfOpen:
		return true
		
	default:
		return false
	}
}

// RecordSuccess records a successful operation
func (cb *CircuitBreaker) RecordSuccess() {
	cb.mu.Lock()
	defer cb.mu.Unlock()
	
	cb.successfulRequests++
	cb.failureCount = 0
	cb.consecutiveFailures = 0
	
	if cb.state == CircuitHalfOpen {
		cb.state = CircuitClosed
		log.Printf("✅ Circuit breaker closed after successful operation")
	}
}

// RecordFailure records a failed operation
func (cb *CircuitBreaker) RecordFailure() {
	cb.mu.Lock()
	defer cb.mu.Unlock()
	
	cb.failedRequests++
	cb.failureCount++
	cb.consecutiveFailures++
	cb.lastFailureTime = time.Now()
	
	if cb.failureCount >= cb.maxFailures && cb.state == CircuitClosed {
		cb.state = CircuitOpen
		cb.nextRetryTime = time.Now().Add(cb.cooldownPeriod)
		log.Printf("🚫 Circuit breaker opened due to %d consecutive failures", cb.consecutiveFailures)
	}
}

// GetStats returns circuit breaker statistics
func (cb *CircuitBreaker) GetStats() map[string]interface{} {
	cb.mu.RLock()
	defer cb.mu.RUnlock()
	
	return map[string]interface{}{
		"state":                cb.state.String(),
		"total_requests":       cb.totalRequests,
		"successful_requests":  cb.successfulRequests,
		"failed_requests":      cb.failedRequests,
		"current_failures":     cb.failureCount,
		"consecutive_failures": cb.consecutiveFailures,
		"last_failure_time":    cb.lastFailureTime,
		"next_retry_time":      cb.nextRetryTime,
	}
}

// IdempotencyManager handles idempotency key generation and tracking
type IdempotencyManager struct {
	keys   map[string]time.Time
	mu     sync.RWMutex
	maxAge time.Duration
}

// NewIdempotencyManager creates a new idempotency manager
func NewIdempotencyManager(maxAge time.Duration) *IdempotencyManager {
	im := &IdempotencyManager{
		keys:   make(map[string]time.Time),
		maxAge: maxAge,
	}
	
	// Start cleanup goroutine
	go im.cleanupExpiredKeys()
	
	return im
}

// GenerateKey generates a stable idempotency key for an event
func (im *IdempotencyManager) GenerateKey(discussionID, eventType string, timestamp time.Time) string {
	// Create 5-minute time buckets to handle slight timing differences
	bucket := timestamp.Truncate(5 * time.Minute)
	
	// Generate stable hash
	data := fmt.Sprintf("%s_%s_%d", discussionID, eventType, bucket.Unix())
	hash := sha256.Sum256([]byte(data))
	return fmt.Sprintf("hmmm_%x", hash[:8]) // Use first 8 bytes for shorter key
}

// IsProcessed checks if an idempotency key has been processed recently
func (im *IdempotencyManager) IsProcessed(key string) bool {
	im.mu.RLock()
	defer im.mu.RUnlock()
	
	processTime, exists := im.keys[key]
	if !exists {
		return false
	}
	
	// Check if key is still valid (not expired)
	return time.Since(processTime) <= im.maxAge
}

// MarkProcessed marks an idempotency key as processed
func (im *IdempotencyManager) MarkProcessed(key string) {
	im.mu.Lock()
	defer im.mu.Unlock()
	
	im.keys[key] = time.Now()
}

// cleanupExpiredKeys periodically removes expired idempotency keys
func (im *IdempotencyManager) cleanupExpiredKeys() {
	ticker := time.NewTicker(im.maxAge / 2) // Cleanup twice as often as expiry
	defer ticker.Stop()
	
	for range ticker.C {
		im.mu.Lock()
		now := time.Now()
		expired := make([]string, 0)
		
		for key, processTime := range im.keys {
			if now.Sub(processTime) > im.maxAge {
				expired = append(expired, key)
			}
		}
		
		for _, key := range expired {
			delete(im.keys, key)
		}
		
		if len(expired) > 0 {
			log.Printf("🧹 Cleaned up %d expired idempotency keys", len(expired))
		}
		
		im.mu.Unlock()
	}
}

// DeadLetterQueue handles failed events that need to be retried later
type DeadLetterQueue struct {
	queueDir  string
	mu        sync.RWMutex
	items     map[string]*DLQItem
	maxRetries int
}

