302 lines
7.1 KiB
Markdown
302 lines
7.1 KiB
Markdown
# Part 6: 性能与可靠性
|
||
|
||
本文档详细说明 Ops 模块的性能优化策略和可靠性保障措施。
|
||
|
||
---
|
||
|
||
## 6.1 性能优化
|
||
|
||
### 6.1.1 Redis 缓存策略
|
||
|
||
**热点数据缓存**：
|
||
|
||
```java
|
||
@Service
|
||
public class CleanerStatusService {
|
||
|
||
@Cacheable(value = "cleaner:status", key = "#userId", unless = "#result == null")
|
||
public CleanerStatusDO getStatus(Long userId) {
|
||
return cleanerStatusMapper.selectByUserId(userId);
|
||
}
|
||
|
||
@CacheEvict(value = "cleaner:status", key = "#userId")
|
||
public void updateStatus(Long userId, CleanerStatusEnum status) {
|
||
cleanerStatusMapper.updateStatus(userId, status);
|
||
}
|
||
}
|
||
```
|
||
|
||
**缓存更新策略**：
|
||
- **Cache Aside**: 先更新数据库，再删除缓存
|
||
- **过期时间**: 5-30 分钟，避免数据过期
|
||
- **空值缓存**: 防止缓存穿透
|
||
|
||
### 6.1.2 数据库优化
|
||
|
||
**索引设计**：
|
||
|
||
```sql
|
||
-- 工单表索引
|
||
CREATE INDEX idx_status ON ops_order(status);
|
||
CREATE INDEX idx_assignee_id ON ops_order(assignee_id);
|
||
CREATE INDEX idx_priority_status ON ops_order(priority, status);
|
||
CREATE INDEX idx_create_time ON ops_order(create_time);
|
||
|
||
-- 队列表索引
|
||
CREATE UNIQUE INDEX uk_user_order ON ops_order_queue(user_id, ops_order_id);
|
||
CREATE INDEX idx_status ON ops_order_queue(status);
|
||
CREATE INDEX idx_priority ON ops_order_queue(priority);
|
||
```
|
||
|
||
**查询优化**：
|
||
|
||
```java
|
||
// ❌ N+1 查询
|
||
List<OpsOrderDO> orders = orderMapper.selectList(wrapper);
|
||
for (OpsOrderDO order : orders) {
|
||
UserDO user = userMapper.selectById(order.getAssigneeId()); // N次查询
|
||
}
|
||
|
||
// ✅ 批量查询
|
||
List<OpsOrderDO> orders = orderMapper.selectList(wrapper);
|
||
Set<Long> userIds = orders.stream().map(OpsOrderDO::getAssigneeId).collect(Collectors.toSet());
|
||
List<UserDO> users = userMapper.selectBatchIds(userIds); // 1次查询
|
||
Map<Long, UserDO> userMap = users.stream().collect(Collectors.toMap(UserDO::getId, u -> u));
|
||
```
|
||
|
||
### 6.1.3 异步处理
|
||
|
||
**事件异步发布**：
|
||
|
||
```java
|
||
@Service
|
||
public class OrderEventPublisherImpl {
|
||
|
||
@Async("eventExecutor")
|
||
public void publishStateChanged(OrderStateChangedEvent event) {
|
||
applicationEventPublisher.publishEvent(event);
|
||
}
|
||
}
|
||
|
||
@Configuration
|
||
public class AsyncConfig {
|
||
|
||
@Bean("eventExecutor")
|
||
public Executor eventExecutor() {
|
||
ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor();
|
||
executor.setCorePoolSize(5);
|
||
executor.setMaxPoolSize(10);
|
||
executor.setQueueCapacity(100);
|
||
executor.setThreadNamePrefix("event-");
|
||
executor.initialize();
|
||
return executor;
|
||
}
|
||
}
|
||
```
|
||
|
||
### 6.1.4 批量操作优化
|
||
|
||
```java
|
||
// 批量入队
|
||
public void batchEnqueue(List<OrderQueueDTO> queueList) {
|
||
// MySQL 批量插入
|
||
orderQueueMapper.insertBatch(queueList);
|
||
|
||
// Redis 批量写入
|
||
CompletableFuture.runAsync(() -> {
|
||
redisQueueService.batchEnqueue(queueList);
|
||
});
|
||
}
|
||
```
|
||
|
||
---
|
||
|
||
## 6.2 并发控制
|
||
|
||
### 6.2.1 分布式锁
|
||
|
||
**Redis 分布式锁**：
|
||
|
||
```java
|
||
@Service
|
||
public class OrderQueueService {
|
||
|
||
public void enqueue(Long orderId, Long userId) {
|
||
String lockKey = "queue:lock:" + userId;
|
||
|
||
// 获取分布式锁
|
||
Boolean lock = redisTemplate.opsForValue()
|
||
.setIfAbsent(lockKey, "1", 10, TimeUnit.SECONDS);
|
||
|
||
if (Boolean.TRUE.equals(lock)) {
|
||
try {
|
||
// 执行入队操作
|
||
