futriix/internal/storage/transactions.go

/*
 * Copyright 2026 Safronov Grigorii
 *
 * Licensed under the CDDL, Version 1.0 (the "License");
 * you may not use this file except in compliance with the License.
 *
 * You may obtain a copy of the License at
 * https://opensource.org/licenses/CDDL-1.0
 */

// Файл: internal/storage/transactions.go
// Назначение: Batch-операции (замена полноценных ACID-транзакций) +
// Time-travel queries на основе WAL.
//
// АРХИТЕКТУРНЫЕ ИЗМЕНЕНИЯ (2026):
//   - Удалены все мьютексы и MVCC (MVCCManager, VisibilityMap,
//     ReadTimestampCache, DocumentVersion). Они давали огромную сложность
//     без реальной пользы для in-memory СУБД.
//   - Полноценные ACID-транзакции заменены на batch-операции.
//     Batch — это группа операций, которые применяются атомарно:
//     либо все, либо ни одна. Гарантия атомарности обеспечивается
//     через WAL: сначала пишем весь batch в WAL, потом применяем.
//   - SAGA-шаги используют batch для гарантии атомарности шага.
//   - Time-travel реализован через WAL: для получения версии документа
//     на момент времени проигрываем WAL до нужного LSN.
//   - Удалён AsyncRecoveryManager (был источником race conditions).
//     Восстановление теперь синхронное при старте.
//
// ИСПРАВЛЕНО (2026-10): добавлен реестр activeBatches в BatchManager,
// который используется eviction-механизмом (runtime_limits.go), чтобы
// не вытеснять документы, участвующие в незакоммиченной batch-операции.
// Ранее runtime_limits.go ссылался на удалённые globalTxManager и
// Transaction — теперь эти ссылки заменены на GetBatchManager() и *Batch.
//
// ДОБАВЛЕНО (2026-10, идемпотентность SAGA + выборы лидера):
//   - SagaPersistentStorage ведёт персистентную таблицу выполненных
//     ExecutionID (saga_executed_keys.json). Перед выполнением шага
//     и перед его компенсацией оркестратор проверяет эту таблицу,
//     чтобы повторный запуск SAGA после падения не приводил к
//     дублированию операций на удалённых узлах.
//   - SagaOrchestrator получил SagaLeaderElector, который берёт лидера
//     из Raft (через SagaRaftAccessor). Если Raft недоступен —
//     используется резервный режим «единственный узел — лидер».
//   - isLeader больше не выставляется в true безусловно: значение
//     обновляется фоновой горутиной leaderElectionLoop, которая
//     опрашивает Raft.State() и публикует изменения в orchestrator.
//   - Добавлен SagaLeaderState (персистентный) — чтобы после рестарта
//     узел знал, был ли он лидером, и корректно себя вёл до первых
//     выборов.
//   - Добавлен интерфейс SagaRaftAccessor, чтобы storage не зависел
//     напрямую от hashicorp/raft.
//
// ДОБАВЛЕНО (2026-10, ревизия):
//   - GetAllSagas у SagaManager — читает все SAGA (включая завершённые)
//     с диска через SagaPersistentStorage.ListAll(). Используется
//     REPL-командой "saga list --all".
//
// ИЗМЕНЕНО (2026-10): удалён AOF. BatchManager больше не содержит
// поле aof, и Commit не пишет в AOF. Единственный источник истины — WAL.

package storage

import (
	"bufio"
	"encoding/binary"
	"encoding/json"
	"fmt"
	"io"
	"os"
	"path/filepath"
	"sort"
	"sync"
	"sync/atomic"
	"time"

	"futriis/internal/config"
)

// =============================================================================
// БАЗОВЫЕ ТИПЫ
// =============================================================================

type TransactionID uint64

// BatchOperation — одна операция в batch.
type BatchOperation struct {
	Type       string                 `json:"type"` // insert, update, delete, restore
	Database   string                 `json:"database"`
	Collection string                 `json:"collection"`
	DocumentID string                 `json:"document_id"`
	Data       map[string]interface{} `json:"data,omitempty"`
	Version    uint64                 `json:"version,omitempty"`
}

// Batch представляет группу операций, применяемых атомарно.
type Batch struct {
	ID         TransactionID    `json:"id"`
	Operations []BatchOperation `json:"operations"`
	Timestamp  int64            `json:"timestamp"`
	Database   string           `json:"database,omitempty"`
	Collection string           `json:"collection,omitempty"`
	LSN        uint64           `json:"lsn,omitempty"`
}

// WALRecord — запись в WAL.
type WALRecord struct {
	CRC       uint32 `json:"crc"`
	Length    uint32 `json:"length"`
	Type      byte   `json:"type"` // 1 = batch
	Data      []byte `json:"data"`
	Timestamp int64  `json:"timestamp"`
	LSN       uint64 `json:"lsn"`
}

// =============================================================================
// КОНСТАНТЫ
// =============================================================================

const (
	WALSegmentSize   = 64 * 1024 * 1024
	WALSegmentPrefix = "wal_segment_"
	WALIndexPrefix   = "wal_index_"

	VersionPruneInterval = 5 * time.Minute

	SagaStateDir           = "saga_states"
	SagaStateFilePrefix    = "saga_state_"
	SagaStateFileSuffix    = ".json"
	SagaCheckpointInterval = 30 * time.Second
	SagaMaxRetries         = 5
	SagaRetryBackoff       = 100 * time.Millisecond

	SagaOrphanTimeout   = 5 * time.Minute
	SagaCleanupInterval = 1 * time.Minute

	// SagaExecutedKeysFile — персистентная таблица выполненных ExecutionID.
	SagaExecutedKeysFile = "saga_executed_keys.json"

	// SagaLeaderStateFile — персистентное состояние лидера SAGA.
	SagaLeaderStateFile = "saga_leader_state.json"

	// SagaLeaderElectionInterval — как часто опрашиваем Raft на предмет лидерства.
	SagaLeaderElectionInterval = 2 * time.Second
)

// =============================================================================
// CRC32
// =============================================================================

var crc32Table = [256]uint32{
	0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 0x076dc419, 0x706af48f,
	0xe963a535, 0x9e6495a3, 0x0edb8832, 0x79dcb8a4, 0xe0d5e91e, 0x97d2d988,
	0x09b64c2b, 0x7eb17cbd, 0xe7b82d07, 0x90bf1d91, 0x1db71064, 0x6ab020f2,
	0xf3b97148, 0x84be41de, 0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7,
	0x136c9856, 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 0x14015c4f, 0x63066cd9,
	0xfa0f3d63, 0x8d080df5, 0x3b6e20c8, 0x4c69105e, 0xd56041e4, 0xa2677172,
	0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b, 0x35b5a8fa, 0x42b2986c,
	0xdbbbc9d6, 0xacbcf940, 0x32d86ce3, 0x45df5c75, 0xdcd60dcf, 0xabd13d59,
	0x26d930ac, 0x51de003a, 0xc8d75180, 0xbfd06116, 0x21b4f4b5, 0x56b3c423,
	0xcfba9599, 0xb8bda50f, 0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924,
	0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 0x76dc4190, 0x01db7106,
	0x98d220bc, 0xefd5102a, 0x71b18589, 0x06b6b51f, 0x9fbfe4a5, 0xe8b8d433,
	0x7807c9a2, 0x0f00f934, 0x9609a88e, 0xe10e9818, 0x7f6a0dbb, 0x086d3d2d,
	0x91646c97, 0xe6635c01, 0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e,
	0x6c0695ed, 0x1b01a57b, 0x8208f4c1, 0xf50fc457, 0x65b0d9c6, 0x12b7e950,
	0x8bbeb8ea, 0xfcb9887c, 0x62dd1ddf, 0x15da2d49, 0x8cd37cf3, 0xfbd44c65,
	0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2, 0x4adfa541, 0x3dd895d7,
	0xa4d1c46d, 0xd3d6f4fb, 0x4369e96a, 0x346ed9fc, 0xad678846, 0xda60b8d0,
	0x44042d73, 0x33031de5, 0xaa0a4c5f, 0xdd0d7cc9, 0x5005713c, 0x270241aa,
	0xbe0b1010, 0xc90c2086, 0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f,
	0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 0x59b33d17, 0x2eb40d81,
	0xb7bd5c3b, 0xc0ba6cad, 0xedb88320, 0x9abfb3b6, 0x03b6e20c, 0x74b1d29a,
	0xead54739, 0x9dd277af, 0x04db2615, 0x73dc1683, 0xe3630b12, 0x94643b84,
	0x0d6d6a3e, 0x7a6a5aa8, 0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1,
	0xf00f9344, 0x8708a3d2, 0x1e01f268, 0x6906c2fe, 0xf762575d, 0x806567cb,
	0x196c3671, 0x6e6b06e7, 0xfed41b76, 0x89d32be0, 0x10da7a5a, 0x67dd4acc,
	0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5, 0xd6d6a3e8, 0xa1d1937e,
	0x38d8c2c4, 0x4fdff252, 0xd1bb67f1, 0xa6bc5767, 0x3fb506dd, 0x48b2364b,
	0xd80d2bda, 0xaf0a1a4c, 0x36034af6, 0x41047a60, 0xdf60efc3, 0xa867df55,
	0x316e8eef, 0x4669be79, 0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236,
	0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 0xc5ba3bbe, 0xb2bd0b28,
	0x2bb45a92, 0x5cb36a04, 0xc2d7ffa7, 0xb5d0cf31, 0x2cd99e8b, 0x5bdeae1d,
	0x9b64c2b0, 0xec63f226, 0x756aa39c, 0x026d930a, 0x9c0906a9, 0xeb0e363f,
	0x72076785, 0x05005713, 0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38,
	0x92d28e9b, 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 0x86d3d2d4, 0xf1d4e242,
	0x68ddb3f8, 0x1fda836e, 0x81be16cd, 0xf6b9265b, 0x6fb077e1, 0x18b74777,
	0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c, 0x8f659eff, 0xf862ae69,
	0x616bffd3, 0x166ccf45, 0xa00ae278, 0xd70dd2ee, 0x4e048354, 0x3903b3c2,
	0xa7672661, 0xd06016f7, 0x4969474d, 0x3e6e77db, 0xaed16a4a, 0xd9d65adc,
	0x40df0b66, 0x37d83bf8, 0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9,
	0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605, 0xcdd70693,
	0x54de5729, 0x23d967bf, 0xb3667a2e, 0xc4614ab8, 0x5d681b02, 0x2a6f2b94,
	0xb40bbe37, 0xc30c8ea1, 0x5a05df1b, 0x2d02ef8d,
}

func crc32(data []byte) uint32 {
	crc := uint32(0xFFFFFFFF)
	for _, b := range data {
		crc = (crc >> 8) ^ crc32Table[(crc^uint32(b))&0xFF]
	}
	return crc ^ 0xFFFFFFFF
}