// DLQItem represents an item in the dead letter queue
type DLQItem struct {
	Event         SlurpEvent `json:"event"`
	FailureReason string     `json:"failure_reason"`
	RetryCount    int        `json:"retry_count"`
	NextRetryTime time.Time  `json:"next_retry_time"`
	FirstFailed   time.Time  `json:"first_failed"`
	LastFailed    time.Time  `json:"last_failed"`
}

// NewDeadLetterQueue creates a new dead letter queue
func NewDeadLetterQueue(queueDir string, maxRetries int) (*DeadLetterQueue, error) {
	if err := os.MkdirAll(queueDir, 0755); err != nil {
		return nil, fmt.Errorf("failed to create queue directory: %w", err)
	}
	
	dlq := &DeadLetterQueue{
		queueDir:   queueDir,
		items:      make(map[string]*DLQItem),
		maxRetries: maxRetries,
	}
	
	// Load existing items from disk
	if err := dlq.loadFromDisk(); err != nil {
		log.Printf("⚠️ Failed to load DLQ from disk: %v", err)
	}
	
	return dlq, nil
}

// Enqueue adds a failed event to the dead letter queue
func (dlq *DeadLetterQueue) Enqueue(event SlurpEvent, reason string) error {
	dlq.mu.Lock()
	defer dlq.mu.Unlock()
	
	eventID := dlq.generateEventID(event)
	now := time.Now()
	
	// Check if event already exists in DLQ
	if existing, exists := dlq.items[eventID]; exists {
		existing.RetryCount++
		existing.FailureReason = reason
		existing.LastFailed = now
		existing.NextRetryTime = dlq.calculateNextRetry(existing.RetryCount)
		
		log.Printf("💀 Updated DLQ item %s (retry %d/%d)", eventID, existing.RetryCount, dlq.maxRetries)
	} else {
		// Create new DLQ item
		item := &DLQItem{
			Event:         event,
			FailureReason: reason,
			RetryCount:    1,
			NextRetryTime: dlq.calculateNextRetry(1),
			FirstFailed:   now,
			LastFailed:    now,
		}
		
		dlq.items[eventID] = item
		log.Printf("💀 Added new item to DLQ: %s", eventID)
	}
	
	// Persist to disk
	return dlq.saveToDisk()
}

// GetReadyItems returns items that are ready for retry
func (dlq *DeadLetterQueue) GetReadyItems() []*DLQItem {
	dlq.mu.RLock()
	defer dlq.mu.RUnlock()
	
	now := time.Now()
	ready := make([]*DLQItem, 0)
	
	for _, item := range dlq.items {
		if item.RetryCount <= dlq.maxRetries && now.After(item.NextRetryTime) {
			ready = append(ready, item)
		}
	}
	
	return ready
}

// MarkSuccess removes an item from the DLQ after successful retry
func (dlq *DeadLetterQueue) MarkSuccess(eventID string) error {
	dlq.mu.Lock()
	defer dlq.mu.Unlock()
	
	delete(dlq.items, eventID)
	log.Printf("✅ Removed successfully retried item from DLQ: %s", eventID)
	
	return dlq.saveToDisk()
}

// MarkFailure updates retry count for failed retry attempt
func (dlq *DeadLetterQueue) MarkFailure(eventID string, reason string) error {
	dlq.mu.Lock()
	defer dlq.mu.Unlock()
	
	if item, exists := dlq.items[eventID]; exists {
		item.RetryCount++
		item.FailureReason = reason
		item.LastFailed = time.Now()
		item.NextRetryTime = dlq.calculateNextRetry(item.RetryCount)
		
		if item.RetryCount > dlq.maxRetries {
			log.Printf("💀 Item exceeded max retries, keeping in DLQ for manual review: %s", eventID)
		}
	}
	
