1. ZLMediaKit快速入门与环境搭建第一次接触ZLMediaKit时我被它的轻量级和高性能所吸引。作为一款开源的流媒体服务器它支持RTSP、RTMP、HLS等多种协议特别适合中小型视频项目的快速部署。记得当时为了测试性能我在一台2核4G的云服务器上同时推了20路720P视频流CPU占用居然不到30%安装过程比想象中简单很多。如果你是Ubuntu用户直接运行官方的一键安装脚本就行# 安装依赖 sudo apt-get install build-essential cmake # 克隆仓库 git clone --depth 1 https://github.com/ZLMediaKit/ZLMediaKit.git # 编译安装 cd ZLMediaKit ./build.shWindows用户可以用VS2019打开工程直接编译。不过在实际项目中我更推荐用Docker部署特别是需要集群化的时候docker run -d -p 1935:1935 -p 8080:8080 \ -e ZLM_SECRET_KEYyour_password \ zlmediakit/zlmediakit:latest配置文件config.ini藏在安装目录的conf文件夹里有几个关键参数需要特别注意[api]下的secret要记牢这是调用API的钥匙[hook]里的admin_params建议改成复杂字符串[http]下的port别跟现有服务冲突踩坑提醒第一次启动时我遇到端口占用问题用netstat -tulnp查了下发现是Nginx占用了8080端口。解决方法要么改ZLMediaKit的http端口要么停掉Nginx服务。2. 核心API实战解析ZLMediaKit的API设计非常RESTful所有接口都通过HTTP调用。我习惯用Postman先测试接口再写到代码里。这里分享几个最常用的API2.1 流代理管理添加拉流代理时新手常犯的三个错误忘记传vhost参数可以用__defaultVhost__url格式错误必须包含rtsp://或rtmp://前缀没处理返回的key后续操作都要用到实战示例代码public JSONObject addStreamProxy(String streamId, String sourceUrl) { MapString, Object params new HashMap(); params.put(vhost, __defaultVhost__); params.put(app, live); params.put(stream, streamId); params.put(url, sourceUrl); JSONObject result zlMediaKit.sendPost(/addStreamProxy, params); if(result.getInteger(code) 0) { String key result.getString(key); // 记得存储这个key到数据库 redisTemplate.opsForValue().set(stream:streamId, key); } return result; }2.2 播放地址生成不同协议生成的播放地址格式不同HLS.m3u8后缀FLV.flv后缀TS.ts后缀我封装了一个智能生成方法public String generatePlayUrl(String streamId, Protocol protocol) { String suffix switch(protocol) { case HLS - m3u8; case FLV - flv; case TS - ts; default - throw new IllegalArgumentException(不支持的协议类型); }; return String.format(http://%s:%s/%s/%s.%s, ip, port, app, streamId, suffix); }3. SpringBoot深度集成3.1 配置自动化在application.yml里我习惯这样配置zlmediakit: server: ip: 192.168.1.100 port: 8080 secret: your_secret_key timeout: 5000 stream: default-app: live default-vhost: __defaultVhost__然后用ConfigurationProperties自动绑定Getter Setter ConfigurationProperties(prefix zlmediakit) public class ZLMediaKitProperties { private Server server; private Stream stream; Getter Setter public static class Server { private String ip; private Integer port; private String secret; private Integer timeout; } Getter Setter public static class Stream { private String defaultApp; private String defaultVhost; } }3.2 连接池优化OkHttpClient的配置直接影响性能这是我的优化方案Bean public OkHttpClient zlMediaKitClient(ZLMediaKitProperties properties) { return new OkHttpClient.Builder() .connectTimeout(properties.getServer().getTimeout(), TimeUnit.MILLISECONDS) .readTimeout(properties.getServer().getTimeout(), TimeUnit.MILLISECONDS) .connectionPool(new ConnectionPool(32, 5, TimeUnit.MINUTES)) .retryOnConnectionFailure(true) .addInterceptor(new RetryInterceptor(3)) .build(); }其中RetryInterceptor是我自定义的重试拦截器public class RetryInterceptor implements Interceptor { private final int maxRetries; public RetryInterceptor(int maxRetries) { this.maxRetries maxRetries; } Override public Response intercept(Chain chain) throws IOException { Request request chain.request(); Response response null; IOException exception null; for (int i 0; i maxRetries; i) { try { response chain.proceed(request); if (response.isSuccessful()) { return response; } } catch (IOException e) { exception e; } } if (exception ! null) throw exception; return response; } }4. 