1. 为什么我们需要处理WebSocket离线消息在企业级应用中关键通知的可靠传递直接影响业务连续性和用户体验。想象一下支付成功通知、系统告警或即时通讯消息因为用户短暂离线而丢失的场景——这可能导致客户投诉、订单纠纷甚至财务损失。WebSocket作为HTML5标准协议相比传统HTTP轮询具有显著优势全双工通信服务端可以主动推送低延迟建立连接后无需重复握手高效性头部信息只有2-10字节但原生WebSocket存在一个致命缺陷当用户网络中断或关闭浏览器时服务端无法感知连接断开导致推送消息石沉大海。我们团队曾因此损失过重要客户——他们的运维人员因未及时收到服务器宕机通知导致业务中断3小时。2. SpringBoot中的WebSocket增强方案2.1 基础配置与心跳检测首先在SpringBoot中启用STOMP协议支持Configuration EnableWebSocketMessageBroker public class WebSocketConfig implements WebSocketMessageBrokerConfigurer { Override public void configureMessageBroker(MessageBrokerRegistry config) { config.enableSimpleBroker(/topic); config.setApplicationDestinationPrefixes(/app); } Override public void registerStompEndpoints(StompEndpointRegistry registry) { registry.addEndpoint(/ws) .setAllowedOriginPatterns(*) .withSockJS(); // 关键启用SockJS回退方案 } Override public void configureWebSocketTransport(WebSocketTransportRegistration registry) { // 设置心跳间隔单位毫秒 registry.setSendTimeLimit(15 * 1000) .setSendBufferSizeLimit(512 * 1024); } }关键配置项说明withSockJS()为不支持WebSocket的浏览器提供降级方案心跳检测通过setSendTimeLimit设置15秒发送超时消息缓冲区限制单个消息不超过512KB2.2 连接状态监听实现创建事件监听器捕获连接状态变化Component public class WebSocketEventListener implements ApplicationListenerAbstractSubProtocolEvent { private static final Logger log LoggerFactory.getLogger(WebSocketEventListener.class); Override public void onApplicationEvent(AbstractSubProtocolEvent event) { if (event instanceof SessionConnectEvent) { String sessionId ((SessionConnectEvent) event).getMessage().getHeaders().get(simpSessionId).toString(); log.info(客户端连接建立: {}, sessionId); } else if (event instanceof SessionDisconnectEvent) { String sessionId ((SessionDisconnectEvent) event).getSessionId(); String reason ((SessionDisconnectEvent) event).getCloseStatus().getReason(); log.warn(客户端断开连接: {}, 原因: {}, sessionId, reason); // 触发离线消息处理 handleOfflineMessage(sessionId); } } private void handleOfflineMessage(String sessionId) { // 实际业务中应查询该session对应的用户ID String userId sessionUserMap.get(sessionId); if(userId ! null) { ListMessage pendingMessages messageService.getPendingMessages(userId); if(!pendingMessages.isEmpty()) { // 进入离线消息处理流程 offlineMessageProcessor.process(userId, pendingMessages); } } } }3. 离线消息存储与补发机制3.1 消息持久化设计建议采用三级存储策略存储层级介质选择保留时间适用场景一级缓存Redis5分钟高频访问的近期消息二级存储MongoDB7天结构化消息主体三级归档文件系统30天审计合规需求消息实体示例Data Document(collection offline_messages) public class OfflineMessage { Id private String id; private String userId; // 目标用户ID private String sessionId; // 最后活跃会话ID private MessageType type; // 消息类型 private String content; // 消息内容JSON格式 private MessageStatus status MessageStatus.PENDING; CreatedDate private LocalDateTime createTime; private LocalDateTime deliverTime; public enum MessageStatus { PENDING, DELIVERED, FAILED } }3.2 补发策略实现基于Spring的定时任务实现分级重试Slf4j Component public class MessageRedeliveryScheduler { Autowired private MessageRepository messageRepo; Autowired private SimpMessagingTemplate messagingTemplate; // 初始延迟5秒之后每30秒执行 Scheduled(initialDelay 5000, fixedRate 30000) public void redeliverPendingMessages() { LocalDateTime cutoffTime LocalDateTime.now().minusMinutes(5); messageRepo.findByStatusAndCreateTimeBefore( MessageStatus.PENDING, cutoffTime ).forEach(msg - { try { messagingTemplate.convertAndSendToUser( msg.getUserId(), /queue/offline, msg.getContent() ); msg.setStatus(MessageStatus.DELIVERED); msg.setDeliverTime(LocalDateTime.now()); messageRepo.save(msg); } catch (Exception e) { log.error(消息重发失败: {}, msg.getId(), e); if(msg.getRetryCount() 3) { msg.setStatus(MessageStatus.FAILED); messageRepo.save(msg); } } }); } }关键参数说明首次重试延迟网络闪断恢复通常需要5秒内最大重试次数3次避免无限循环消息超时5分钟未读转存长期存储4. 生产环境中的性能优化4.1 连接管理优化使用WebSocket连接池避免重复创建Bean public WebSocketClient webSocketClient() { ListTransport transports new ArrayList(); transports.add(new WebSocketTransport(new StandardWebSocketClient())); transports.add(new RestTemplateXhrTransport()); SockJsClient sockJsClient new SockJsClient(transports); sockJsClient.setHttpHeaderNames(X-Requested-With); // 连接池配置 sockJsClient.setDisconnectTimeout(30_000); sockJsClient.setMessageCodec(new Jackson2SockJsMessageCodec()); return sockJsClient; }4.2 消息压缩配置在application.properties中启用消息压缩# 启用消息压缩 spring.websocket.compression.enabledtrue # 消息大小超过2KB时压缩 spring.websocket.compression.message-size-threshold2048 # 压缩缓冲区8KB spring.websocket.compression.buffer-size81924.3 负载测试指标参考使用JMeter压测得到的基准数据单节点4核8G并发连接数平均延迟吞吐量CPU占用1,00028ms1,200 msg/s45%5,00053ms3,800 msg/s78%10,000217ms5,200 msg/s92%当连接数超过5000时建议启用集群模式使用Redis作为消息代理考虑引入Kafka削峰5. 常见问题排查指南5.1 连接不稳定问题现象客户端频繁断开重连排查步骤检查Nginx配置proxy_connect_timeout 7d; proxy_send_timeout 7d; proxy_read_timeout 7d; proxy_http_version 1.1; proxy_set_header Upgrade $http_upgrade; proxy_set_header Connection upgrade;验证心跳配置是否生效检查防火墙是否拦截WebSocket帧5.2 消息堆积问题现象Redis内存持续增长解决方案// 在消息存储时添加TTL Bean public RedisTemplateString, OfflineMessage redisTemplate(RedisConnectionFactory factory) { RedisTemplateString, OfflineMessage template new RedisTemplate(); template.setConnectionFactory(factory); template.setDefaultSerializer(new Jackson2JsonRedisSerializer(OfflineMessage.class)); template.setEnableDefaultSerializer(true); template.setKeySerializer(new StringRedisSerializer()); template.setValueSerializer(new GenericJackson2JsonRedisSerializer()); template.afterPropertiesSet(); // 设置全局过期时间 template.expire(offline:msg:*, 1, TimeUnit.HOURS); return template; }5.3 集群环境下的会话同步使用Spring Session实现分布式会话管理Configuration EnableRedisHttpSession public class SessionConfig { Bean public RedisSerializerObject springSessionDefaultRedisSerializer() { return new GenericJackson2JsonRedisSerializer(); } Bean public CookieSerializer cookieSerializer() { DefaultCookieSerializer serializer new DefaultCookieSerializer(); serializer.setCookieName(JSESSIONID); serializer.setCookiePath(/); serializer.setDomainNamePattern(^.?\\.(\\w\\.[a-z])$); return serializer; } }6. 进阶消息可靠投递保障6.1 事务型消息处理结合本地事务表确保消息不丢失CREATE TABLE message_transaction ( id VARCHAR(36) PRIMARY KEY, business_id VARCHAR(64) NOT NULL, content TEXT NOT NULL, status ENUM(PENDING,PROCESSED) NOT NULL, created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP, INDEX idx_business (business_id) );Spring事务管理器配置Transactional public void processWithTransaction(Message message) { // 1. 