// =============================================================================
// BATCH MANAGER — замена TransactionManager
// =============================================================================

// BatchManager управляет batch-операциями и WAL.
type BatchManager struct {
	nextBatchID atomic.Uint64
	wal         *SegmentedWALManager
	logger      LoggerInterface
	walPath     string
	stats       *BatchStats

	// activeBatches — реестр активных (незакоммиченных) batch'ей.
	activeBatches sync.Map
}

// BatchStats — статистика batch-операций.
type BatchStats struct {
	TotalBatches    atomic.Uint64
	TotalOperations atomic.Uint64
	TotalFailed     atomic.Uint64
	StartTime       time.Time
}

// =============================================================================
// ГЛОБАЛЬНЫЕ ПЕРЕМЕННЫЕ
// =============================================================================

var (
	globalBatchManager *BatchManager
	batchManagerOnce   sync.Once

	globalStorage *Storage
)

// InitBatchManager инициализирует batch-менеджер.
func InitBatchManager(walPath string) error {
	return InitBatchManagerWithConfig(walPath, nil)
}

// InitBatchManagerWithConfig инициализирует batch-менеджер с конфигурацией.
func InitBatchManagerWithConfig(walPath string, config map[string]interface{}) error {
	var err error
	batchManagerOnce.Do(func() {
		fsyncEnabled := true
		if config != nil {
			if v, ok := config["fsync_enabled"].(bool); ok {
				fsyncEnabled = v
			}
		}

		globalBatchManager = &BatchManager{
			walPath: walPath,
			stats:   &BatchStats{StartTime: time.Now()},
		}
		globalBatchManager.nextBatchID.Store(1)

		globalBatchManager.wal, err = NewSegmentedWALManager(
			filepath.Dir(walPath), fsyncEnabled, nil)
		if err != nil {
			return
		}
	})
	return err
}

// GetBatchManager возвращает глобальный batch-менеджер.
func GetBatchManager() *BatchManager { return globalBatchManager }

// SetBatchLogger устанавливает логгер.
func SetBatchLogger(logger LoggerInterface) {
	if globalBatchManager != nil {
		globalBatchManager.logger = logger
		if globalBatchManager.wal != nil {
			globalBatchManager.wal.logger = logger
		}
	}
}

// SetGlobalStorage устанавливает глобальное хранилище.
func SetGlobalStorage(s *Storage) { globalStorage = s }

// GetGlobalStorage возвращает глобальное хранилище.
func GetGlobalStorage() *Storage { return globalStorage }

// =============================================================================
// РЕЕСТР АКТИВНЫХ BATCH'ЕЙ
// =============================================================================

// RegisterBatch регистрирует batch как активный.
func (bm *BatchManager) RegisterBatch(b *Batch) {
	if bm == nil || b == nil {
		return
	}
	bm.activeBatches.Store(b.ID, b)
}

// UnregisterBatch снимает регистрацию batch.
func (bm *BatchManager) UnregisterBatch(b *Batch) {
	if bm == nil || b == nil {
		return
	}
	bm.activeBatches.Delete(b.ID)
}

// IsDocInActiveBatch проверяет, участвует ли документ в активной batch-операции.
func (bm *BatchManager) IsDocInActiveBatch(docID, dbName, collName string) bool {
	if bm == nil {
		return false
	}
	found := false
	bm.activeBatches.Range(func(key, value interface{}) bool {
		b, ok := value.(*Batch)
		if !ok || b == nil {
			return true
		}
		for _, op := range b.Operations {
			if op.DocumentID == docID &&
				op.Database == dbName &&
				op.Collection == collName {
				found = true
				return false
			}
		}
		return true
	})
	return found
}

// GetActiveBatchCount возвращает количество активных batch'ей.
func (bm *BatchManager) GetActiveBatchCount() int {
	if bm == nil {
		return 0
	}
	count := 0
	bm.activeBatches.Range(func(_, _ interface{}) bool {
		count++
		return true
	})
	return count
}

// =============================================================================
// BATCH API
// =============================================================================

// NewBatch создаёт новый batch.
func NewBatch(database, collection string) *Batch {
	if globalBatchManager == nil {
		_ = InitBatchManager("futriis.wal")
	}
	b := &Batch{
		ID:         TransactionID(globalBatchManager.nextBatchID.Add(1) - 1),
		Operations: make([]BatchOperation, 0, 16),
		Timestamp:  time.Now().UnixMilli(),
		Database:   database,
		Collection: collection,
	}
	globalBatchManager.RegisterBatch(b)
	return b
}

// AddInsert добавляет операцию вставки.
func (b *Batch) AddInsert(doc *Document) {
	b.Operations = append(b.Operations, BatchOperation{
		Type:       "insert",
		Database:   b.Database,
		Collection: b.Collection,
		DocumentID: doc.ID,
		Data:       doc.GetFields(),
		Version:    doc.Version,
	})
}

// AddUpdate добавляет операцию обновления.
func (b *Batch) AddUpdate(docID string, updates map[string]interface{}) {
	b.Operations = append(b.Operations, BatchOperation{
		Type:       "update",
		Database:   b.Database,
		Collection: b.Collection,
		DocumentID: docID,
		Data:       updates,
	})
}

// AddDelete добавляет операцию удаления.
func (b *Batch) AddDelete(docID string) {
	b.Operations = append(b.Operations, BatchOperation{
		Type:       "delete",
		Database:   b.Database,
		Collection: b.Collection,
		DocumentID: docID,
	})
}

// AddRestore добавляет операцию восстановления.
func (b *Batch) AddRestore(docID string) {
	b.Operations = append(b.Operations, BatchOperation{
		Type:       "restore",
		Database:   b.Database,
		Collection: b.Collection,
		DocumentID: docID,
	})
}

// Commit применяет batch атомарно.
func (b *Batch) Commit() error {
	if globalBatchManager == nil {
		return fmt.Errorf("batch manager not initialized")
	}

	defer globalBatchManager.UnregisterBatch(b)

	if len(b.Operations) == 0 {
		return nil
	}

	data, err := json.Marshal(b)
	if err != nil {
		return fmt.Errorf("failed to marshal batch: %v", err)
	}

	record := &WALRecord{Type: 1, Data: data}
	if err := globalBatchManager.wal.WriteSync(record); err != nil {
		globalBatchManager.stats.TotalFailed.Add(1)
		return fmt.Errorf("failed to write batch to WAL: %v", err)
	}

	if err := b.apply(); err != nil {
		globalBatchManager.stats.TotalFailed.Add(1)
		return fmt.Errorf("batch apply failed: %v", err)
	}

	globalBatchManager.stats.TotalBatches.Add(1)
	globalBatchManager.stats.TotalOperations.Add(uint64(len(b.Operations)))
	return nil
}

// apply применяет все операции batch.
func (b *Batch) apply() error {
	if globalStorage == nil {
		return fmt.Errorf("storage not initialized")
	}

	type appliedOp struct {
		opType string
		db     string
		coll   string
		docID  string
		oldDoc *Document
	}
	applied := make([]appliedOp, 0, len(b.Operations))

	rollback := func() {
		for i := len(applied) - 1; i >= 0; i-- {
			ao := applied[i]
			db, err := globalStorage.GetDatabase(ao.db)
			if err != nil {
				continue
			}
			coll, err := db.GetCollection(ao.coll)
			if err != nil {
				continue
			}
			switch ao.opType {
			case "insert":
				coll.PermanentDelete(ao.docID)
			case "update":
				if ao.oldDoc != nil {
					coll.Update(ao.docID, ao.oldDoc.GetFields())
				}
			case "delete":
				coll.RestoreDeleted(ao.docID)
			}
		}
	}

	for _, op := range b.Operations {
		db, err := globalStorage.GetDatabase(op.Database)
		if err != nil {
			rollback()
			return fmt.Errorf("database not found: %s", op.Database)
		}
		coll, err := db.GetCollection(op.Collection)
		if err != nil {
			rollback()
			return fmt.Errorf("collection not found: %s", op.Collection)
		}

		switch op.Type {
		case "insert":
			doc := NewDocumentWithID(op.DocumentID)
			for k, v := range op.Data {
				doc.SetField(k, v)
			}
			doc.Version = op.Version
			if err := coll.Insert(doc); err != nil {
				rollback()
				return err
			}
			applied = append(applied, appliedOp{
				opType: "insert", db: op.Database, coll: op.Collection, docID: op.DocumentID,
			})

		case "update":
			oldDoc, _ := coll.Find(op.DocumentID)
			var oldCopy *Document
			if oldDoc != nil {
				oldCopy = oldDoc.Clone()
			}
			if err := coll.Update(op.DocumentID, op.Data); err != nil {
				rollback()
				return err
			}
			applied = append(applied, appliedOp{
				opType: "update", db: op.Database, coll: op.Collection,
				docID: op.DocumentID, oldDoc: oldCopy,
			})

		case "delete":
			if err := coll.Delete(op.DocumentID); err != nil {
				rollback()
				return err
			}
			applied = append(applied, appliedOp{
				opType: "delete", db: op.Database, coll: op.Collection, docID: op.DocumentID,
			})

		case "restore":
			if err := coll.RestoreDeleted(op.DocumentID); err != nil {
				rollback()
				return err
			}
			applied = append(applied, appliedOp{
				opType: "restore", db: op.Database, coll: op.Collection, docID: op.DocumentID,
			})
		}
	}
	return nil
}

// =============================================================================
// TIME-TRAVEL QUERIES (WAL-based)
// =============================================================================

// GetDocumentAtTimestamp возвращает документ на указанный момент времени.
func GetDocumentAtTimestamp(docID string, timestamp int64) (*Document, error) {
	if globalBatchManager == nil {
		return nil, fmt.Errorf("batch manager not initialized")
	}
	if globalBatchManager.wal == nil {
		return nil, fmt.Errorf("WAL not initialized")
	}

	records, err := globalBatchManager.wal.ReadAll()
	if err != nil {
		return nil, fmt.Errorf("failed to read WAL: %v", err)
	}

	var current *Document
	for _, rec := range records {
		if rec.Timestamp > timestamp {
			break
		}
		if rec.Type != 1 {
			continue
		}
		var batch Batch
		if err := json.Unmarshal(rec.Data, &batch); err != nil {
			continue
		}
		for _, op := range batch.Operations {
			if op.DocumentID != docID {
				continue
			}
			switch op.Type {
			case "insert":
				doc := NewDocumentWithID(op.DocumentID)
				for k, v := range op.Data {
					doc.SetField(k, v)
				}
				doc.Version = op.Version
				doc.CreatedAt = batch.Timestamp
				doc.UpdatedAt = batch.Timestamp
				current = doc
			case "update":
				if current != nil {
					current.Update(op.Data)
					current.UpdatedAt = batch.Timestamp
				}
			case "delete":
				if current != nil {
					current.SoftDelete()
					current.DeletedAt = batch.Timestamp
				}
			case "restore":
				if current != nil {
					current.Restore()
					current.UpdatedAt = batch.Timestamp
				}
			}
		}
	}

	if current == nil {
		return nil, fmt.Errorf("document %s not found at timestamp %d", docID, timestamp)
	}
	return current, nil
}