生产环境实战技巧4.1 流状态监控通过定时调用/getMediaList接口可以实现流状态监控Scheduled(fixedRate 30000) public void monitorStreams() { JSONObject result zlMediaKit.sendGet(/getMediaList); ListStreamInfo activeStreams parseStreams(result); activeStreams.forEach(stream - { if (stream.getAliveSecond() 3600) { log.warn(流 {} 已持续 {} 秒, stream.getStreamId(), stream.getAliveSecond()); } }); }4.2 自动清理空闲流结合Spring的定时任务可以实现自动清理Scheduled(cron 0 0 3 * * ?) public void cleanupIdleStreams() { JSONObject result zlMediaKit.sendGet(/getMediaList); ListStreamInfo streams parseStreams(result); streams.stream() .filter(stream - stream.getAliveSecond() 86400) .forEach(stream - { String key redisTemplate.opsForValue().get(stream:stream.getStreamId()); if (key ! null) { zlMediaKit.delStreamProxy(key); } }); }4.3 负载均衡策略当单节点压力过大时我采用DNS轮询健康检查的方案部署多个ZLMediaKit节点用Nginx做负载均衡通过/getStat接口实现健康检查Nginx配置示例upstream zlm_servers { server 192.168.1.101:8080; server 192.168.1.102:8080; server 192.168.1.103:8080; } server { location /api/health { proxy_pass http://zlm_servers/index/api/getStat; health_check; } }5. 常见问题解决方案5.1 401鉴权失败这个问题困扰了我整整一天根本原因是hook鉴权未关闭请求时缺少secret参数secret与config.ini配置不一致解决方案三步走检查config.ini的[api]部分secret配置确保每次请求都带secret参数用Postman先测试基础接口5.2 流延迟过高遇到直播延迟超过5秒的情况可以尝试调整播放协议FLV通常比HLS快修改config.ini的[rtmp]项timeout_ms开启TCP_NODELAY// 在创建OkHttpClient时加入 SocketFactory socketFactory new SocketFactory() { Override public Socket createSocket() throws IOException { Socket socket new Socket(); socket.setTcpNoDelay(true); return socket; } }; new OkHttpClient.Builder() .socketFactory(socketFactory) // 其他配置...5.3 内存泄漏排查通过jmap和jstat工具发现内存缓慢增长最终定位到OkHttpResponse未关闭JSON解析器存在循环引用修复方案try (Response response client.newCall(request).execute()) { try (ResponseBody body response.body()) { // 处理响应 } }6. 性能优化实战6.1 API调用优化批量操作时建议使用连接池我设置maxIdleConnections50开启HTTP/2支持添加GZIP压缩new OkHttpClient.Builder() .protocols(Arrays.asList(Protocol.H2_PRIOR_KNOWLEDGE, Protocol.HTTP_1_1)) .addInterceptor(new GzipRequestInterceptor()) // 其他配置...6.2 JVM参数调优生产环境推荐配置-Xms和-Xmx设为相同值避免动态调整使用G1垃圾回收器添加OOM时heapdump参数java -jar your-app.jar \ -Xms4g -Xmx4g \ -XX:UseG1GC \ -XX:HeapDumpOnOutOfMemoryError \ -XX:HeapDumpPath/tmp6.3 异步处理方案对于非实时操作我采用Spring的AsyncAsync(taskExecutor) public CompletableFutureJSONObject asyncAddStream(String streamId, String url) { JSONObject result addStreamProxy(streamId, url); return CompletableFuture.completedFuture(result); } // 配置线程池 Bean(name taskExecutor) public Executor taskExecutor() { ThreadPoolTaskExecutor executor new ThreadPoolTaskExecutor(); executor.setCorePoolSize(10); executor.setMaxPoolSize(20); executor.setQueueCapacity(500); executor.setThreadNamePrefix(ZLMediaKit-Async-); executor.initialize(); return executor; }7. 扩展功能开发7.1 录制功能集成通过/openRtpServer接口实现自动录制public JSONObject startRecording(String streamId, String savePath) { MapString, Object params new HashMap(); params.put(stream_id, streamId); params.put(save_path, savePath); params.put(max_second, 3600); return sendPost(/openRtpServer, params); }7.2 智能流量控制基于QPS的动态限流Aspect Component public class RateLimitAspect { private final RateLimiter rateLimiter RateLimiter.create(50); // 50 QPS Around(execution(* com.example.zlm..*(..))) public Object limit(ProceedingJoinPoint pjp) throws Throwable { if (rateLimiter.tryAcquire()) { return pjp.proceed(); } throw new RuntimeException(API调用过于频繁); } }7.3 安全加固方案接口签名验证IP白名单限制请求频率监控public class SecurityInterceptor implements HandlerInterceptor { Override public boolean preHandle(HttpServletRequest request, HttpServletResponse response, Object handler) { String clientIp getClientIp(request); if (!ipWhitelist.contains(clientIp)) { response.setStatus(403); return false; } String signature request.getHeader(X-Signature); if (!verifySignature(signature)) { response.setStatus(401); return false; } return true; } }