	return dlq.saveToDisk()
}

// GetStats returns DLQ statistics
func (dlq *DeadLetterQueue) GetStats() map[string]interface{} {
	dlq.mu.RLock()
	defer dlq.mu.RUnlock()
	
	ready := 0
	exhausted := 0
	waiting := 0
	
	now := time.Now()
	for _, item := range dlq.items {
		if item.RetryCount > dlq.maxRetries {
			exhausted++
		} else if now.After(item.NextRetryTime) {
			ready++
		} else {
			waiting++
		}
	}
	
	return map[string]interface{}{
		"total_items":     len(dlq.items),
		"ready_for_retry": ready,
		"waiting":         waiting,
		"exhausted":       exhausted,
		"max_retries":     dlq.maxRetries,
	}
}

// calculateNextRetry calculates the next retry time using exponential backoff with jitter
func (dlq *DeadLetterQueue) calculateNextRetry(retryCount int) time.Time {
	// Exponential backoff: 2^retryCount minutes with jitter
	baseDelay := time.Duration(math.Pow(2, float64(retryCount))) * time.Minute
	
	// Add jitter (±25% random variation)
	jitter := time.Duration(rand.Float64()*0.5-0.25) * baseDelay
	delay := baseDelay + jitter
	
	// Cap at 1 hour maximum
	if delay > time.Hour {
		delay = time.Hour
	}
	
	return time.Now().Add(delay)
}

// generateEventID creates a unique ID for an event
func (dlq *DeadLetterQueue) generateEventID(event SlurpEvent) string {
	data := fmt.Sprintf("%s_%s_%s_%d", 
		event.EventType, 
		event.Path, 
		event.CreatedBy,
		event.Timestamp.Unix())
	
	hash := sha256.Sum256([]byte(data))
	return fmt.Sprintf("dlq_%x", hash[:8])
}

// saveToDisk persists the DLQ to disk
func (dlq *DeadLetterQueue) saveToDisk() error {
	filePath := filepath.Join(dlq.queueDir, "dlq_items.json")
	
	data, err := json.MarshalIndent(dlq.items, "", "  ")
	if err != nil {
		return fmt.Errorf("failed to marshal DLQ items: %w", err)
	}
	
	return os.WriteFile(filePath, data, 0644)
}

// loadFromDisk loads the DLQ from disk
func (dlq *DeadLetterQueue) loadFromDisk() error {
	filePath := filepath.Join(dlq.queueDir, "dlq_items.json")
	
	data, err := os.ReadFile(filePath)
	if err != nil {
		if os.IsNotExist(err) {
			return nil // No existing queue file, start fresh
		}
		return fmt.Errorf("failed to read DLQ file: %w", err)
	}
	
	return json.Unmarshal(data, &dlq.items)
}

// BackoffStrategy calculates retry delays with exponential backoff and jitter
type BackoffStrategy struct {
	initialDelay time.Duration
	maxDelay     time.Duration
	multiplier   float64
	jitterFactor float64
}

// NewBackoffStrategy creates a new backoff strategy
func NewBackoffStrategy(initialDelay, maxDelay time.Duration, multiplier, jitterFactor float64) *BackoffStrategy {
	return &BackoffStrategy{
		initialDelay: initialDelay,
		maxDelay:     maxDelay,
		multiplier:   multiplier,
		jitterFactor: jitterFactor,
	}
}

// GetDelay calculates the delay for a given attempt number
func (bs *BackoffStrategy) GetDelay(attempt int) time.Duration {
	if attempt <= 0 {
		return bs.initialDelay
	}
	
	// Exponential backoff
	delay := time.Duration(float64(bs.initialDelay) * math.Pow(bs.multiplier, float64(attempt-1)))
	
	// Apply maximum delay cap
	if delay > bs.maxDelay {
		delay = bs.maxDelay
	}
	
	// Add jitter to avoid thundering herd
	jitter := time.Duration(rand.Float64()*bs.jitterFactor*2-bs.jitterFactor) * delay
	delay += jitter
	
	// Ensure delay is never negative
	if delay < 0 {
		delay = bs.initialDelay
	}
	
	return delay
}