// GetDocumentVersionsSince возвращает все версии документа после указанного времени.
func GetDocumentVersionsSince(docID string, since int64) ([]*Document, error) {
	if globalBatchManager == nil || globalBatchManager.wal == nil {
		return nil, fmt.Errorf("batch manager not initialized")
	}
	records, err := globalBatchManager.wal.ReadAll()
	if err != nil {
		return nil, err
	}

	versions := make([]*Document, 0)
	for _, rec := range records {
		if rec.Timestamp <= since || rec.Type != 1 {
			continue
		}
		var batch Batch
		if err := json.Unmarshal(rec.Data, &batch); err != nil {
			continue
		}
		for _, op := range batch.Operations {
			if op.DocumentID != docID {
				continue
			}
			doc := NewDocumentWithID(op.DocumentID)
			if op.Type == "insert" || op.Type == "update" {
				for k, v := range op.Data {
					doc.SetField(k, v)
				}
				doc.Version = op.Version
				doc.UpdatedAt = batch.Timestamp
				versions = append(versions, doc)
			}
		}
	}
	return versions, nil
}

// =============================================================================
// WAL MANAGER (Segmented)
// =============================================================================

// WALSegment представляет один сегмент WAL.
type WALSegment struct {
	ID       uint32
	File     *os.File
	Writer   *bufio.Writer
	Path     string
	StartLSN uint64
	EndLSN   uint64
	Size     int64
	mu       sync.Mutex
}

// WALIndexEntry — запись индекса WAL.
type WALIndexEntry struct {
	LSN       uint64
	SegmentID uint32
	Offset    int64
	Length    uint32
	Checksum  uint32
}

// WALIndexManager — индекс WAL для быстрого доступа.
type WALIndexManager struct {
	index     map[uint64]*WALIndexEntry
	segments  map[uint32]*WALSegment
	mu        sync.RWMutex
	indexPath string
}

// SegmentedWALManager — сегментированный WAL.
type SegmentedWALManager struct {
	segmentsDir      string
	segments         map[uint32]*WALSegment
	currentSegment   *WALSegment
	currentSegmentID uint32
	index            *WALIndexManager
	mu               sync.RWMutex
	writeChan        chan *WALRecord
	stopCh           chan struct{}
	wg               sync.WaitGroup
	batchSize        int
	logger           LoggerInterface
	fsyncEnabled     bool

	ackMu  sync.Mutex
	ackMap map[uint64]chan struct{}
}

// NewSegmentedWALManager создаёт новый сегментированный WAL.
func NewSegmentedWALManager(segmentsDir string, fsyncEnabled bool, logger LoggerInterface) (*SegmentedWALManager, error) {
	if err := os.MkdirAll(segmentsDir, 0755); err != nil {
		return nil, fmt.Errorf("failed to create segments dir: %v", err)
	}
	wm := &SegmentedWALManager{
		segmentsDir: segmentsDir,
		segments:    make(map[uint32]*WALSegment),
		index: &WALIndexManager{
			index:     make(map[uint64]*WALIndexEntry),
			segments:  make(map[uint32]*WALSegment),
			indexPath: filepath.Join(segmentsDir, WALIndexPrefix+"index.json"),
		},
		writeChan:    make(chan *WALRecord, 10000),
		stopCh:       make(chan struct{}),
		batchSize:    100,
		logger:       logger,
		fsyncEnabled: fsyncEnabled,
		ackMap:       make(map[uint64]chan struct{}),
	}
	if err := wm.loadExistingSegments(); err != nil {
		return nil, err
	}
	if err := wm.index.load(); err != nil {
		if logger != nil {
			logger.Warn(fmt.Sprintf("Failed to load WAL index: %v", err))
		}
	}
	if wm.currentSegment == nil {
		if err := wm.rotateSegmentInternal(); err != nil {
			return nil, err
		}
	}
	wm.wg.Add(1)
	go wm.writerLoop()
	return wm, nil
}

func (wm *SegmentedWALManager) loadExistingSegments() error {
	files, err := filepath.Glob(filepath.Join(wm.segmentsDir, WALSegmentPrefix+"*"))
	if err != nil {
		return err
	}
	for _, filePath := range files {
		var segmentID uint32
		if _, err := fmt.Sscanf(filepath.Base(filePath), WALSegmentPrefix+"%d.log", &segmentID); err != nil {
			continue
		}
		file, err := os.OpenFile(filePath, os.O_RDWR, 0644)
		if err != nil {
			continue
		}
		stat, _ := file.Stat()
		segment := &WALSegment{
			ID:       segmentID,
			File:     file,
			Writer:   bufio.NewWriterSize(file, 64*1024),
			Path:     filePath,
			Size:     stat.Size(),
			StartLSN: uint64(segmentID) * WALSegmentSize / 100,
		}
		wm.segments[segmentID] = segment
		if segmentID > wm.currentSegmentID {
			wm.currentSegmentID = segmentID
			wm.currentSegment = segment
		}
	}
	return nil
}

// rotateSegmentInternal — вызывается только при захваченном wm.mu.
func (wm *SegmentedWALManager) rotateSegmentInternal() error {
	newSegmentID := wm.currentSegmentID + 1
	segmentPath := filepath.Join(wm.segmentsDir, fmt.Sprintf(WALSegmentPrefix+"%d.log", newSegmentID))
	file, err := os.OpenFile(segmentPath, os.O_CREATE|os.O_APPEND|os.O_RDWR, 0644)
	if err != nil {
		return fmt.Errorf("failed to create segment: %v", err)
	}
	newSegment := &WALSegment{
		ID:       newSegmentID,
		File:     file,
		Writer:   bufio.NewWriterSize(file, 64*1024),
		Path:     segmentPath,
		StartLSN: wm.getCurrentLSNLocked(),
	}
	if wm.currentSegment != nil {
		wm.currentSegment.Writer.Flush()
		if wm.fsyncEnabled {
			RealFsyncWithRetry(wm.currentSegment.File, 3, 100*time.Millisecond)
		}
		wm.currentSegment.EndLSN = wm.getCurrentLSNLocked()
		wm.currentSegment.File.Close()
		FsyncDir(wm.segmentsDir)
	}
	wm.currentSegment = newSegment
	wm.currentSegmentID = newSegmentID
	wm.segments[newSegmentID] = newSegment
	if wm.logger != nil {
		wm.logger.Info(fmt.Sprintf("Created new WAL segment: %d", newSegmentID))
	}
	return nil
}

func (wm *SegmentedWALManager) getCurrentLSNLocked() uint64 {
	if wm.currentSegment == nil {
		return 1
	}
	return wm.currentSegment.StartLSN + uint64(wm.currentSegment.Size/100)
}

// Write — асинхронная запись.
func (wm *SegmentedWALManager) Write(record *WALRecord) error {
	record.Timestamp = time.Now().UnixMilli()
	wm.writeChan <- record
	return nil
}

// WriteSync — синхронная запись с гарантией fsync.
func (wm *SegmentedWALManager) WriteSync(record *WALRecord) error {
	record.Timestamp = time.Now().UnixMilli()

	wm.mu.Lock()
	lsn := wm.getCurrentLSNLocked() + 1
	record.LSN = lsn
	ackCh := make(chan struct{})
	wm.ackMu.Lock()
	wm.ackMap[lsn] = ackCh
	wm.ackMu.Unlock()
	wm.mu.Unlock()

	wm.writeChan <- record

	select {
	case <-ackCh:
		return nil
	case <-time.After(10 * time.Second):
		wm.ackMu.Lock()
		delete(wm.ackMap, lsn)
		wm.ackMu.Unlock()
		return fmt.Errorf("WAL write sync timeout (LSN %d)", lsn)
	}
}

func (wm *SegmentedWALManager) writerLoop() {
	defer wm.wg.Done()
	batch := make([]*WALRecord, 0, wm.batchSize)
	ticker := time.NewTicker(5 * time.Second)
	defer ticker.Stop()
	for {
		select {
		case record, ok := <-wm.writeChan:
			if !ok {
				if len(batch) > 0 {
					wm.flushBatch(batch)
				}
				return
			}
			batch = append(batch, record)
			if len(batch) >= wm.batchSize {
				wm.flushBatch(batch)
				batch = batch[:0]
			}
		case <-ticker.C:
			if len(batch) > 0 {
				wm.flushBatch(batch)
				batch = batch[:0]
			}
		case <-wm.stopCh:
			if len(batch) > 0 {
				wm.flushBatch(batch)
			}
			return
		}
	}
}

// flushBatch — записывает batch записей и подтверждает через ackMap.
func (wm *SegmentedWALManager) flushBatch(batch []*WALRecord) {
	wm.mu.Lock()

	for _, record := range batch {
		if wm.currentSegment.Size >= WALSegmentSize {
			if err := wm.rotateSegmentInternal(); err != nil {
				if wm.logger != nil {
					wm.logger.Error(fmt.Sprintf("Failed to rotate segment: %v", err))
				}
				continue
			}
		}
		data, err := json.Marshal(record)
		if err != nil {
			continue
		}
		lsnBytes := make([]byte, 8)
		binary.BigEndian.PutUint64(lsnBytes, record.LSN)
		crcData := append(lsnBytes, data...)
		record.CRC = crc32(crcData)
		header := make([]byte, 8)
		binary.BigEndian.PutUint32(header[0:4], uint32(len(data)))
		binary.BigEndian.PutUint32(header[4:8], record.CRC)
		if _, err := wm.currentSegment.Writer.Write(header); err != nil {
			continue
		}
		if _, err := wm.currentSegment.Writer.Write(data); err != nil {
			continue
		}
		wm.index.addEntry(&WALIndexEntry{
			LSN:       record.LSN,
			SegmentID: wm.currentSegment.ID,
			Offset:    wm.currentSegment.Size,
			Length:    uint32(len(data)),
			Checksum:  record.CRC,
		})
		wm.currentSegment.Size += int64(8 + len(data))
		wm.currentSegment.EndLSN = record.LSN
	}
	wm.currentSegment.Writer.Flush()
	if wm.fsyncEnabled {
		RealFsyncWithRetry(wm.currentSegment.File, 3, 100*time.Millisecond)
	}
	wm.index.save()

	wm.mu.Unlock()

	wm.ackMu.Lock()
	for _, record := range batch {
		if ch, ok := wm.ackMap[record.LSN]; ok {
			close(ch)
			delete(wm.ackMap, record.LSN)
		}
	}
	wm.ackMu.Unlock()
}

