标签：
本章将描述motan部分的特性并对源码进行分析。
1.requestid的维护，使用了当前时间左移20位，再和一个自增变量组合
public class RequestIdGenerator { protected static final AtomicLong offset = new AtomicLong(0); protected static final int BITS = 20; protected static final long MAX_COUNT_PER_MILLIS = 1 << BITS; /** * 获取 requestId * * @return */ public static long getRequestId() { long currentTime = System.currentTimeMillis(); long count = offset.incrementAndGet(); while(count >= MAX_COUNT_PER_MILLIS){ synchronized (RequestIdGenerator.class){ if(offset.get() >= MAX_COUNT_PER_MILLIS){ offset.set(0); } } count = offset.incrementAndGet(); } return (currentTime << BITS) + count; } public static long getRequestIdFromClient() { // TODO 上下文 requestid return 0; } }
2.限流，motan支持简单的限流，是利用filter来实现的
@SpiMeta(name = "active")
@Activation(sequence = 1)
public class ActiveLimitFilter implements Filter {
@Override
public Response filter(Caller<?> caller, Request request) {
int maxAcvitivyCount = caller.getUrl().getIntParameter(URLParamType.actives.getName(), URLParamType.actives.getIntValue());
if (maxAcvitivyCount > 0) {
int activeCount = RpcStats.getServiceStat(caller.getUrl()).getActiveCount();
if (activeCount >= maxAcvitivyCount) {
throw new MotanServiceException(String.format("Request(%s) active count exceed the limit (%s), referer:%s", request,
maxAcvitivyCount, caller.getUrl()), MotanErrorMsgConstant.SERVICE_REJECT);
}
}
long startTime = System.currentTimeMillis();
RpcStats.beforeCall(caller.getUrl(), request);
try {
Response rs = caller.call(request);
RpcStats.afterCall(caller.getUrl(), request, true, System.currentTimeMillis() - startTime);
return rs;
} catch (RuntimeException re) {
RpcStats.afterCall(caller.getUrl(), request, false, System.currentTimeMillis() - startTime);
throw re;
}
}
}
3.对于连续失败的client进行不可用操作
void incrErrorCount() { long count = errorCount.incrementAndGet(); // 如果节点是可用状态，同时当前连续失败的次数超过限制maxClientConnection次，那么把该节点标示为不可用 if (count >= maxClientConnection && state.isAliveState()) { synchronized (this) { count = errorCount.longValue(); if (count >= maxClientConnection && state.isAliveState()) { LoggerUtil.error("NettyClient unavailable Error: url=" + url.getIdentity() + " " + url.getServerPortStr()); state = ChannelState.UNALIVE; } } } } void resetErrorCount() { errorCount.set(0); if (state.isAliveState()) { return; } synchronized (this) { if (state.isAliveState()) { return; } // 如果节点是unalive才进行设置，而如果是 close 或者 uninit，那么直接忽略 if (state.isUnAliveState()) { long count = errorCount.longValue(); // 过程中有其他并发更新errorCount的，因此这里需要进行一次判断 if (count < maxClientConnection) { state = ChannelState.ALIVE; LoggerUtil.info("NettyClient recover available: url=" + url.getIdentity() + " " + url.getServerPortStr()); } } } }
4.支持多注册中心，因此cluster的refer集合是所有注册中心包含服务器的集合，如果同一个服务器在不同的注册中心注册，则cluster中当作两个服务器
5.服务端的采用boss线程池+工作线程池+业务线程池的处理方式
private final static ChannelFactory channelFactory = new NioServerSocketChannelFactory(//boss线程池和工作线程池，主要负责接收消息
Executors.newCachedThreadPool(new DefaultThreadFactory("nettyServerBoss", true)),
Executors.newCachedThreadPool(new DefaultThreadFactory("nettyServerWorker", true)));
private StandardThreadExecutor standardThreadExecutor = null;//业务线程池，负责具体的业务处理
standardThreadExecutor = (standardThreadExecutor != null && !standardThreadExecutor.isShutdown()) ? standardThreadExecutor
: new StandardThreadExecutor(minWorkerThread, maxWorkerThread, workerQueueSize,
new DefaultThreadFactory("NettyServer-" + url.getServerPortStr(), true));
standardThreadExecutor.prestartAllCoreThreads();
final NettyChannelHandler handler = new NettyChannelHandler(NettyServer.this, messageHandler,
standardThreadExecutor);//handler使用业务线程池今天处理具体的业务
标签：
原文地址：http://www.cnblogs.com/mantu/p/5887224.html