doEnqueue(orderId, userId);
|
||
} finally {
|
||
// 释放锁
|
||
redisTemplate.delete(lockKey);
|
||
}
|
||
} else {
|
||
throw new ServiceException("系统繁忙，请稍后重试");
|
||
}
|
||
}
|
||
}
|
||
```
|
||
|
||
### 6.2.2 乐观锁
|
||
|
||
```java
|
||
@TableName("ops_order")
|
||
public class OpsOrderDO {
|
||
private Long id;
|
||
private String status;
|
||
|
||
@Version // 乐观锁字段
|
||
private Integer version;
|
||
}
|
||
|
||
// 更新时自动检查版本号
|
||
orderMapper.updateById(order); // 如果 version 不匹配，更新失败
|
||
```
|
||
|
||
---
|
||
|
||
## 6.3 可靠性保障
|
||
|
||
### 6.3.1 Redis + MySQL 双写策略
|
||
|
||
**写入流程**：
|
||
|
||
```java
|
||
@Transactional(rollbackFor = Exception.class)
|
||
public Long enqueue(OrderQueueDTO dto) {
|
||
// 1. 先写 MySQL（保证数据不丢失）
|
||
Long queueId = orderQueueMapper.insert(dto);
|
||
|
||
// 2. 异步写 Redis（失败不影响主流程）
|
||
CompletableFuture.runAsync(() -> {
|
||
try {
|
||
redisQueueService.enqueue(dto);
|
||
} catch (Exception e) {
|
||
log.error("Redis 队列写入失败，依赖定时同步任务补偿", e);
|
||
}
|
||
});
|
||
|
||
return queueId;
|
||
}
|
||
```
|
||
|
||
### 6.3.2 定时同步任务
|
||
|
||
```java
|
||
@Scheduled(cron = "0 */5 * * * ?") // 每5分钟同步一次
|
||
public void syncMySQLToRedis() {
|
||
// 查询最近变更的数据
|
||
List<OrderQueueDTO> changedTasks =
|
||
orderQueueMapper.selectChangedAfter(DateUtils.addHours(-1));
|
||
|
||
// 同步到 Redis
|
||
redisQueueService.batchEnqueue(changedTasks);
|
||
|
||
log.info("定时同步完成，同步{}条记录", changedTasks.size());
|
||
}
|
||
```
|
||
|
||
### 6.3.3 故障恢复机制
|
||
|
||
**Redis 宕机降级**：
|
||
|
||
```java
|
||
public List<OrderQueueDTO> getTasksByUserId(Long userId) {
|
||
try {
|
||
// 1. 优先从 Redis 获取
|
||
List<OrderQueueDTO> redisTasks = redisQueueService.getTasksByUserId(userId);
|
||
if (redisTasks != null && !redisTasks.isEmpty()) {
|
||
return redisTasks;
|
||
}
|
||
} catch (Exception e) {
|
||
log.error("Redis 查询失败，降级到 MySQL", e);
|
||
}
|
||
|
||
// 2. Redis 失败，从 MySQL 获取
|
||
return orderQueueMapper.selectListByUserId(userId);
|
||
}
|
||
```
|
||
|
||
---
|
||
|
||
## 6.4 监控与告警
|
||
|
||
### 6.4.1 关键指标监控
|
||
|
||
**系统指标**：
|
||
- QPS（每秒请求数）
|
||
- 响应时间（P50、P95、P99）
|
||
- 错误率
|
||
- 线程池使用率
|
||
|
||
**业务指标**：
|
||
- 工单创建速率
|
||
- 派单成功率
|
||
- 平均响应时长
|
||
- 队列积压数量
|
||
|
||
### 6.4.2 告警规则
|
||
|
||
| 告警项 | 触发条件 | 级别 | 处理方式 |
|
||
|--------|---------|------|----------|
|
||
| P0 工单超时 | 超时未接单 > 3 分钟 | P0 | 钉钉 + 短信 |
|
||
| 队列积压 | 某区域积压 > 10 个 | P1 | 钉钉通知 |
|
||
| 派单失败率 | 失败率 > 5% | P1 | 钉钉通知 |
|
||
| API 错误率 | 错误率 > 1% | P2 | 邮件通知 |
|
||
|
||
---
|
||
|
||
## 6.5 容灾设计
|
||
|
||
### 6.5.1 服务降级
|
||
|
||
```java
|
||
@Service
|
||
public class DispatchEngineService {
|
||
|
||
@Resource
|
||
private DispatchEngine dispatchEngine;
|
||
|
||
public DispatchResult dispatch(DispatchContext context) {
|
||
try {
|
||
// 正常派单流程
|
||
return dispatchEngine.dispatch(context);
|
||
} catch (Exception e) {
|
||
log.error("自动派单失败，降级为手动派单", e);
|
||
|
||
// 降级方案：创建待分配工单
|
||
return createPendingOrder(context);
|
||
}
|
||
}
|
||
}
|
||
```
|
||
|
||
### 6.5.2 数据备份
|
||
|
||
**MySQL 备份**：
|
||
```bash
|
||
# 每日备份
|
||
mysqldump -h localhost -u root -p aiot_ops > backup_$(date +%Y%m%d).sql
|
||
```
|
||
|
||
**Redis 持久化**：
|
||
```bash
|
||
# redis.conf
|
||
save 900 1 # 900秒内至少1个key变化，执行BGSAVE
|
||
save 300 10 # 300秒内至少10个key变化
|
||
save 60 10000 # 60秒内至少10000个key变化
|
||
|
||
appendonly yes # 开启 AOF
|
||
appendfsync everysec # 每秒同步
|
||
```
|
||
|
||
---
|
||
|
||
**下一章**：[Part 7: 扩展性设计](./part7-扩展性设计.md)
|