// Sync — синхронизация.
func (wm *SegmentedWALManager) Sync() error {
	wm.mu.Lock()
	defer wm.mu.Unlock()
	if wm.currentSegment == nil {
		return nil
	}
	if err := wm.currentSegment.Writer.Flush(); err != nil {
		return err
	}
	if wm.fsyncEnabled {
		return RealFsyncWithRetry(wm.currentSegment.File, 3, 100*time.Millisecond)
	}
	return nil
}

// ReadAll — читает все записи WAL.
func (wm *SegmentedWALManager) ReadAll() ([]*WALRecord, error) {
	wm.mu.RLock()
	segments := make([]*WALSegment, 0, len(wm.segments))
	for _, seg := range wm.segments {
		segments = append(segments, seg)
	}
	wm.mu.RUnlock()
	sort.Slice(segments, func(i, j int) bool {
		return segments[i].ID < segments[j].ID
	})
	records := make([]*WALRecord, 0)
	for _, seg := range segments {
		segRecords, err := wm.readSegmentRecords(seg)
		if err != nil {
			if wm.logger != nil {
				wm.logger.Warn(fmt.Sprintf("Failed to read segment %d: %v", seg.ID, err))
			}
			continue
		}
		records = append(records, segRecords...)
	}
	return records, nil
}

// ReadSince — читает записи после указанного LSN.
func (wm *SegmentedWALManager) ReadSince(lsn uint64) ([]*WALRecord, error) {
	allRecords, err := wm.ReadAll()
	if err != nil {
		return nil, err
	}
	result := make([]*WALRecord, 0)
	for _, record := range allRecords {
		if record.LSN > lsn {
			result = append(result, record)
		}
	}
	return result, nil
}

// GetCurrentLSN — текущий LSN.
func (wm *SegmentedWALManager) GetCurrentLSN() uint64 {
	wm.mu.RLock()
	defer wm.mu.RUnlock()
	if wm.currentSegment == nil {
		return 1
	}
	return wm.currentSegment.EndLSN
}

func (wm *SegmentedWALManager) readSegmentRecords(seg *WALSegment) ([]*WALRecord, error) {
	seg.mu.Lock()
	defer seg.mu.Unlock()
	if seg.File == nil {
		return nil, nil
	}
	seg.Writer.Flush()
	seg.File.Seek(0, 0)
	records := make([]*WALRecord, 0)
	reader := bufio.NewReader(seg.File)
	headerBuf := make([]byte, 8)
	for {
		_, err := io.ReadFull(reader, headerBuf)
		if err != nil {
			break
		}
		recordLen := binary.BigEndian.Uint32(headerBuf[0:4])
		expectedCRC := binary.BigEndian.Uint32(headerBuf[4:8])
		if recordLen == 0 || recordLen > 100*1024*1024 {
			break
		}
		recordData := make([]byte, recordLen)
		_, err = io.ReadFull(reader, recordData)
		if err != nil {
			break
		}
		var record WALRecord
		if err := json.Unmarshal(recordData, &record); err != nil {
			continue
		}
		lsnBytes := make([]byte, 8)
		binary.BigEndian.PutUint64(lsnBytes, record.LSN)
		crcData := append(lsnBytes, recordData...)
		calculatedCRC := crc32(crcData)
		if calculatedCRC != expectedCRC && calculatedCRC != record.CRC {
			continue
		}
		records = append(records, &record)
	}
	return records, nil
}

// Close — закрывает WAL.
func (wm *SegmentedWALManager) Close() error {
	close(wm.stopCh)
	wm.wg.Wait()
	close(wm.writeChan)

	wm.mu.Lock()
	defer wm.mu.Unlock()
	if wm.currentSegment != nil {
		wm.currentSegment.Writer.Flush()
		if wm.fsyncEnabled {
			RealFsyncWithRetry(wm.currentSegment.File, 3, 100*time.Millisecond)
		}
		wm.currentSegment.File.Close()
	}
	wm.index.save()
	return nil
}

func (im *WALIndexManager) addEntry(entry *WALIndexEntry) {
	im.mu.Lock()
	defer im.mu.Unlock()
	im.index[entry.LSN] = entry
}

func (im *WALIndexManager) save() error {
	im.mu.RLock()
	defer im.mu.RUnlock()
	data, err := json.Marshal(im.index)
	if err != nil {
		return err
	}
	return os.WriteFile(im.indexPath, data, 0644)
}

func (im *WALIndexManager) load() error {
	data, err := os.ReadFile(im.indexPath)
	if err != nil {
		if os.IsNotExist(err) {
			return nil
		}
		return err
	}
	if len(data) == 0 {
		return nil
	}
	return json.Unmarshal(data, &im.index)
}

// =============================================================================
// SAGA — ИНТЕРФЕЙСЫ ДЛЯ RAFT
// =============================================================================

// SagaRaftAccessor — минимальный интерфейс к Raft, необходимый
// оркестратору SAGA для выборов лидера.
//
// Реализуется в internal/cluster (SagaLeaderBridge), чтобы storage
// не зависел напрямую от hashicorp/raft.
type SagaRaftAccessor interface {
	// IsSagaLeader возвращает true, если текущий узел является лидером
	// в Raft-группе.
	IsSagaLeader() bool

	// GetSagaLeaderID возвращает ID текущего лидера SAGA (может быть
	// пустым, если лидер ещё не выбран).
	GetSagaLeaderID() string

	// GetSagaCurrentTerm возвращает текущий терм Raft.
	GetSagaCurrentTerm() uint64

	// RegisterSagaLeaderObserver регистрирует callback, который будет
	// вызван при смене лидера SAGA.
	RegisterSagaLeaderObserver(id string, cb func(isLeader bool, leaderID string))
}

// SagaLeaderState — персистентное состояние лидера SAGA.
type SagaLeaderState struct {
	LeaderID  string `json:"leader_id"`
	IsLeader  bool   `json:"is_leader"`
	Term      uint64 `json:"term"`
	UpdatedAt int64  `json:"updated_at"`
	NodeID    string `json:"node_id"`
}

// =============================================================================
// SAGA PERSISTENT STORAGE
// =============================================================================

// SagaState — состояние SAGA.
type SagaState struct {
	ID                   string                 `json:"id"`
	Status               string                 `json:"status"`
	CurrentStep          int                    `json:"current_step"`
	Steps                []SagaStepState        `json:"steps"`
	Data                 map[string]interface{} `json:"data"`
	CreatedAt            int64                  `json:"created_at"`
	UpdatedAt            int64                  `json:"updated_at"`
	CompletedAt          int64                  `json:"completed_at,omitempty"`
	CompensationExecuted bool                   `json:"compensation_executed"`
	NodeID               string                 `json:"node_id"`
	CoordinatorID        string                 `json:"coordinator_id"`
	Version              uint64                 `json:"version"`
	RetryCount           int                    `json:"retry_count"`
	LastError            string                 `json:"last_error,omitempty"`
	ExecutionID          string                 `json:"execution_id"`
}

// SagaStepState — состояние шага SAGA.
type SagaStepState struct {
	ID            string                 `json:"id"`
	Name          string                 `json:"name"`
	Status        string                 `json:"status"`
	Data          map[string]interface{} `json:"data"`
	StartedAt     int64                  `json:"started_at"`
	CompletedAt   int64                  `json:"completed_at"`
	ExecutionID   string                 `json:"execution_id"`
	RetryCount    int                    `json:"retry_count"`
	LastError     string                 `json:"last_error,omitempty"`
	CompensatedAt int64                  `json:"compensated_at,omitempty"`
}

// SagaPersistentStorage — персистентное хранилище SAGA.
//
// ДОБАВЛЕНО: executedKeys — персистентная таблица уже выполненных
// ExecutionID. Заполняется после успешного выполнения шага и после
// успешной компенсации. Проверяется перед выполнением/компенсацией.
type SagaPersistentStorage struct {
	baseDir         string
	mu              sync.RWMutex
	cache           map[string]*SagaState
	maxCache        int
	logger          LoggerInterface
	fsyncEnabled    bool
	fsyncMaxRetries int
	fsyncRetryDelay time.Duration

	executedMu   sync.RWMutex
	executedKeys map[string]int64
}

// NewSagaPersistentStorageWithConfig создаёт хранилище SAGA.
func NewSagaPersistentStorageWithConfig(cfg *config.SagaConfig, logger LoggerInterface) (*SagaPersistentStorage, error) {
	baseDir := SagaStateDir
	maxCache := 10000
	fsyncEnabled := true
	fsyncMaxRetries := 3
	fsyncRetryDelay := 100 * time.Millisecond

	if cfg != nil {
		baseDir = cfg.GetStateDir()
		maxCache = cfg.GetMaxCacheSize()
		fsyncEnabled = cfg.IsFsyncEnabled()
		fsyncMaxRetries = cfg.GetFsyncMaxRetries()
		fsyncRetryDelay = cfg.GetFsyncRetryDelay()
	}

	fullPath := filepath.Join(baseDir)
	if err := os.MkdirAll(fullPath, 0755); err != nil {
		return nil, fmt.Errorf("failed to create saga state directory: %v", err)
	}

	sps := &SagaPersistentStorage{
		baseDir:         fullPath,
		cache:           make(map[string]*SagaState),
		maxCache:        maxCache,
		logger:          logger,
		fsyncEnabled:    fsyncEnabled,
		fsyncMaxRetries: fsyncMaxRetries,
		fsyncRetryDelay: fsyncRetryDelay,
		executedKeys:    make(map[string]int64),
	}

	if err := sps.loadExecutedKeys(); err != nil {
		if logger != nil {
			logger.Warn(fmt.Sprintf("Failed to load saga executed keys: %v", err))
		}
	}

	return sps, nil
}

// getStatePath — путь к файлу состояния SAGA.
func (sps *SagaPersistentStorage) getStatePath(sagaID string) string {
	return filepath.Join(sps.baseDir, fmt.Sprintf("%s%s%s", SagaStateFilePrefix, sagaID, SagaStateFileSuffix))
}

// getExecutedKeysPath — путь к файлу таблицы выполненных ключей.
func (sps *SagaPersistentStorage) getExecutedKeysPath() string {
	return filepath.Join(sps.baseDir, SagaExecutedKeysFile)
}

// loadExecutedKeys загружает таблицу выполненных ключей с диска.
func (sps *SagaPersistentStorage) loadExecutedKeys() error {
	data, err := os.ReadFile(sps.getExecutedKeysPath())
	if err != nil {
		if os.IsNotExist(err) {
			return nil
		}
		return err
	}
	if len(data) == 0 {
		return nil
	}
	var loaded map[string]int64
	if err := json.Unmarshal(data, &loaded); err != nil {
		return err
	}
	sps.executedMu.Lock()
	sps.executedKeys = loaded
	sps.executedMu.Unlock()
	return nil
}