业务处理 orderService.createOrder(message); // 2. 记录事务状态 transactionRepo.save( new MessageTransaction( message.getId(), ORDER_CREATED, PROCESSED ) ); // 3. 发送WebSocket通知 messagingTemplate.convertAndSend( /topic/orders, new OrderEvent(message) ); }6.2 终端状态同步方案对于移动端实现状态同步RestController RequestMapping(/api/push) public class PushStateController { GetMapping(/ack/{messageId}) public ResponseEntity? acknowledgeMessage( PathVariable String messageId, RequestHeader(X-Device-ID) String deviceId) { OptionalOfflineMessage message messageRepo.findById(messageId); if (message.isPresent()) { message.get().setStatus(MessageStatus.DELIVERED); message.get().setDeliverTime(LocalDateTime.now()); messageRepo.save(message.get()); // 更新设备最后活跃时间 deviceService.updateLastActive(deviceId); return ResponseEntity.ok().build(); } return ResponseEntity.notFound().build(); } }在Android端实现消息确认fun acknowledgeMessage(messageId: String) { val deviceId Settings.Secure.getString( context.contentResolver, Settings.Secure.ANDROID_ID ) RetrofitClient.instance.pushService .acknowledgeMessage(messageId, deviceId) .enqueue(object : CallbackVoid { override fun onResponse(call: CallVoid, response: ResponseVoid) { if (response.isSuccessful) { Log.d(Push, Message $messageId acknowledged) } } override fun onFailure(call: CallVoid, t: Throwable) { Log.e(Push, Ack failed, t) // 加入重试队列 RetryQueue.add(messageId) } }) }7. 监控与告警体系搭建7.1 Prometheus监控指标暴露WebSocket关键指标Configuration public class WebsocketMetrics { Bean public MeterRegistryCustomizerPrometheusMeterRegistry websocketMetricsConfig() { return registry - { Gauge.builder(websocket.sessions.active, () - SimpUserRegistry.getUserCount()) .description(当前活跃WebSocket会话数) .register(registry); Counter.builder(websocket.messages.sent) .description(已发送消息总数) .tag(direction, outbound) .register(registry); Timer.builder(websocket.message.processing.time) .description(消息处理耗时) .publishPercentiles(0.5, 0.95, 0.99) .register(registry); }; } }7.2 关键告警规则示例Grafana告警配置建议紧急告警P0连续5分钟会话断开率 20%消息积压量超过10,000条警告级别P1平均消息延迟 500ms节点内存使用率 80%告警通知应包含以下关键信息{ alert_name: WebSocket_High_Disconnect_Rate, severity: critical, current_value: 34.7%, threshold: 20%, occur_time: 2023-08-20T14:32:45Z, affected_services: [order-service, notification-service], troubleshooting: 检查网络负载均衡器状态验证Redis连接池配置 }8. 实际案例电商订单状态通知某跨境电商平台接入离线消息方案后的效果对比指标改造前改造后提升幅度通知到达率82%99.97%17.97%客诉率6.2%1.1%-82.3%支付超时率12%3%-75%服务器负载峰值CPU 85%峰值CPU 62%-27%核心实现代码片段public class OrderStatusNotifier { Autowired private SimpMessagingTemplate messagingTemplate; Autowired private OfflineMessageService offlineService; TransactionalEventListener(phase TransactionPhase.AFTER_COMMIT) public void onOrderEvent(OrderStatusChangedEvent event) { String userId event.getUserId(); String destination /user/ userId /queue/order-updates; try { messagingTemplate.convertAndSend( destination, new OrderStatusMessage(event) ); } catch (MessageDeliveryException e) { // 记录离线消息 offlineService.saveOfflineMessage( userId, ORDER_UPDATE, objectMapper.writeValueAsString(event) ); } } }这个案例中我们特别处理了事务提交后发送通知避免脏读使用TransactionalEventListener确保消息与数据库事务一致自动降级到离线存储的异常处理机制