// saveExecutedKeys сохраняет таблицу выполненных ключей на диск.
func (sps *SagaPersistentStorage) saveExecutedKeys() error {
	sps.executedMu.RLock()
	data, err := json.Marshal(sps.executedKeys)
	sps.executedMu.RUnlock()
	if err != nil {
		return err
	}
	path := sps.getExecutedKeysPath()
	tmpPath := path + ".tmp"
	if err := os.WriteFile(tmpPath, data, 0644); err != nil {
		return err
	}
	if sps.fsyncEnabled {
		if f, err := os.OpenFile(tmpPath, os.O_RDWR, 0644); err == nil {
			RealFsyncWithRetry(f, sps.fsyncMaxRetries, sps.fsyncRetryDelay)
			f.Close()
		}
	}
	if err := os.Rename(tmpPath, path); err != nil {
		return err
	}
	if sps.fsyncEnabled {
		FsyncDir(sps.baseDir)
	}
	return nil
}

// IsExecuted проверяет, был ли ExecutionID уже выполнен.
func (sps *SagaPersistentStorage) IsExecuted(executionID string) bool {
	sps.executedMu.RLock()
	defer sps.executedMu.RUnlock()
	_, ok := sps.executedKeys[executionID]
	return ok
}

// MarkExecuted атомарно отмечает ExecutionID как выполненный
// и сохраняет таблицу на диск.
//
// Возвращает true, если ключ был добавлен впервые.
func (sps *SagaPersistentStorage) MarkExecuted(executionID string) bool {
	sps.executedMu.Lock()
	if _, exists := sps.executedKeys[executionID]; exists {
		sps.executedMu.Unlock()
		return false
	}
	sps.executedKeys[executionID] = time.Now().UnixMilli()
	sps.executedMu.Unlock()

	if err := sps.saveExecutedKeys(); err != nil {
		if sps.logger != nil {
			sps.logger.Warn(fmt.Sprintf("Failed to persist executed keys: %v", err))
		}
	}
	return true
}

// Save сохраняет состояние SAGA.
func (sps *SagaPersistentStorage) Save(state *SagaState) error {
	sps.mu.Lock()
	defer sps.mu.Unlock()

	state.UpdatedAt = time.Now().UnixMilli()
	state.Version++

	if len(sps.cache) >= sps.maxCache {
		var oldestKey string
		var oldestTime int64 = time.Now().UnixMilli()
		for k, v := range sps.cache {
			if v.UpdatedAt < oldestTime {
				oldestTime = v.UpdatedAt
				oldestKey = k
			}
		}
		if oldestKey != "" {
			delete(sps.cache, oldestKey)
		}
	}
	sps.cache[state.ID] = state

	path := sps.getStatePath(state.ID)
	data, err := json.MarshalIndent(state, "", "  ")
	if err != nil {
		return fmt.Errorf("failed to marshal saga state: %v", err)
	}

	tmpPath := path + ".tmp"
	if err := os.WriteFile(tmpPath, data, 0644); err != nil {
		return fmt.Errorf("failed to write saga state: %v", err)
	}

	if sps.fsyncEnabled {
		if f, err := os.OpenFile(tmpPath, os.O_RDWR, 0644); err == nil {
			RealFsyncWithRetry(f, sps.fsyncMaxRetries, sps.fsyncRetryDelay)
			f.Close()
		}
	}

	if err := os.Rename(tmpPath, path); err != nil {
		return fmt.Errorf("failed to rename saga state: %v", err)
	}

	if sps.fsyncEnabled {
		FsyncDir(sps.baseDir)
	}
	return nil
}

// Load загружает состояние SAGA.
func (sps *SagaPersistentStorage) Load(sagaID string) (*SagaState, error) {
	sps.mu.RLock()
	if state, ok := sps.cache[sagaID]; ok {
		sps.mu.RUnlock()
		return state, nil
	}
	sps.mu.RUnlock()

	path := sps.getStatePath(sagaID)
	data, err := os.ReadFile(path)
	if err != nil {
		if os.IsNotExist(err) {
			return nil, nil
		}
		return nil, fmt.Errorf("failed to read saga state: %v", err)
	}

	var state SagaState
	if err := json.Unmarshal(data, &state); err != nil {
		return nil, fmt.Errorf("failed to unmarshal saga state: %v", err)
	}

	sps.mu.Lock()
	sps.cache[sagaID] = &state
	sps.mu.Unlock()
	return &state, nil
}

// Delete удаляет состояние SAGA.
func (sps *SagaPersistentStorage) Delete(sagaID string) error {
	sps.mu.Lock()
	delete(sps.cache, sagaID)
	sps.mu.Unlock()

	path := sps.getStatePath(sagaID)
	if err := os.Remove(path); err != nil && !os.IsNotExist(err) {
		return fmt.Errorf("failed to delete saga state: %v", err)
	}
	return nil
}

// ListAll возвращает все состояния SAGA.
func (sps *SagaPersistentStorage) ListAll() ([]*SagaState, error) {
	pattern := filepath.Join(sps.baseDir, fmt.Sprintf("%s*%s", SagaStateFilePrefix, SagaStateFileSuffix))
	files, err := filepath.Glob(pattern)
	if err != nil {
		return nil, err
	}
	states := make([]*SagaState, 0, len(files))
	for _, file := range files {
		data, err := os.ReadFile(file)
		if err != nil {
			continue
		}
		var state SagaState
		if err := json.Unmarshal(data, &state); err != nil {
			continue
		}
		states = append(states, &state)
	}
	return states, nil
}

// ListPending возвращает незавершённые SAGA.
func (sps *SagaPersistentStorage) ListPending() ([]*SagaState, error) {
	all, err := sps.ListAll()
	if err != nil {
		return nil, err
	}
	pending := make([]*SagaState, 0)
	for _, state := range all {
		if state.Status == "pending" || state.Status == "running" || state.Status == "compensating" {
			pending = append(pending, state)
		}
	}
	return pending, nil
}

// SaveLeaderState сохраняет состояние лидера SAGA на диск.
func (sps *SagaPersistentStorage) SaveLeaderState(state *SagaLeaderState) error {
	path := filepath.Join(sps.baseDir, SagaLeaderStateFile)
	data, err := json.MarshalIndent(state, "", "  ")
	if err != nil {
		return err
	}
	tmpPath := path + ".tmp"
	if err := os.WriteFile(tmpPath, data, 0644); err != nil {
		return err
	}
	if sps.fsyncEnabled {
		if f, err := os.OpenFile(tmpPath, os.O_RDWR, 0644); err == nil {
			RealFsyncWithRetry(f, sps.fsyncMaxRetries, sps.fsyncRetryDelay)
			f.Close()
		}
	}
	if err := os.Rename(tmpPath, path); err != nil {
		return err
	}
	if sps.fsyncEnabled {
		FsyncDir(sps.baseDir)
	}
	return nil
}

// LoadLeaderState загружает состояние лидера SAGA с диска.
func (sps *SagaPersistentStorage) LoadLeaderState() (*SagaLeaderState, error) {
	path := filepath.Join(sps.baseDir, SagaLeaderStateFile)
	data, err := os.ReadFile(path)
	if err != nil {
		if os.IsNotExist(err) {
			return nil, nil
		}
		return nil, err
	}
	var state SagaLeaderState
	if err := json.Unmarshal(data, &state); err != nil {
		return nil, err
	}
	return &state, nil
}

// =============================================================================
// SAGA LEADER ELECTOR
// =============================================================================

// SagaLeaderElector — выборы лидера SAGA через Raft.
type SagaLeaderElector struct {
	nodeID      string
	raft        SagaRaftAccessor
	storage     *SagaPersistentStorage
	logger      LoggerInterface
	isLeader    atomic.Bool
	leaderID    atomic.Value
	currentTerm atomic.Uint64

	observersMu sync.RWMutex
	observers   map[string]func(isLeader bool, leaderID string)

	stopChan chan struct{}
	wg       sync.WaitGroup
}

// NewSagaLeaderElector создаёт выборщик лидера SAGA.
func NewSagaLeaderElector(nodeID string, raft SagaRaftAccessor, storage *SagaPersistentStorage, logger LoggerInterface) *SagaLeaderElector {
	e := &SagaLeaderElector{
		nodeID:    nodeID,
		raft:      raft,
		storage:   storage,
		logger:    logger,
		observers: make(map[string]func(isLeader bool, leaderID string)),
		stopChan:  make(chan struct{}),
	}
	e.leaderID.Store("")

	if storage != nil {
		if state, err := storage.LoadLeaderState(); err == nil && state != nil {
			e.isLeader.Store(state.IsLeader)
			e.currentTerm.Store(state.Term)
			e.leaderID.Store(state.LeaderID)
		}
	}

	if raft != nil {
		raft.RegisterSagaLeaderObserver(nodeID, func(isLeader bool, leaderID string) {
			e.updateLeadership(isLeader, leaderID)
		})
	}

	e.wg.Add(1)
	go e.leaderElectionLoop()

	return e
}

// updateLeadership — центральная точка обновления лидерства.
func (e *SagaLeaderElector) updateLeadership(isLeader bool, leaderID string) {
	oldLeader := e.isLeader.Load()
	oldLeaderID, _ := e.leaderID.Load().(string)

	e.isLeader.Store(isLeader)
	e.leaderID.Store(leaderID)

	term := uint64(0)
	if e.raft != nil {
		term = e.raft.GetSagaCurrentTerm()
	}
	e.currentTerm.Store(term)

	if e.storage != nil {
		_ = e.storage.SaveLeaderState(&SagaLeaderState{
			LeaderID:  leaderID,
			IsLeader:  isLeader,
			Term:      term,
			UpdatedAt: time.Now().UnixMilli(),
			NodeID:    e.nodeID,
		})
	}

	if oldLeader != isLeader || oldLeaderID != leaderID {
		e.observersMu.RLock()
		obs := make([]func(isLeader bool, leaderID string), 0, len(e.observers))
		for _, cb := range e.observers {
			obs = append(obs, cb)
		}
		e.observersMu.RUnlock()

		for _, cb := range obs {
			func(cb func(isLeader bool, leaderID string)) {
				defer func() {
					if r := recover(); r != nil && e.logger != nil {
						e.logger.Error(fmt.Sprintf("Saga leader observer panicked: %v", r))
					}
				}()
				cb(isLeader, leaderID)
			}(cb)
		}

		if e.logger != nil {
			e.logger.Info(fmt.Sprintf("Saga leadership changed: isLeader=%v, leaderID=%s, term=%d",
				isLeader, leaderID, term))
		}
	}
}

// leaderElectionLoop — фоновый опрос Raft на предмет лидерства.
func (e *SagaLeaderElector) leaderElectionLoop() {
	defer e.wg.Done()
	ticker := time.NewTicker(SagaLeaderElectionInterval)
	defer ticker.Stop()

	for {
		select {
		case <-ticker.C:
			if e.raft == nil {
				continue
			}
			isLeader := e.raft.IsSagaLeader()
			leaderID := e.raft.GetSagaLeaderID()
			e.updateLeadership(isLeader, leaderID)
		case <-e.stopChan:
			return
		}
	}
}

// IsLeader — является ли текущий узел лидером SAGA.
func (e *SagaLeaderElector) IsLeader() bool {
	return e.isLeader.Load()
}

// GetLeaderID — ID текущего лидера SAGA.
func (e *SagaLeaderElector) GetLeaderID() string {
	v, _ := e.leaderID.Load().(string)
	return v
}

// GetCurrentTerm — текущий терм Raft.
func (e *SagaLeaderElector) GetCurrentTerm() uint64 {
	return e.currentTerm.Load()
}

// RegisterObserver регистрирует наблюдателя за сменой лидера.
func (e *SagaLeaderElector) RegisterObserver(id string, cb func(isLeader bool, leaderID string)) {
	e.observersMu.Lock()
	defer e.observersMu.Unlock()
	e.observers[id] = cb
}

// UnregisterObserver снимает регистрацию наблюдателя.
func (e *SagaLeaderElector) UnregisterObserver(id string) {
	e.observersMu.Lock()
	defer e.observersMu.Unlock()
	delete(e.observers, id)
}

// Stop останавливает выборщик.
func (e *SagaLeaderElector) Stop() {
	close(e.stopChan)
	e.wg.Wait()
}

// =============================================================================
// SAGA ORCHESTRATOR
// =============================================================================

// SagaOrchestrator управляет SAGA.
type SagaOrchestrator struct {
	storage       *SagaPersistentStorage
	mu            sync.RWMutex
	logger        LoggerInterface
	nodeID        string
	stopChan      chan struct{}
	wg            sync.WaitGroup
	isLeader      atomic.Bool
	leaderID      string
	config        *config.SagaConfig
	metrics       *SagaMetrics
	activeSagas   sync.Map
	leaderElector *SagaLeaderElector
	raftAccessor  SagaRaftAccessor
}

// SagaMetrics — метрики SAGA.
type SagaMetrics struct {
	TotalStarted     atomic.Uint64
	TotalCompleted   atomic.Uint64
	TotalAborted     atomic.Uint64
	TotalFailed      atomic.Uint64
	TotalCompensated atomic.Uint64
	ActiveCount      atomic.Int64
	RecoveryCount    atomic.Uint64
	IdempotentSkips  atomic.Uint64
	NotLeaderRejects atomic.Uint64
}

// NewSagaOrchestratorWithConfig создаёт оркестратор без Raft.
func NewSagaOrchestratorWithConfig(cfg *config.SagaConfig, nodeID string, logger LoggerInterface) (*SagaOrchestrator, error) {
	return NewSagaOrchestratorWithRaft(cfg, nodeID, nil, logger)
}

// NewSagaOrchestratorWithRaft создаёт оркестратор с привязкой к Raft.
func NewSagaOrchestratorWithRaft(cfg *config.SagaConfig, nodeID string, raftAccessor SagaRaftAccessor, logger LoggerInterface) (*SagaOrchestrator, error) {
	if cfg == nil {
		cfg = &config.SagaConfig{
			Enabled:               true,
			StateDir:              "saga_states",
			MaxRetries:            5,
			RetryBackoffMs:        100,
			SagaTimeoutSec:        300,
			StuckCheckIntervalSec: 10,
			RecoveryIntervalSec:   30,
			CleanupPeriodHours:    24,
			MaxCacheSize:          10000,
			FsyncEnabled:          true,
			FsyncMaxRetries:       3,
			FsyncRetryDelayMs:     100,
		}
	}

	storage, err := NewSagaPersistentStorageWithConfig(cfg, logger)
	if err != nil {
		return nil, err
	}

	o := &SagaOrchestrator{
		storage:      storage,
		logger:       logger,
		nodeID:       nodeID,
		stopChan:     make(chan struct{}),
		metrics:      &SagaMetrics{},
		config:       cfg,
		raftAccessor: raftAccessor,
	}

	o.leaderElector = NewSagaLeaderElector(nodeID, raftAccessor, storage, logger)

	if raftAccessor == nil {
		o.isLeader.Store(true)
		o.leaderID = nodeID
		o.leaderElector.updateLeadership(true, nodeID)
	} else {
		o.isLeader.Store(o.leaderElector.IsLeader())
		o.leaderID = o.leaderElector.GetLeaderID()
	}

	o.leaderElector.RegisterObserver("orchestrator", func(isLeader bool, leaderID string) {
		o.mu.Lock()
		o.leaderID = leaderID
		o.mu.Unlock()
	})

	o.wg.Add(1)
	go o.recoveryLoop()
	o.wg.Add(1)
	go o.orphanCleanupLoop()

	return o, nil
}

// IsLeader — является ли узел лидером.
func (o *SagaOrchestrator) IsLeader() bool {
	if o.leaderElector == nil {
		return true
	}
	return o.leaderElector.IsLeader()
}

// GetLeaderID — ID лидера.
func (o *SagaOrchestrator) GetLeaderID() string {
	if o.leaderElector == nil {
		return o.nodeID
	}
	if id := o.leaderElector.GetLeaderID(); id != "" {
		return id
	}
	o.mu.RLock()
	defer o.mu.RUnlock()
	return o.leaderID
}

// IsExecuted проверяет, был ли ExecutionID уже выполнен.
func (o *SagaOrchestrator) IsExecuted(executionID string) bool {
	if o.storage == nil {
		return false
	}
	return o.storage.IsExecuted(executionID)
}

// MarkExecuted помечает ExecutionID как выполненный.
func (o *SagaOrchestrator) MarkExecuted(executionID string) bool {
	if o.storage == nil {
		return true
	}
	first := o.storage.MarkExecuted(executionID)
	if !first {
		o.metrics.IdempotentSkips.Add(1)
	}
	return first
}

func (o *SagaOrchestrator) recoveryLoop() {
	defer o.wg.Done()
	ticker := time.NewTicker(o.config.GetRecoveryInterval())
	defer ticker.Stop()
	for {
		select {
		case <-ticker.C:
			if !o.IsLeader() {
				continue
			}
			o.recoverPendingSagas()
		case <-o.stopChan:
			return
		}
	}
}

func (o *SagaOrchestrator) orphanCleanupLoop() {
	defer o.wg.Done()
	ticker := time.NewTicker(SagaCleanupInterval)
	defer ticker.Stop()
	for {
		select {
		case <-ticker.C:
			if !o.IsLeader() {
				continue
			}
			o.cleanupOrphanSteps()
		case <-o.stopChan:
			return
		}
	}
}

// cleanupOrphanSteps — очистка «осиротевших» SAGA.
func (o *SagaOrchestrator) cleanupOrphanSteps() {
	pending, err := o.storage.ListPending()
	if err != nil {
		return
	}
	now := time.Now().UnixMilli()
	for _, state := range pending {
		if state.Status == "running" || state.Status == "compensating" {
			elapsed := now - state.UpdatedAt
			if elapsed > SagaOrphanTimeout.Milliseconds() {
				if o.logger != nil {
					o.logger.Warn(fmt.Sprintf("Found orphan saga %s, cleaning up", state.ID))
				}
				state.Status = "orphaned"
				state.LastError = fmt.Sprintf("orphaned after %d ms", elapsed)
				o.storage.Save(state)
				o.compensateOrphanSaga(state)
			}
		}
	}
}

// compensateOrphanSaga — компенсация «осиротевшей» SAGA.
func (o *SagaOrchestrator) compensateOrphanSaga(state *SagaState) {
	if !o.IsLeader() {
		return
	}
	if state.CompensationExecuted {
		return
	}
	if o.logger != nil {
		o.logger.Info(fmt.Sprintf("Compensating orphan saga %s", state.ID))
	}

	saga := o.stateToTransaction(state)
	if err := o.executeCompensation(saga, len(saga.Steps)-1); err != nil {
		if o.logger != nil {
			o.logger.Error(fmt.Sprintf("Failed to compensate orphan saga %s: %v", state.ID, err))
		}
		state.Status = "compensation_failed"
		state.LastError = err.Error()
		o.storage.Save(state)
		return
	}
	state.CompensationExecuted = true
	state.Status = "orphaned_compensated"
	o.storage.Save(state)
}

// stateToTransaction конвертирует SagaState в SagaTransaction.
func (o *SagaOrchestrator) stateToTransaction(state *SagaState) *SagaTransaction {
	saga := &SagaTransaction{
		ID:                   state.ID,
		Status:               state.Status,
		CurrentStep:          state.CurrentStep,
		Data:                 state.Data,
		CreatedAt:            state.CreatedAt,
		UpdatedAt:            state.UpdatedAt,
		CompletedAt:          state.CompletedAt,
		compensationExecuted: state.CompensationExecuted,
		executedOps:          make(map[string]bool),
	}
	for _, stepState := range state.Steps {
		step := &SagaStep{
			ID:          stepState.ID,
			Name:        stepState.Name,
			Status:      stepState.Status,
			Data:        stepState.Data,
			StartedAt:   stepState.StartedAt,
			CompletedAt: stepState.CompletedAt,
			ExecutionID: stepState.ExecutionID,
			RetryCount:  stepState.RetryCount,
			LastError:   stepState.LastError,
		}
		saga.Steps = append(saga.Steps, step)
		if stepState.Status == "completed" {
			saga.executedOps[stepState.ExecutionID] = true
		}
	}
	return saga
}

// recoverPendingSagas — восстановление незавершённых SAGA.
func (o *SagaOrchestrator) recoverPendingSagas() {
	if !o.IsLeader() {
		return
	}
	pending, err := o.storage.ListPending()
	if err != nil {
		return
	}
	if len(pending) == 0 {
		return
	}
	if o.logger != nil {
		o.logger.Info(fmt.Sprintf("Recovering %d pending sagas", len(pending)))
	}
	for _, state := range pending {
		if err := o.resumeSaga(state); err != nil {
			if o.logger != nil {
				o.logger.Error(fmt.Sprintf("Failed to resume saga %s: %v", state.ID, err))
			}
		} else {
			o.metrics.RecoveryCount.Add(1)
		}
	}
}

// resumeSaga — возобновление SAGA.
func (o *SagaOrchestrator) resumeSaga(state *SagaState) error {
	saga := o.stateToTransaction(state)
	o.activeSagas.Store(state.ID, saga)
	return o.executeSagaInternal(saga)
}

// BeginSaga — начать SAGA.
func (o *SagaOrchestrator) BeginSaga(id string) (*SagaTransaction, error) {
	if !o.IsLeader() {
		o.metrics.NotLeaderRejects.Add(1)
		return nil, fmt.Errorf("node is not the saga leader (leader: %s)", o.GetLeaderID())
	}

	existing, err := o.storage.Load(id)
	if err != nil {
		return nil, err
	}
	if existing != nil {
		return nil, fmt.Errorf("saga %s already exists", id)
	}
	now := time.Now().UnixMilli()
	state := &SagaState{
		ID:          id,
		Status:      "pending",
		CurrentStep: 0,
		Data:        make(map[string]interface{}),
		CreatedAt:   now,
		UpdatedAt:   now,
		NodeID:      o.nodeID,
		Version:     1,
		ExecutionID: fmt.Sprintf("%s_%d", id, now),
	}
	if err := o.storage.Save(state); err != nil {
		return nil, err
	}
	saga := &SagaTransaction{
		ID:          id,
		Steps:       make([]*SagaStep, 0),
		CurrentStep: 0,
		Status:      "pending",
		CreatedAt:   now,
		UpdatedAt:   now,
		Data:        make(map[string]interface{}),
		executedOps: make(map[string]bool),
	}
	o.activeSagas.Store(id, saga)
	o.metrics.TotalStarted.Add(1)
	o.metrics.ActiveCount.Add(1)
	return saga, nil
}

// Execute — выполнить SAGA.
func (o *SagaOrchestrator) Execute(saga *SagaTransaction) error {
	if !o.IsLeader() {
		o.metrics.NotLeaderRejects.Add(1)
		return fmt.Errorf("node is not the saga leader (leader: %s)", o.GetLeaderID())
	}
	return o.executeSagaInternal(saga)
}

// executeSagaInternal — внутренняя логика выполнения SAGA.
func (o *SagaOrchestrator) executeSagaInternal(saga *SagaTransaction) error {
	if !o.IsLeader() {
		return fmt.Errorf("node is not the saga leader")
	}

	saga.mu.Lock()
	if saga.Status != "pending" && saga.Status != "running" {
		saga.mu.Unlock()
		return fmt.Errorf("saga %s is not in pending or running state", saga.ID)
	}
	saga.Status = "running"
	saga.UpdatedAt = time.Now().UnixMilli()
	saga.mu.Unlock()

	if err := o.saveSagaState(saga); err != nil {
		return err
	}

	for i := saga.CurrentStep; i < len(saga.Steps); i++ {
		if !o.IsLeader() {
			return fmt.Errorf("lost saga leadership during execution of %s", saga.ID)
		}

		saga.mu.Lock()
		step := saga.Steps[i]
		saga.CurrentStep = i
		if step.Status == "completed" {
			saga.mu.Unlock()
			continue
		}
		step.Status = "running"
		step.StartedAt = time.Now().UnixMilli()
		saga.mu.Unlock()

		if o.IsExecuted(step.ExecutionID) || saga.HasExecuted(step.ExecutionID) {
			saga.mu.Lock()
			step.Status = "completed"
			step.CompletedAt = time.Now().UnixMilli()
			saga.mu.Unlock()
			if o.logger != nil {
				o.logger.Debug(fmt.Sprintf("Saga step %s (execution_id=%s) already executed, skipping",
					step.Name, step.ExecutionID))
			}
			continue
		}

		var err error
		maxRetries := SagaMaxRetries
		if o.config != nil {
			maxRetries = o.config.GetMaxRetries()
		}
		for retry := 0; retry < maxRetries; retry++ {
			saga.mu.Lock()
			step.RetryCount = retry + 1
			saga.mu.Unlock()
			if err = step.Execute(); err == nil {
				break
			}
			saga.mu.Lock()
			step.LastError = err.Error()
			saga.mu.Unlock()
			backoff := time.Duration(o.config.RetryBackoffMs*(1<<retry)) * time.Millisecond
			time.Sleep(backoff)
		}
		saga.mu.Lock()
		step.CompletedAt = time.Now().UnixMilli()
		saga.mu.Unlock()

		if err != nil {
			saga.mu.Lock()
			step.Status = "failed"
			saga.Status = "compensating"
			saga.UpdatedAt = time.Now().UnixMilli()
			saga.mu.Unlock()
			o.saveSagaState(saga)

			compensationErr := o.executeCompensation(saga, i)
			if compensationErr != nil {
				saga.mu.Lock()
				saga.Status = "compensation_failed"
				saga.UpdatedAt = time.Now().UnixMilli()
				saga.mu.Unlock()
				o.saveSagaState(saga)
				o.metrics.TotalFailed.Add(1)
				o.metrics.ActiveCount.Add(-1)
				return fmt.Errorf("saga %s compensation failed: %v", saga.ID, compensationErr)
			}
			saga.mu.Lock()
			saga.Status = "aborted"
			saga.UpdatedAt = time.Now().UnixMilli()
			saga.mu.Unlock()
			o.saveSagaState(saga)
			o.metrics.TotalAborted.Add(1)
			o.metrics.ActiveCount.Add(-1)
			return fmt.Errorf("saga %s aborted at step %s: %v", saga.ID, step.Name, err)
		}

		o.MarkExecuted(step.ExecutionID)

		saga.mu.Lock()
		step.Status = "completed"
		saga.MarkExecuted(step.ExecutionID)
		saga.UpdatedAt = time.Now().UnixMilli()
		saga.mu.Unlock()
		o.saveSagaState(saga)
	}

	saga.mu.Lock()
	saga.Status = "completed"
	saga.CompletedAt = time.Now().UnixMilli()
	saga.UpdatedAt = saga.CompletedAt
	saga.mu.Unlock()
	o.saveSagaState(saga)
	o.metrics.TotalCompleted.Add(1)
	o.metrics.ActiveCount.Add(-1)
	return nil
}

// saveSagaState сохраняет состояние SAGA.
func (o *SagaOrchestrator) saveSagaState(saga *SagaTransaction) error {
	saga.mu.RLock()
	state := &SagaState{
		ID:                   saga.ID,
		Status:               saga.Status,
		CurrentStep:          saga.CurrentStep,
		Data:                 saga.Data,
		CreatedAt:            saga.CreatedAt,
		UpdatedAt:            saga.UpdatedAt,
		CompletedAt:          saga.CompletedAt,
		CompensationExecuted: saga.compensationExecuted,
		NodeID:               o.nodeID,
		ExecutionID:          fmt.Sprintf("%s_%d", saga.ID, saga.CreatedAt),
	}
	state.Steps = make([]SagaStepState, len(saga.Steps))
	for i, step := range saga.Steps {
		state.Steps[i] = SagaStepState{
			ID:          step.ID,
			Name:        step.Name,
			Status:      step.Status,
			Data:        step.Data,
			StartedAt:   step.StartedAt,
			CompletedAt: step.CompletedAt,
			ExecutionID: step.ExecutionID,
			RetryCount:  step.RetryCount,
			LastError:   step.LastError,
		}
	}
	saga.mu.RUnlock()
	return o.storage.Save(state)
}

// executeCompensation — выполнение компенсации.
func (o *SagaOrchestrator) executeCompensation(saga *SagaTransaction, failedStep int) error {
	if !o.IsLeader() {
		return fmt.Errorf("node is not the saga leader")
	}

	saga.mu.Lock()
	if saga.compensationExecuted {
		saga.mu.Unlock()
		return nil
	}
	saga.mu.Unlock()

	for j := failedStep; j >= 0; j-- {
		if !o.IsLeader() {
			return fmt.Errorf("lost saga leadership during compensation of %s", saga.ID)
		}

		saga.mu.RLock()
		step := saga.Steps[j]
		status := step.Status
		saga.mu.RUnlock()

		if status == "compensated" || status == "pending" {
			continue
		}
		if status != "completed" && status != "failed" {
			continue
		}

		compExecID := step.ExecutionID + ":compensate"
		if o.IsExecuted(compExecID) {
			saga.mu.Lock()
			step.Status = "compensated"
			step.CompletedAt = time.Now().UnixMilli()
			saga.mu.Unlock()
			o.saveSagaState(saga)
			continue
		}

		if step.Compensate == nil {
			saga.mu.Lock()
			step.Status = "compensated"
			step.CompletedAt = time.Now().UnixMilli()
			saga.mu.Unlock()
			o.MarkExecuted(compExecID)
			o.saveSagaState(saga)
			continue
		}

		var err error
		maxRetries := SagaMaxRetries
		if o.config != nil {
			maxRetries = o.config.GetMaxRetries()
		}
		for retry := 0; retry < maxRetries; retry++ {
			if err = step.Compensate(); err == nil {
				break
			}
			backoff := time.Duration(o.config.RetryBackoffMs*(1<<retry)) * time.Millisecond
			time.Sleep(backoff)
		}
		if err != nil {
			saga.mu.Lock()
			step.Status = "compensation_failed"
			step.LastError = err.Error()
			saga.mu.Unlock()
			o.saveSagaState(saga)
			return fmt.Errorf("compensation for step %s failed: %v", step.Name, err)
		}
		o.MarkExecuted(compExecID)
		saga.mu.Lock()
		step.Status = "compensated"
		step.CompletedAt = time.Now().UnixMilli()
		saga.mu.Unlock()
		o.saveSagaState(saga)
	}

	saga.mu.Lock()
	saga.compensationExecuted = true
	saga.mu.Unlock()
	o.metrics.TotalCompensated.Add(1)
	return nil
}

// GetSaga — получить SAGA.
func (o *SagaOrchestrator) GetSaga(id string) (*SagaTransaction, error) {
	if val, ok := o.activeSagas.Load(id); ok {
		return val.(*SagaTransaction), nil
	}
	state, err := o.storage.Load(id)
	if err != nil {
		return nil, err
	}
	if state == nil {
		return nil, fmt.Errorf("saga %s not found", id)
	}
	return o.stateToTransaction(state), nil
}

// GetMetrics — метрики SAGA.
func (o *SagaOrchestrator) GetMetrics() map[string]interface{} {
	return map[string]interface{}{
		"total_started":      o.metrics.TotalStarted.Load(),
		"total_completed":    o.metrics.TotalCompleted.Load(),
		"total_aborted":      o.metrics.TotalAborted.Load(),
		"total_failed":       o.metrics.TotalFailed.Load(),
		"total_compensated":  o.metrics.TotalCompensated.Load(),
		"active_count":       o.metrics.ActiveCount.Load(),
		"recovery_count":     o.metrics.RecoveryCount.Load(),
		"idempotent_skips":   o.metrics.IdempotentSkips.Load(),
		"not_leader_rejects": o.metrics.NotLeaderRejects.Load(),
		"is_leader":          o.IsLeader(),
		"leader_id":          o.GetLeaderID(),
		"node_id":            o.nodeID,
	}
}

// Stop — остановка оркестратора.
func (o *SagaOrchestrator) Stop() {
	close(o.stopChan)
	if o.leaderElector != nil {
		o.leaderElector.Stop()
	}
	o.wg.Wait()
	if o.logger != nil {
		o.logger.Info("Saga orchestrator stopped")
	}
}

// =============================================================================
// SAGA TRANSACTION
// =============================================================================

// SagaStep — шаг SAGA.
type SagaStep struct {
	ID          string                 `json:"id"`
	Name        string                 `json:"name"`
	Execute     func() error           `json:"-"`
	Compensate  func() error           `json:"-"`
	Status      string                 `json:"status"`
	Data        map[string]interface{} `json:"data"`
	StartedAt   int64                  `json:"started_at"`
	CompletedAt int64                  `json:"completed_at"`
	ExecutionID string                 `json:"execution_id"`
	RetryCount  int                    `json:"retry_count"`
	LastError   string                 `json:"last_error,omitempty"`
}

// SagaTransaction — SAGA-транзакция.
type SagaTransaction struct {
	ID                   string                 `json:"id"`
	Steps                []*SagaStep            `json:"steps"`
	CurrentStep          int                    `json:"current_step"`
	Status               string                 `json:"status"`
	CreatedAt            int64                  `json:"created_at"`
	UpdatedAt            int64                  `json:"updated_at"`
	CompletedAt          int64                  `json:"completed_at,omitempty"`
	Data                 map[string]interface{} `json:"data"`
	mu                   sync.RWMutex
	executedOps          map[string]bool `json:"-"`
	compensationExecuted bool            `json:"-"`
}

// AddStep добавляет шаг.
func (s *SagaTransaction) AddStep(name string, execute, compensate func() error, data map[string]interface{}) *SagaTransaction {
	s.mu.Lock()
	defer s.mu.Unlock()
	stepID := fmt.Sprintf("%s_step_%d_%d", s.ID, len(s.Steps), time.Now().UnixNano())
	step := &SagaStep{
		ID:          stepID,
		Name:        name,
		Execute:     execute,
		Compensate:  compensate,
		Status:      "pending",
		Data:        data,
		StartedAt:   time.Now().UnixMilli(),
		ExecutionID: stepID,
		RetryCount:  0,
	}
	s.Steps = append(s.Steps, step)
	return s
}

// SetData устанавливает данные.
func (s *SagaTransaction) SetData(key string, value interface{}) {
	s.mu.Lock()
	defer s.mu.Unlock()
	s.Data[key] = value
}

// GetData получает данные.
func (s *SagaTransaction) GetData(key string) (interface{}, bool) {
	s.mu.RLock()
	defer s.mu.RUnlock()
	val, ok := s.Data[key]
	return val, ok
}

// HasExecuted — был ли шаг выполнен.
func (s *SagaTransaction) HasExecuted(operationID string) bool {
	s.mu.RLock()
	defer s.mu.RUnlock()
	_, ok := s.executedOps[operationID]
	return ok
}

// MarkExecuted — отметить шаг как выполненный.
func (s *SagaTransaction) MarkExecuted(operationID string) {
	s.executedOps[operationID] = true
}

// =============================================================================
// SAGA MANAGER (обёртка)
// =============================================================================

// SagaManager — обёртка над оркестратором.
type SagaManager struct {
	orchestrator *SagaOrchestrator
	logger       LoggerInterface
	stopChan     chan struct{}
	wg           sync.WaitGroup
	maxRetries   int
}

// NewSagaManager создаёт SAGA-менеджер (single-node, без Raft).
func NewSagaManager(logger LoggerInterface) *SagaManager {
	orchestrator, err := NewSagaOrchestratorWithConfig(nil, "default-node", logger)
	if err != nil {
		if logger != nil {
			logger.Error(fmt.Sprintf("Failed to create saga orchestrator: %v", err))
		}
		return &SagaManager{
			logger:     logger,
			stopChan:   make(chan struct{}),
			maxRetries: 3,
		}
	}
	return &SagaManager{
		orchestrator: orchestrator,
		logger:       logger,
		stopChan:     make(chan struct{}),
		maxRetries:   3,
	}
}

// NewSagaManagerWithRaft создаёт SAGA-менеджер с привязкой к Raft.
func NewSagaManagerWithRaft(nodeID string, raftAccessor SagaRaftAccessor, logger LoggerInterface) *SagaManager {
	orchestrator, err := NewSagaOrchestratorWithRaft(nil, nodeID, raftAccessor, logger)
	if err != nil {
		if logger != nil {
			logger.Error(fmt.Sprintf("Failed to create saga orchestrator with raft: %v", err))
		}
		return &SagaManager{
			logger:     logger,
			stopChan:   make(chan struct{}),
			maxRetries: 3,
		}
	}
	return &SagaManager{
		orchestrator: orchestrator,
		logger:       logger,
		stopChan:     make(chan struct{}),
		maxRetries:   3,
	}
}

// BeginSaga — начать SAGA.
func (sm *SagaManager) BeginSaga(id string) *SagaTransaction {
	if sm.orchestrator != nil {
		saga, err := sm.orchestrator.BeginSaga(id)
		if err != nil {
			if sm.logger != nil {
				sm.logger.Error(fmt.Sprintf("Failed to begin saga: %v", err))
			}
			return &SagaTransaction{
				ID:          id,
				Status:      "pending",
				CreatedAt:   time.Now().UnixMilli(),
				UpdatedAt:   time.Now().UnixMilli(),
				Data:        make(map[string]interface{}),
				executedOps: make(map[string]bool),
			}
		}
		return saga
	}
	return &SagaTransaction{
		ID:          id,
		Steps:       make([]*SagaStep, 0),
		CurrentStep: 0,
		Status:      "pending",
		CreatedAt:   time.Now().UnixMilli(),
		UpdatedAt:   time.Now().UnixMilli(),
		Data:        make(map[string]interface{}),
		executedOps: make(map[string]bool),
	}
}

// Execute — выполнить SAGA.
func (sm *SagaManager) Execute(saga *SagaTransaction) error {
	if sm.orchestrator != nil {
		return sm.orchestrator.Execute(saga)
	}
	return fmt.Errorf("orchestrator not available")
}

// GetSaga — получить SAGA.
func (sm *SagaManager) GetSaga(id string) (*SagaTransaction, error) {
	if sm.orchestrator != nil {
		return sm.orchestrator.GetSaga(id)
	}
	return nil, fmt.Errorf("saga %s not found", id)
}

// GetSagaStatus — статус SAGA.
func (sm *SagaManager) GetSagaStatus(id string) (string, error) {
	saga, err := sm.GetSaga(id)
	if err != nil {
		return "", err
	}
	saga.mu.RLock()
	defer saga.mu.RUnlock()
	return saga.Status, nil
}

// GetOrchestrator — получить оркестратор.
func (sm *SagaManager) GetOrchestrator() *SagaOrchestrator {
	return sm.orchestrator
}

// GetMetrics — метрики SAGA.
func (sm *SagaManager) GetMetrics() map[string]interface{} {
	if sm.orchestrator != nil {
		return sm.orchestrator.GetMetrics()
	}
	return map[string]interface{}{
		"total_started":   0,
		"total_completed": 0,
		"total_aborted":   0,
		"total_failed":    0,
		"active_count":    0,
		"is_leader":       false,
		"leader_id":       "",
		"node_id":         "",
	}
}

// GetActiveSagas — активные SAGA.
func (sm *SagaManager) GetActiveSagas() []*SagaTransaction {
	result := make([]*SagaTransaction, 0)
	if sm.orchestrator == nil {
		return result
	}
	sm.orchestrator.activeSagas.Range(func(key, value interface{}) bool {
		saga := value.(*SagaTransaction)
		saga.mu.RLock()
		status := saga.Status
		saga.mu.RUnlock()
		if status == "pending" || status == "running" {
			result = append(result, saga)
		}
		return true
	})
	return result
}

// GetAllSagas возвращает все SAGA транзакции — и активные, и завершённые.
//
// ДОБАВЛЕНО (2026-10): используется REPL-командой "saga list --all".
// В отличие от GetActiveSagas, читает состояние с диска через
// SagaPersistentStorage.ListAll() и конвертирует каждую запись в
// SagaTransaction через stateToTransaction.
func (sm *SagaManager) GetAllSagas() []*SagaTransaction {
	result := make([]*SagaTransaction, 0)
	if sm.orchestrator == nil || sm.orchestrator.storage == nil {
		return result
	}
	states, err := sm.orchestrator.storage.ListAll()
	if err != nil {
		return result
	}
	for _, st := range states {
		if st == nil {
			continue
		}
		result = append(result, sm.orchestrator.stateToTransaction(st))
	}
	return result
}

// Stop — остановка.
func (sm *SagaManager) Stop() {
	close(sm.stopChan)
	sm.wg.Wait()
	if sm.orchestrator != nil {
		sm.orchestrator.Stop()
	}
}

// =============================================================================
// ГЛОБАЛЬНЫЕ SAGA
// =============================================================================

var globalSagaOrchestrator *SagaOrchestrator
var sagaOrchestratorMu sync.RWMutex

// SetGlobalSagaOrchestrator — установить глобальный оркестратор.
func SetGlobalSagaOrchestrator(o *SagaOrchestrator) {
	sagaOrchestratorMu.Lock()
	defer sagaOrchestratorMu.Unlock()
	globalSagaOrchestrator = o
}

// GetGlobalSagaOrchestrator — получить глобальный оркестратор.
func GetGlobalSagaOrchestrator() *SagaOrchestrator {
	sagaOrchestratorMu.RLock()
	defer sagaOrchestratorMu.RUnlock()
	return globalSagaOrchestrator
}

// =============================================================================
// СТАТИСТИКА
// =============================================================================

// GetBatchStats возвращает статистику batch-операций.
func GetBatchStats() map[string]interface{} {
	if globalBatchManager == nil {
		return map[string]interface{}{"error": "batch manager not initialized"}
	}
	return map[string]interface{}{
		"total_batches":    globalBatchManager.stats.TotalBatches.Load(),
		"total_operations": globalBatchManager.stats.TotalOperations.Load(),
		"total_failed":     globalBatchManager.stats.TotalFailed.Load(),
		"active_batches":   globalBatchManager.GetActiveBatchCount(),
		"uptime_seconds":   time.Since(globalBatchManager.stats.StartTime).Seconds(),
	}
}

// StopBatchManager останавливает batch-менеджер.
func StopBatchManager() error {
	if globalBatchManager == nil {
		return nil
	}
	if globalBatchManager.wal != nil {
		globalBatchManager.wal.Sync()
		globalBatchManager.wal.Close()
	}
	return nil
}

// =============================================================================
// СОВМЕСТИМОСТЬ: псевдонимы для старого API
// =============================================================================

// GetTransactionStats — псевдоним для GetBatchStats.
func GetTransactionStats() map[string]interface{} {
	return GetBatchStats()
}